Biomimetics and Marine Technology

JournalVolume 45 Number 4 July/August 2011

The International, Interdisciplinary Society Devoted to Ocean and Marine Engineering, Science, and Policy

Biomimetics and Marine Technology

www.benthos.com

INTRODUCINGThe Next Generation

in Underwater Acoustic Modems

ATM-900 Series

The ATM-900 Series Modems from Teledyne Benthos

Available April 2011See us at Ocean Business - Booth #S1.

A Member of Teledyne Marine49 Edgerton Drive, North Falmouth, MA 02556 USA

T: +1 508.563.1000 • F: +1 508.563.6444E: [email protected]

• Advanced Data Logger Capabilities

– Increased from 704K to 6MB (flash file format)

– Searchable by record number and time

• Dual Serial Port – Interface with more than one sensor at a time (optional)

• Integrated real-time clock for data time stamping

• New Command Structure & Modem Parameters

• Pinger Mode

• Arbitrary Waveform Generator (optional)

8Biomimetics and Marine Technology: An IntroductionFrank E. Fish, Donna M. Kocak

14Biomimicking Marine Mechanisms and Organizational PrinciplesCommentary by Yoseph Bar-Cohen

16Sink and Swim: Clues From Nature for Aquatic RoboticsCommentary by Jeannette Yen

19Developing Bioinspired Autonomous SystemsCommentary by Thomas M. McKenna

24GhostSwimmer™ AUV: Applying Biomimetics to Underwater Robotics for Achievement of Tactical Relevance Commentary by Michael Rufo, Mark Smithers

31Autonomous Robotic Fish as Mobile Sensor Platforms: Challenges and Potential SolutionsXiaobo Tan

41Robotic Models for Studying Undulatory Locomotion in FishesGeorge V. Lauder, Jeanette Lim, Ryan Shelton, Chuck Witt, Erik Anderson, James L. Tangorra

56Thrust Production in Highly Flexible Pectoral Fins: A Computational DissectionSrinivas Ramakrishnan, Meliha Bozkurttas, Rajat Mittal, George V. Lauder

Volume 45, Number 4, July/August 2011

Biomimetics and Marine Technology Guest Editors: Frank E. Fish and Donna M. Kocak

65Learning From the Fins of Ray-Finned Fish for the Propulsors of Unmanned Undersea VehiclesJames L. Tangorra, Timo Gericke, George V. Lauder

74Bioinspired Design Process for an Underwater Flying and Hovering VehicleJason D. Geder, John S. Palmisano, Ravi Ramamurti, Marius Pruessner, Banahalli Ratna, William C. Sandberg

83A Twistable Ionic Polymer-Metal Composite Artificial Muscle for Marine ApplicationsKwang J. Kim, David Pugal, Kam K. Leang

99Batoid Fishes: Inspiration for the Next Generation of Underwater RobotsKeith W. Moored, Frank E. Fish, Trevor H. Kemp, Hilary Bart-Smith

110Bioinspired Propulsion Mechanisms Based on Manta Ray LocomotionKeith W. Moored, Peter A. Dewey, Megan C. Leftwich, Hilary Bart-Smith, Alexander J. Smits

119Inspired by Sharks: A Biomimetic Skeleton for the Flapping, Propulsive Tail of an Aquatic RobotJohn H. Long, Jr., Tom Koob, Justin Schaefer, Adam Summers, Kurt Bantilan, Sindre Grotmol, Marianne Porter

In This IssueTurn to page 7 for a key to the cover images.

The Marine Technology Society Journal (ISSN 0025-3324) is published by the Marine Technology Society, Inc., 5565 Sterrett Place, Suite 108, Columbia, MD 21044.

MTS members can purchase the printed Journal for $27 domestic and $50 (plus $50 S&H) international. Non-members and library subscriptions are $420 online only, $124 print—domestic, $140 (plus $50 S&H) print— international, $435 print and online (worldwide); Single-issue (hardcopy) is $20 plus $7.50 S&H (domestic), $24.50 S&H (international); Pay-per-view (worldwide): $15/article. Postage for periodicals is paid at Columbia, MD, and additional mailing offices.

P O S T M A S T E R :Please send address changes to:

Marine Technology Society Journal5565 Sterrett PlaceSuite 108Columbia, Maryland 21044

Copyright © 2011 Marine Technology Society, Inc.

Text: SPi Cover and Graphics:Michele A. Danoff, Graphics By Design

In This Issue130Lateral-Line-Inspired Sensor Arrays for Navigation and Object IdentificationVicente I. Fernandez, Audrey Maertens, Frank M. Yaul, Jason Dahl, Jeffrey H. Lang, Michael S. Triantafyllou

147A Conserved Neural Circuit-Based Architecture for Ambulatory and Undulatory Biomimetic RobotsJoseph Ayers, Anthony Westphal, Daniel Blustein

153A Hybrid Class Underwater Vehicle: Bioinspired Propulsion, Embedded System, and Acoustic Communication and Localization SystemMichael Krieg, Peter Klein, Robert Hodgkinson, Kamran Mohseni

165Modeling of Artificial Aurelia aurita Bell DeformationKeyur B. Joshi, Alex Villanueva, Colin F. Smith, Shashank Priya

181Swimming and Walking of an Amphibious Robot With Fin ActuatorsNaomi Kato

198Marine Applications of the Biomimetic Humpback Whale Flipper Frank E. Fish, Paul W. Weber, Mark M. Murray, Laurens E. Howle

208Shark Skin Separation Control Mechanisms Amy Lang, Philip Motta, Maria Laura Habegger, Robert Hueter, Farhana Afroz

216Can Biomimicry and Bioinspiration Provide Solutions for Fouling Control? Emily Ralston, Geoffrey Swain

228BOOK REVIEW: Sex, Drugs, and Sea Slime: The Oceans’ Oddest Creatures and Why They Matterby Ellen Prager Reviewed by Jason Goldberg

Deep Tracks!

Teledyne RD Instruments’ new PAVS150 DVL keeps your vehicle on course up to 500 meters above the seafloor.

The 150 kHz Phased Array Velocity Sensor (PAVS150) Doppler Velocity Log delivers over 2x the range of our industry-standard Workhorse Navigator. Now that’s deep tracks!

The PAVS150’s 1000M depth rating, combined with its increased range allows you to submerge your manned or unmanned vehicle sooner and transit longer, saving you time, money, and peace of mind.

Stay the course with PAVS!

Full details at: http://www.rdinstruments.com/PAVS.aspx

Poway, CA USA • Tel. +1-858-842-2600www.rdinstruments.com

Learn how MTS can help your company reach the technology leaders and innovators

in the marine technology industry.

Opportunities include advertising in Currents, the MTS Journal,

sponsorships, online and custom-tailored programs.

MTS — here for the marine technology community,

here for you!

Email Mary Beth Loutinsky at [email protected] or

call 703-629-3810

BOA RD OF D IREC T ORSPresidentJerry BoatmanQinetiQ North America – Technology SolutionsGroupPresident-electDrew MichelROV Technologies, Inc.Immediate Past PresidentElizabeth CorbinVP—Section AffairsLisa MedeirosOceanWorks InternationalVP—Education and ResearchJill ZandeMATE CenterVP—Industry and TechnologyJerry C. WilsonFugro Pelagos, Inc.VP—PublicationsKarin LynnTreasurer and VP—Budget and FinanceDebra KillInternational Submarine EngineeringVP—Government and Public AffairsJustin ManleyLiquid Robotics

SEC T IONSCanadian MaritimeVacantFloridaVacantGulf CoastLaurie JuganConsultantHampton RoadsRaymond TollSAICHawaiiStewart BurleyStrategic Theories UnlimitedHoustonRobert KeithPhoenix International Holdings, Inc.JapanProf. Toshitsugu SakouTokai UniversityMontereyJill ZandeMATE CenterNew EnglandChris JakubiakUMASS Dartmouth-SMASTNewfoundland and LabradorBill O’KeefeSurmount Technologies, Inc.OregonTBDPuget SoundFritz StahrUniversity of WashingtonSan DiegoScott MauSouthwest Fisheries Sciences CenterSouth KoreaDr. Seok Won HongMaritime & Ocean Engineering Research Inst.(MOERI/KORDI)Washington, D.C.Brent EversHadal Technologies, Inc.

P ROF E S SION A L C OMMI T T EE S

Industry and TechnologyBuoy TechnologyDr. Walter PaulWoods Hole Oceanographic InstitutionCables and ConnectorsHelmut H. PortmannNational Data Buoy CenterDeepwater Field Development TechnologyDr. Benton BaughRadoil, Inc.DivingDavid C. BerrySubsea Construction and Diving ConsultantDynamic PositioningHoward ShattoShatto EngineeringManned Underwater VehiclesWilliam KohnenSEAmagine Hydrospace CorporationMooringsJack RowleySAICOceanographic InstrumentationDr. Jim IrishUniversity of New HampshireOffshore StructuresDr. Peter W. MarshallMHP Systems EngineeringRemotely Operated VehiclesDrew MichelROV Technologies, Inc.Renewable EnergyRich ChwaszczewskiSAICRopes and Tension MembersEvan ZimmermanDelmar Systems, Inc.Seafloor EngineeringHerb Herrmann Naval Seafloor Cable Protection OfficeUnderwater ImagingDr. Fraser DalgleishHarbor Branch Oceanographic InstituteUnmanned Maritime VehiclesRafael MandujanoVehicle Control Technologies, Inc.

Education and ResearchMarine ArchaeologyDan WarrenC & C TechnologiesMarine EducationErica MoultonMATE CenterMarine Geodetic Information SystemsDave ZilkoskiNOAAMarine MaterialsVacantOcean ExplorationGuillermo SöhnleinOceanGatePhysical Oceanography/MeteorologyDr. Richard L. CroutNational Data Buoy CenterRemote SensingHerb RipleyHyperspectral Imaging Limited

Government and Public AffairsMarine Law and PolicyMontserrat Gorina-YsernHealthy Children–Healthy Oceans FoundationMarine Mineral ResourcesDr. John C. WiltshireUniversity of HawaiiMarine SecurityDallas MeggittSound & Sea TechnologyOcean Economic PotentialJames MarshUniversity of HawaiiOcean Observing SystemsDonna KocakHARRIS CapRock CommunicationsOcean PollutionJacob SobinNOAA Coastal Services Center

S T UDEN T SEC T IONSDuke UniversityCounselor: Douglas Nowacek, Ph.D.Florida Atlantic UniversityCounselor: Douglas A. Briggs, Ph.D.Florida Institute of TechnologyCounselor: Stephen Wood, Ph.D., P.E.Long Beach City CollegeCounselor: Scott FraserMassachusetts Institute of TechnologyCounselor: Alexandra Techet, Ph.D.Monterey Peninsula College/Hartnell CollegeCounselor: Jeremy R. HertzbergTexas A&M University—College StationCounselor: Patrick LynettTexas A&M—Corpus ChristiCounselor: Lea-Der Chen, Ph.D.Texas A&M University—GalvestonCounselor: Frank Warnakula, Ph.D.United States Naval AcademyCounselors: Capt. Joseph T. Arcano (USN Ret), Ph.D.

Cmdr. David J. RobillardUniversity of HawaiiCounselor: R. Cengiz Ertekin, Ph.D.University of HoustonCounselors: Raresh Pascali, P.E., Chuck RichardsUniversity of North Carolina—CharlotteCounselor: James Conrad, Ph.D.University of Southern MississippiCounselor: Stephen Howden, Ph.D. Webb InstituteCounselor: Matthew Werner

HONOR A RY MEMBERS†Robert B. Abel†Charles H. BussmannJohn C. Calhoun, Jr.John P. Craven†Paul M. FyeDavid S. Potter†Athelstan Spilhaus†E. C. Stephan†Allyn C. Vine†James H. Wakelin, Jr.†deceased

Marine Technology Society Officers

Brian Bingham, Ph.D.EditorUniversity of Hawaii at Manoa

Corey JaskolskiHydro Technologies

Donna KocakHARRIS CapRock Communications

Scott Kraus, Ph.D.New England Aquarium

Dhugal Lindsay, Ph.D.Japan Agency for Marine-Earth Science & Technology

Justin ManleyLiquid Robotics

Stephanie ShowalterNational Sea Grant Law Center

Jason StanleySchilling Robotics

Edith Widder, Ph.D.Ocean Research and Conservation Association

Jill ZandeMATE Center

EditorialKarin LynnVP of Publications

Brian Bingham, Ph.D.Editor

Amy MorganteManaging Editor

AdministrationJerry BoatmanPresident

Richard LawsonExecutive Director

Jeanne GloverMembership and Marketing Manager

Michael HallMember Groups Manager

Chris BarrettDirector of Professional Development and Meetings

Suzanne VoelkerSubscription Manager

Editorial BoardThe Marine Technology Society is a not-for-profit, international professional society. Established in 1963, the Society’s mission is to promote the exchange of information in ocean and marine engineer-ing, technology, science, and policy.

Please send all correspondence to:The Marine Technology Society5565 Sterrett Place, Suite 108Columbia, MD 21044(410) 884-5330 Tel.(410) 884-9060 FAXMTS Journal: [email protected]: [email protected]: [email protected]: [email protected]: [email protected]: www.mtsociety.org

MEMBERSHIP INFORMATIONmay be obtained by contacting the Marine Technology Society. Benefits include: Free subscription to the online Marine

Technology Society Journal, with highly reduced rates for the paper version

Free subscription to the bimonthly news -let ter, Currents, covering events, business news, science and technology, and people in marine technology

Member discounts on all MTS publications Reduced registration rates to all MTS and

MTS-sponsored conferences and workshops Member-only access to an expansive Job

Bank and Member Directory Reduced advertising rates in MTS

publications National recognition through our Awards

ProgramIndividual dues are $75 per year. Life membership is available for a one-time fee of $1,000. Patron, Student, Emeritus, Institutional, Business, and Corporate memberships are also available.

ADVERTISINGAdvertising is accepted by the Marine Tech nology Society Journal. For more informa-tion on MTS advertising and policy, please contact Mary Beth Loutinsky, [email protected].

COPYRIGHTCopyright © 2011 by the Marine Technology Society, Inc. Authorization to photocopy items for internal or personal use, or the internal or personal use of specific clients, is granted by the Marine Technology Society, provided that the base fee of $1.00 per copy, plus .20 per page is paid directly to Copyright Clearance Center, 222 Rosewood Dr., Danvers, MA 01923.

For those organizations that have been granted a photocopy license by CCC, a sepa-rate sys tem of payment has been arranged. The fee code for users of the Transactional Reporting Service is 0025-3324/89 $1.00 + .20. Papers by U.S Government employees are declared works of the U.S. Government and are therefore in the public domain.

The Marine Technology Society cannot be held responsible for the opinions given and the statements made in any of the articles published.

ABSTRACTSMTS Journal article abstracts, if available, can be accessed for free at http://www.ingentaconnect.com/content/mts/mtsj.

Print and electronic abstracts of MTS Journal articles are also available through GeoRef <http://www.agiweb.org/georef/>, Aquatic Sciences and Fisheries Abstracts, published by Cambridge Scientific Abstracts <http://www.csa.com/factsheets/aquclust-set-c.php>, and Geobase’s Oceanbase published by Elsevier Science.

CONTRIBUTORSContributors can obtain an information and style sheet by contacting the managing editor. Sub missions that are relevant to the concerns of the Society are welcome. All papers are sub-jected to a stringent review procedure directed by the editor and the editorial board. The Journal focuses on technical material that may not otherwise be available, and thus technical papers and notes that have not been published previously are given priority. General commen-taries are also accepted and are subject to review and approval by the editorial board.

6 Marine Technology Society Journal

Front cover: GhostSwimmer AUVs from Boston Engineering Advanced Systems Group and Olin College Intelligent Vehicle Lab. Image courtesy of HARRIS CapRock Communication; underwater photo courtesy Dr. Tamara Frank.

Back cover: (Image courtesy of HARRIS CapRock Communications)1. Propeller with tubercles (propeller courtesy of Laurens Howls and morphing courtesy of John A. Lever)2. CephaloBot prototype hybrid vehicle3. Robotic Turtle, “RT-I” 4. Mantabot

5. Artificial Aurelia aurita bell deformation (photo courtesy of Alex Villanueva, CIMSS, Virginia Tech)6. Autonomous robotic fish (photo courtesy of Xiaobo Tan)7. Bluegill Sunfish (Lepomis macrochirus) with robotic pectoral fins (photo courtesy of George Lauder and James Tangorra)8. Four-fin vehicle based on Bird wrasse (Gomphosus varius)9. Tadro4 modeled after living electric ray Narcine (photo courtesy of Dr. Steve Kajiura)10. Lobster-based robot (robot photo courtesy of Brian Tucker, Bresnahan Photography; lobster photo courtesy of Dr. Tamara Frank)

Key to Cover Images

of

I N T R O D U C T I O N

Biomimetics and Marine Technology:An Introduction

Frank E. FishWest Chester University

Donna M. KocakHARRIS CapRock Communications

I nspiration for the development of new technologies is at the heart of the biomimeticapproach. As there is a wide diversity of biological forms, particular attributes can be tar-geted that provide innovative solutions to engineering problems. Biology can provide newtechnological possibilities and enhance performance of existing technologies. Biomimicry isa tool for solving problems in the conceptual and embodiment phases of design (Reapet al., 2005). The goal of biomimetics is to use biological inspiration to engineer machinesthat emulate the performance of animals (Kumph & Triantafyllou, 1998; Taubes, 2000),particularly in instances where the animal’s performance exceeds current technology. Thenatural experimentation that has occurred through the evolutionary process has producedthe plethora of organisms, both living and extinct. Within the phylogenetic lineages ofthese organisms, there has been essentially a “cost-benefit analysis” where particular designsfor specific functions have been optimized to perform with respect to the rigors of theirenvironment (Fish, 2006).

Biologists are well acquainted with the specific adaptations present in animals, which maybe of interest to engineers. For biologists, an adaptationist program has allowed for the iden-tification of novel features of organisms based on engineering principles, whereas for engi-neers, identification of such novel features is necessary to exploit them for biomimeticdevelopment. This new synergy between biologists and engineers can be beneficial in advanc-ing technology by looking to nature to provide solutions to current problems.

For marine technologies, the biomimetic approach particularly holds the promise of en-hanced performance and increased efficiency for operation in the aquatic realm. It was in theoceans that life first evolved and where complex animals have thrived for over 600 millionyears. There are marine representatives from every major animal phylum. Marine animalssurvive in environments as diverse as tropical coral reefs, polar ice-capped oceans, and thelightless abyssal depths. The diversity of habitats available in marine systems has led to avast array of body designs and physiological and behavioral mechanisms. These adaptationsthat evolved in animals are used to overcome the biotic and abiotic challenges in the ocean.

To deal with the rigors of the marine environment, animals have developed specializedsensory systems (e.g., echolocation, electroreception), mechanisms to deal with pressure


(e.g., buoyancy control), strategies to economize on energy (e.g., fusiform body design,schooling, burst-and-glide swimming), armor (e.g., bony scales, mollusk shells), stabilitymechanisms (e.g., paired and median fins), maneuverability (e.g., flexible bodies, vectoredthrust), speed (e.g., high-aspect-ratio oscillatory propulsors, jet propulsion), stealth (e.g.,camouflage, low acoustic signature), and use of compliant materials (e.g., collagen, proteinrubbers, mucous). In using such specializations to enhance their own survival, animals at-tempt to function in a manner to minimize their total energy budget while maximizing theperformance of the specialization. Animals are doing the type of optimization that engi-neers seek to incorporate into designs (Vincent, 1990), and specifically, marine animalsare dealing with the very problems that are of concern for marine engineers.

Marine technology is well suited for the application of bioinspired design, as there is a needfor exploitation of the oceans for new sources of food, energy, and minerals. Exploration ofthe world’s oceans is expensive. Ship time, support personnel, maintenance facilities, andenergy costs can be prohibitively costly. Added to these costs are the expanse of the oceansurface (3.6 × 108 km2) to be explored and the dangers associated with working in thedeep-water environment (average depth = 3,650 m). The Challenger Deep (depth =10,902 m; pressure = 16,500 psi or 113,764 kPa) in the Mariana Trench was only visitedonce by a manned vehicle,Trieste, in 1960. Since that time, only two unmanned expeditions,robotic deep-sea probeKaikō andHROVNereus, have returned to the deepest surveyed pointin the ocean.

Historically, marine animals have served as the inspiration for technological design. Dur-ing the Renaissance, animals were identified as streamlined bodies for drag reduction thatcould be applied to manufactured devices. Between 1505 and 1508, Leonardo da Vinciwas particularly interested in flow in water, as revealed in his notebooks, Codex Leicester(Ball, 2009). Da Vinci wrote on the function of streamlined bodies in reducing drag andnoted the streamlined shape of a fish (Anderson, 1998). He argued that the fish couldmove through the water with little resistance because its shape allowed the water to flowsmoothly over the afterbody without prematurely separating. Da Vinci recognized and dem-onstrated a similar design with the hull shape of ships.

Giovanni Borelli in 1680made an examination of the swimming motions of animals withtheir application to submarine technology (Borelli, 1680). In his book De Motu Animalium(The Movement of Animals), Borelli likened swimming to flying in that both were accom-plished by the displacement of fluids, although he noted the differences in density of air andwater and their effects on stability and buoyancy. Borelli described the design of an early sub-marine that incorporated ideas based in part on animals for buoyancy regulation and propul-sion. The submarine would submerge using a hydrostatic mechanism based on the swimbladder of a fish by filling goatskin bags, located inside the submarine, through holes in

July/August 2011 Volume 45 Number 4 9

the sides of the boat. Propulsion would be accomplished by oars projecting through the hulland fitted with watertight seals. When the submarine was on the bottom, it was envisionedthat the oars would push off the sandy substrate to move the boat along. In mid-water, theoars would paddle like the feet of frogs or geese. During the rearward power stroke, a flexiblepaddle at the end of the oar would expand to work on a largemass of fluid.During the forwardrecovery stroke, the paddle would fold passively to reduce the frontal area and drag on the oar.However, Borelli considered that propulsion of the boat would be easier if a flexible oar werepositioned at the stern, emulating the motion of a fish tail. Despite the elaborate design for itstime, it is doubtful if this early biomimetic experiment was successfully used.

Cayley examined the streamlined body shapes of a trout and a dolphin in 1809 as solids ofleast resistance design (Gibbs-Smith, 1962). Cayley unsuccessfully attempted to apply thesenatural designs to the hull of a boat for moving on the water surface (Vogel, 1998). Therounded configuration of the hull was unstable with respect to roll, and low drag did notoccur. The design of fish and dolphins is similar to the optimal shape for drag reductionof submerged bodies, such as modern submarines. These natural swimmers and submarineshave fusiform body shapes with a rounded leading edge and slowly tapering tail. The fore-runner for hulls used by modern nuclear submarines, the USS Albacore, was built in 1953with a fusiform shape.

Both Cayley’s hull and the USS Albacore demonstrate limitations due to a misuse of thebiomimetic approach. For Cayley, strict adherence to copying biological designs withoutproper insight into the function and limitations of those designs proved disastrous (Vogel,1998; Fish, 2006). While the shapes of fish and dolphins are appropriate for movementunderwater, their shapes are not effective at the water surface. Hulls with broad beams andgreater buoyancy, like those displayed by waterfowl, provide enhanced stability and oppor-tunity for greater speed at the water surface (Aigeldinger & Fish, 1995).

In the case of the USS Albacore, the design is only analogous with fish and dolphins.Although the designs are convergent, there was no information exchange to determine design.The Albacore was likened to the shape of a fish, but the submarine’s design was not biolog-ically inspired (Harris 1997; Largess & Mandelblatt, 1999). This was similar to the descrip-tion of the fictional submarine, theNautilus, in Jules Verne’s Twenty Thousand Leagues underthe Sea:

We were lying upon the back of a sort of submarine boat, which appeared (as far as I couldjudge) like a huge fish of steel.

The hull shape of the Albacorewas based on the “Lyon form” of airship models (Largess &Mandelblatt, 1999). Originally, submarine hulls were designed more as surface ships due to


the limited amount of time that they could operate submerged. The streamlined hull of theAlbacore made it the fastest and most maneuverable submarine of its time.

The convergent designs of the Albacore andmarine animals reflect selection, both artificialand natural, respectively, for optimizing similar performance parameters (i.e., speed, drag re-duction, maneuverability). However, similarity of design does not necessarily always translateinto identical performance and assumptions of performance expectations should be ap-proached with caution. Similar shapes can have different functions or have limitations to afunction when viewed in isolation without consideration of the whole system. Despite itsenhanced maneuverability compared to other submarines, the Albacore ’s rate of turn at2° s−1 is poor when compared with similarly shaped dolphins, which can turn at 453° s−1

and in a confined space of 20% of body length (Fish, 2002). The rigid hull of the submarinein concert with yaw control mainly from an aft-positioned rudder limits agility and maneu-verability compared to flexible-bodied dolphins with multiple fore and aft mobile controlsurfaces (Fish, 2002; Fish & Nicastro, 2003).

The biomimetic approach demands first careful observation of the whole biological sys-tem to identify the principles and attributes of the system. Thus, major limitations and con-straints of any biological design can be defined before translation to an engineered system.The association between biologists and engineers becomes paramount as biomimetic technol-ogies are developed.

This special issue of theMarine Technology Society Journal brings together both biologistsand engineers who are currently involved with the development of biomimetic devices forapplications in themarine environment. The intent is to assess the current state of technologyand incite new applications that apply these and other innovative concepts as technology ad-vances. The research presented in this issue mimics specific aspects of fish, skates, rays, sharks,lobsters, jellyfish, squids, and sea turtles. In many of these papers, the goal is focused on lo-comotion or propulsion in robotic counterparts so they can maneuver more efficiently in theanimal’s designed-for environment. This may be acted out either individually or in multiplesas schools (or swarms) of fish. Swimming and walking using amphibious designs are also con-sidered. A recent student competition sponsored by the Office of Naval Research (ONR) washeld for the first time (involving several of the authors in this issue) to evaluate the perfor-mance of “mantabots,” which aspire to perform as gracefully as the sleek and elegant mantarays they attempt to copy (Pennisi, 2011). A commentary byMcKenna provides an overviewof this and other biomimetic work sponsored by ONR. Other papers in this issue identify asingle unique aspect and strive to achieve the same benefit in an engineered device. Simulat-ing the lateral line system found in most aquatic vertebrates using pressure sensors is one ex-ample that can be used to supplement vision and sonar in turbid waters. Deriving a controlmechanism based on shark skin properties to increase vehicle swimming speeds, imitating the


tubercles of whale fins to enhance hydrodynamics, and engineering a bioinspired solutionfor natural antifouling mechanisms are other examples. Whether your interest is in under-water vehicles, imaging, dynamic positioning, marine materials, ocean pollution, remotesensing, oceanographic instrumentation, ocean observing, marine science, or education,this special issue should provide valuable content.

ReferencesAigeldinger, T.L., & Fish, F.E. 1995. Hydroplaning by ducklings: Overcoming limitations to swimming at the

water surface. J Exp Biol. 198:1567-4.

Anderson, J.D. 1998. A History of Aerodynamics. Cambridge: Cambridge University Press.

Ball, P. 2009. Flow. Oxford: Oxford University Press.

Borelli, G.A. 1680. De Motu Animalium Pars (The Movement of Animals, translated by P. Maquet (1989)).

Berlin: Springer-Verlag.

Fish, F.E. 2002. Balancing requirements for stability and maneuverability in cetaceans. Integr Comp Biol.

42:85-93. doi: 10.1093/icb/42.1.85.

Fish, F.E. 2006. Limits of nature and advances of technology in marine systems: What does biomimetics have

to offer to aquatic robots? Appl Bionics Biomech. 3:49-60. doi: 10.1533/abbi.2004.0028.

Fish, F.E., & Nicastro, A.J. 2003. Aquatic turning performance by the whirligig beetle: constraints on

maneuverability by a rigid biological system. J Exp Biol. 206:1649-56. doi: 10.1242/jeb.00305.

Gibbs-Smith, C.H. 1962. Sir George Cayley’s Aeronautics 1796-1855. London: Her Majesty’s Stationery

Office.

Harris, B. 1997. The Navy Times Book of Submarines: A Political, Social, and Military History. New York:

Berkley.

Kumph, J.M., & Triantafyllou, M.S. 1998. A fast-starting and maneuvering vehicle, the ROBOPIKE.

In: Proceedings of the International Symposium on Seawater Drag Reduction, ed. Meng, J.C.S.,

pp. 485-90. Newport, Rhode Island.

Largess, R.P., & Mandelblatt, J.L. 1999. U.S.S. Albacore: Forerunner of the Future. Portsmouth, NH:

The Portsmouth Marine Society.

Pennisi, E. 2011. Manta Machines, 332:28-9, Science, 27 May 2001, www.sciencemag.org. Retrieved on

June 14, 2011.

Reap, J., Baumeister, D., & Bras, B. 2005. Holism, biomimicry and sustainable engineering. In: Proceedings

of IMECE2005. November 5-11, 2005. Orlando, FL.


Taubes, G. 2000. Biologists and engineers create a new generation of robots that imitate life. Science. 288:80-3.

doi: 10.1126/science.288.5463.80.

Vincent, J. 1990. Structural Biomaterials. Princeton: Princeton Univ. Press.

Vogel, S. 1998. Cat’s Paws and Catapults. New York: W. W. Norton.


C O M M E N T A R Y

Biomimicking Marine Mechanismsand Organizational PrinciplesA U T H O RYoseph Bar-CohenJet Propulsion Laboratory,California Institute of Technology

Nature is effectively a giant labora-tory where trial-and-error evolution-ary experiments are taking place. Asnature performs its experiments, allthe fields of science and engineeringare employed, including physics,chemistry, mechanical engineering,and materials science. The processesrange in scale from nano and micro(e.g., viruses and bacteria) to macroand mega (e.g., our life scale, ele-phants, and whales). To address themany survival challenges, biologicalsystems came up with superb solu-tions. The constraints in addressingthese challenges are similar to thosethat human engineers are facing, in-cluding the need to maximize thefunctionality of their design and pro-duce systems that use minimal re-sources (e.g., materials, energy, cost,etc.). Humans have always made ef-forts to use nature as a model forinspiring innovation and problemsolving. However, biological andbotanical systems have superiorcapabilities, including producingmaterials—they use their body tem-perature, the materials are recyclable,and the process does not involvepollution.

To take advantage of the capabili-ties of nature, the field of biomimeticsinvolves seeking to understand anduse of the capabilities as a model for

copying, adapting, and inspiring con-cepts and designs (Bar-Cohen, 2005;Bar-Cohen, 2011; Benyus, 1998;Vincent, 2001). For this purpose,scientists are seeking rules, concepts,mechanisms, and principles to inspirenew possibilities. Some of the benefitsthat resulted from biomimetic ap-proaches have led to improved struc-tures, actuators, sensors, interfaces,control algorithms, software, drugs,defense, and intelligence, and mayhelp to improve our ability to recyclematerials and protect the environment.Some of the biomimetic characteristicsthat are being developed include shapemorphing, self-repair, self-replication,and self-reconfiguration (Bar-Cohen& Breazeal, 2003; Bar-Cohen, 2011).Increasingly, researchers are workingtowards adapting the capabilities ofmany creatures to perform tasks inhard-to-reach areas and in conditionsthat are too harsh or dangerous forhumans.

Marine biosystems are quite richin capabilities that have and can fur-ther benefit humans from biomimick-ing. There are many examples thatone can list, including one as simpleas the fins that are used by swimmersand divers that significantly enhancetheir performance. While it may bearguable that the fins were a biologi-cally inspired invention, one can statethat it is common knowledge thatswimming creatures (e.g., geese,swans, seagulls, seals, and frogs) havefeet with membranes that help themswim. The stability, maneuvering,and swimming performance of under-

water animals are determined by themorphology, position, and mobilityof their control surfaces. For cetaceans(i.e., whales, dolphins, and porpoises)the pectoral flippers are mobile hydro-foils that generate lift similar to engi-neered hydrofoils. The flippers havevarious shapes and are used to per-form lateral turning, dive, surface,brake, and other mobility-relatedfunctions. Studies of the 3-D geom-etry and hydrodynamic performanceof cetacean flippers with variousmorphologies help provide insightinto the maneuverability, drag andlift performance at high Reynoldsnumbers (Fish et al., 2011). Waggingthe body and tail is the main pro-pulsion method of marine swimmers(e.g., billfish and sailfish), allowingthem to reach significant speeds ofover 75 km/h. Submarines thatcould perform efficiently as marineswimmers using a flexible bodywould be an important explorationtool for scientists and would poten-tially have many military applications.

While significant advances werereached via mimicking and the inspira-tion of marine biology, there are manycapabilities in nature that are still farsuperior to engineered capabilities;examples include the following: The sonar of marine animals is far

superior to any existing marinesonar (Muller & Hallam, 2004).

Sea shells and skeletons of marineinvertebrates are far stronger andlighter than human-made materials,and their fabrication does not createpollution concerns.


Muscles stick to rocks even thoughthe adhesion is done in water, andthey sustain sticksion in spite ofthe strong impacts of ocean waves.On the other hand, most human-made adhesives fail when the adhe-sion is done in on a wet surface.

The chiton, which is a diminutivemollusk, has very strong teeth thatit uses to munch on rocks and ex-tract food (Weaver et al., 2010).This capability may be used to in-spire the development of effectivelightweight bits for in situ plane-tary exploration sampling drills(Figure 1).

One hopes that engineers will beable to rapidly prototype biologicalcapabilities as fast as it is now possibleto graphically edit photos of biologicalsystems, as illustrated in Figure 2. Inthis figure, a photo of a shark was ed-ited to create an imaginary image of aU.S. Navy vessel that is shark-like.While we are somewhat far from thiscapability, significant advances havebeen made by scientists and engineerswho seek to mimic marine biology.This special issue is dedicated to de-scribing and discussing the latest ad-vances that were inspired by marinemechanisms and their organizationalprinciples.

AcknowledgmentSome of the research reported in

this article was conducted at the JetPropulsion Laboratory, California In-stitute of Technology, under a con-tract with the National Aeronauticsand Space Administration.

Author:Yoseph Bar-CohenJet Propulsion Laboratory,California Institute of Technology4800 Oak Grove Drive,Pasadena, CA 91109-8099Email: [email protected]

ReferencesBar-Cohen, Y. (Ed.). 2005. Biomimetics—

Biologically Inspired Technologies. Boca

Raton, FL: CRC Press. pp. 1-527.

Bar-Cohen, Y. (Ed.). 2011. Biomimetics:

Nature-Based Innovation. Boca Raton, FL:

CRC Press, Taylor & Francis Group.

pp. 1-788.

Bar-Cohen, Y., & Breazeal, C. (Eds.). 2003.

Biologically Inspired Intelligent Robots.

Bellingham, WA: SPIE Press. pp. 1-393.

Vol. PM122.

Benyus, J.M. 1998. Biomimicry: Innovation

Inspired by Nature. New York, NY: Harper-

Collins Publishers. pp. 1-302.

Fish, F.E., Smits, A.J., Haj-Hariri, H.,

Iwasaki, T., & Bart-Smith, H. 2011.

Biomimetic swimmer inspired by the manta

ray, Chapter 17. In Biomimetics: Nature-

Based Innovation, ed. Bar-Cohen, Y. Boca

Raton, FL: CRC Press, Taylor & Francis

Group.

Muller, R., & Hallam, J.C.T. 2004. From

bat pinnae to sonar antennae: augmented

obliquely truncated horns as a novel para-

metric shape model. In: Proceeding of the 8th

International Conference on the Simulation

of Adaptive Behavior, SAB’04. Massachusetts,

USA: The International Society for Adaptive

Behavior.

Vincent, J.F.V. 2001. Stealing ideas from

nature, Chapter 3. In: Deployable Structures,

ed. Pellegrino, S. Vienna, Austria: Springer-

Verlag. pp. 51-58.

Weaver, J.C., Wang, Q., Miserez, A.,

Tantuccio, A., Stromberg, R., Bozhilov,

K.N.P., … Kisailus, D. 2010. Analysis of an

ultra hard magnetic biomineral in chiton

radular teeth. Mater Today, 13(1-2):

pp. 42-52.

FIGURE 1

A front view of the Eudoxochiton nobilis(“Noble” chiton). Courtesy of Iain Anderson.

FIGURE 2

An example is shown where software wasused to turn a photo of a natural shark intoa vessel-like naval system. The photograph(a) was taken by the author; the imagebelow (b) shows the modification to a Navymarine vehicle and is courtesy of DavidHanson, Hanson Robotics LLC, TX.


C O M M E N T A R Y

Sink and Swim: Clues From Naturefor Aquatic RoboticsA U T H O RJeannette YenGeorgia Institute of Technology

There are many ways to travel bywater. Watercraft can use propellersand sails while living organisms ex-hibit quite a variety of modes of trans-port. They sink and swim, flap andglide, stroke and jet. From this, wesee a distinction between the limitedhuman solutions and the diverse nat-ural solutions. This natural diversitycontinues to inspire inventions in ro-botics. Leonardo da Vinci envisionedwalking on water (Figure 1), some-thing water striders could do millionsof years ago (Hu et al., 2007). Bath-tub toys such as the TwiddleFish,designed by Chuck Pell of DukeUniversity (Figure 2), taught us theimportance of stiffness in how fish

swim so well. Pell commented, “Inthe hands of kids…they can reallyfeel what’s going on.” To the delightof pre-K–12 children and their par-ents, there are propulsive toys on themarket with remarkably functionalbody parts, serving as inspiration andoutreach to youngsters.

Recent efforts have focused onflapping by fish and mantas (e.g.,Tangorra et al., 2011; Fish et al.,2011a) or jetting by jellyfish andsquid (e.g., Moslemi & Krueger,2011). By replicating nature via ro-botics, we understand the significanceof the number of joints (Dean et al.,2009), the shape of the fin (Curetet al., 2011), and the direction of atail swish (Long et al., 2006) for con-trolling movement. Using these ro-bots to repeatedly vary the timing ofundulations of a fish or jellyfish, wediscover that the frequency of thereal organism is optimized to capture

the energy shed in the vortices left bythe previous pulse (Triantafyllou &Triantafyllou, 1995; Fish, 2006;Ruiz et al., 2010). This is not some-thing you could ask a fish or a jellyfishto do over and over again (though asbiologists, we have patiently waitedand recorded our aquatic creaturesperforming these behaviors over andover and over again; see Catton et al.,2011). We identify the body parts thatare important for propulsion, and wefigure out how many propulsors areneeded, along with their placement.Control systems coordinate the multi-ple fins. Sensors integrating differentsensory modalities are sought to devel-op navigation systems that reliablyguide the robot to reach its destina-tion. We test out new actuators thataccelerate unsteadily to increase thrust.We use new materials for joints orbodies with a compliance to achievethe flexibility needed for realistic un-dulations and squeezes (Lauder et al.,2007; Cutkosky & Kim, 2009; Ruizet al., 2010). By matching nature asclosely as possible, we gain a greaterunderstanding of how nature works.

But is matching reality necessary?What if instead of understandinghow shape and structure are opti-mized for best propulsion, we deter-mine how to achieve stealth bydesigning robots that blend in withthe background turbulence or matchthe disturbance made by the othermembers of the school? A robot likethat would enable us to spy onschools from the inside out, by be-coming a schoolmate. Can we study

FIGURE 1

Leonardo da Vinci: Shoes for walking on water(image from http://en.wikipedia.org/wiki/Walking_on_water).

FIGURE 2

TwiddleFish by Charles Pell, Bio-Design Studio,Duke University (cited by Guterl, 1996). Photocourtesy of F. Fish.


how the sensing system is integratedwith the propulsors to enhance theacuity of information capture? Whatprinciples can we abstract from natu-ral aquatic propulsion to improvehow we save energy or save materialsor improve performance, as all surviv-ing species do in a variety of ways tosuit the constraints of the environ-ment in which they have evolved?Perhaps we can take advantage of theenvironment and glide and surf wherepossible, using the free energy of thesea. Oceanographic gliders like theSlocum glider Scarlet 27 (Figure 3;Schofield et al., 2010) are beautifullydesigned, wherein one version varies itsballast by a phase change in response toambient ocean temperature (Webb et al.,2001). By using whale- or copepod-like buoyancy control (Clarke, 1978;Pond & Tarling, 2011) and harvestingenvironmental energy in the oceantemperature gradient, this glider trav-eled great distances with much lessenergy than other forms of propulsion.

Simple modifications of shape onkey control surfaces can lead to largevariations in the balance of lift anddrag essential for tight maneuvering,as exemplified by the tubercles ofwhale fins that now enable wind tur-bines to capture energy at lower windspeeds (Fish et al., 2011b). Canwe fur-ther achieve an economy of materialsby adopting the streamlined form ofthe shark and applying the hierarchi-cal structure of its denticles to finetune that drag reduction (Dean &Bhushan, 2010)?

Indeed, looking at the familiarplayers in the sea points out manycases of convergent evolution interms of undulatory or jet-like pro-pulsion. A closer look reveals otherunusual adaptations for aquatic mo-bility such as the parachutes of ptero-pods or the floats of siphonophores or

the multiple oars of copepods, krill,and polychaetes. Analyses of theirtransport mechanisms may againopen our eyes to novel designs of fu-

ture underwater vehicles that cansteadily hover, smoothly cruise, rapid-ly escape, quietly sink or perform anygait as needed in tight quarters. With

FIGURE 3

(A) Autonomous robots that follow the routes of swimming penguins are collecting informationthat could help scientists understand why the birds’ populations are dropping rapidly. The un-derwater robots, called gliders, are programmed to record ocean conditions as they follow thetracks of Adelie penguins swimming in the Southern Ocean surrounding Antarctica. (B) Diagramof Teledyne-Webb Corporation’s Slocum Glider (coastal model). The Front Main HousingSection glider’s ballast, and consequently its flight, is controlled by moving water into or outof the Fore Wet Section. From Kahl et al. (2010). (C) The 221-day path taken by Scarlet Knight(adapted from Schofield et al., 2010; Kahl et al., 2010; images courtesy of O. Schofield: http://rucool.marine.rutgers.edu).


the rapid evolution occurring in mate-rials science research and in controlsystems, we may find ourselves travel-ing through fluids in unexpectedways.

Author:Jeannette YenSchool of Biology,Center for Biologically InspiredDesignGeorgia Institute of Technology,Atlanta, GA. 30332-0230Email: [email protected]

ReferencesCatton, K.B., Webster, D.R., Kawaguchi, S.,

& Yen, J. 2011. The hydrodynamic disturbances

of two species of krill: Implications for aggre-

gation structure. J Exp Biol. 214:1845-56.

Clarke, M.R. 1978. Buoyancy control as a

function of the spermaceti organ in the sperm

whale. J Mar Biol Assoc UK. 58:27-71.

doi: 10.1017/S0025315400024395.

Curet, O.M., Patankar, N.A., Lauder, G.V.,

& MacIver, M.A. 2011. Aquatic manoeuver-

ing with counter-propagating waves: A novel

locomotive strategy. J Roy Soc Interface.

8(60):1041-50. doi: 10.1098/rsif.2010.0493.

Cutkosky, M.R., & Kim, S. 2009. Design

and fabrication of multi-material structures for

bioinspired robots. Philos T R Soc A.

367:1799-813. doi: 10.1098/rsta.2009.0013.

Dean, B., & Bhushan, B. 2010. Shark-skin

surfaces for fluid-drag reduction in turbulent

flow: A review. Philos T R Soc A. 368(1929):

4775-806. doi: 10.1098/rsta.2010.0201.

Dean, M.N., Swanson, B.O., & Summers,

A.P. 2009. Biomaterials: Properties, variation

and evolution. Integr Comp Biol. 49(1):

15-20. doi: 10.1093/icb/icp012.

Fish, F.E. 2006. Limits of nature and advances

of technology: What does biomimetics have to

offer to aquatic robots? Appl Bionics Biomech.

3(1):49-60. doi: 10.1533/abbi.2004.0028.

Fish, F.E., Nichols, R.H., Dudas, M.A.,

Moored, K.W., & Bart-Smith, H. 2011a.

Kinematics of swimming in the manta ray

(Manta birostris): 3D analysis of open

water maneuverability. Integr Comp Biol.

51(Suppl. 1):E42.

Fish, F.E., Weber, P.W., Murray, M.M., &

Howle, L.E. 2011b. The tubercles on

humpback whale’s flipper: Application of

bio-inspired technology. Integr Comp Biol.

51(1):203-13. doi: 10.1093/icb/icr016.

Guterl, F. 1996. Playthings of Science.

http://discovermagazine.com/1996/dec/

playthingsofscie946. (accessed 15 July 2011).

Hu, D., Chan, B., & Bush, J.W.M. 2007.

Water-walking devices. Exp Fluids. 43:

769-78. doi: 10.1007/s00348-007-0339-6.

Kahl, L., Oscar Schofield, A., & Fraser, W.R.

2010. Autonomous gliders reveal features

of the water column associated with foraging

by adelie penguins. Integr Comp Biol.

50(6):1041-50. doi: 10.1093/icb/icq098.

Lauder, G.V., Anderson, E.J., Tangorra, J., &

Madden, P.G. 2007. Fish biorobotics: Kine-

matics and hydrodynamics of self-propulsion.

J Exp Biol. 210:2767-80. doi: 10.1242/

jeb.000265.

Long, J.H.J., Koob, T.J., Irving, K., Combie,

K., Engel, V., Livingston, N.,… Schumacher,

J. 2006. Biomimetic evolutionary analysis:

Testing the adaptive value of vertebrate tail

stiffness in autonomous swimming robots.

J Exp Biol. 209:4732-46. doi: 10.1242/

jeb.02559.

Moslemi, A.A., & Krueger, P.S. 2011. The

effect of Reynolds number on the propulsive

efficiency of a biomorphic pulsed-jet

underwater vehicle. Bioinspir Biomim.

6(2):026001. doi: 10.1088/1748-

3182/6/2/026001.

Pond, D.W., & Tarling, G.A. 2011. Phase

transitions of wax esters adjust buoyancy

in diapausing Calanoides acutus. Limnol

Oceanogr. 56(4):1310-8.

Ruiz, L.A., Whittlesey, R.W., & Dabiri, J.O.

2010. Vortex-enhanced propulsion.

J Fluid Mech. 668:5-32. doi: 10.1017/

S0022112010004908.

Schofield, O., Kohut, J., Glenn, S., Morell,

J., Capella, J., Corredor, J., … Boicourt, W.

2010. A regional Slocum glider network in the

Mid-Atlantic coastal waters leverages broad

community engagement. Mar Technol Soc J.

44(6):64-74.

Tangorra, J., Phelan, C., Esposito, C., &

Lauder, G.V. 2011. Use of biorobotic models

of highly deformable fins for studying the

mechanics and control of fin forces in fishes.

Integr Comp Biol. 51(1):176-89. doi:

10.1093/icb/icr036.

Triantafyllou, M.S., & Triantafyllou, G.S.

1995. An efficient swimming machine.

Sci Am. 272:64-70. doi: 10.1038/

scientificamerican0395-64.

Webb, D.C., Simonetti, P.J., & Jones, C.P.

2001. SLOCUM: An underwater glider

propelled by environmental energy.

IEEE J Oceanic Eng. 26(4):447-52.

doi: 10.1109/48.972077.


C O M M E N T A R Y

Developing Bioinspired Autonomous SystemsA U T H O RThomas M. McKennaOffice of Naval Research

Research that seeks to identify theprinciples, strategies, and mechanismsused by motile aquatic animals offersopportunities to develop underseavehicles that exceed current capabili-ties and enable the Navy to expandthe operational envelope of autono-mous undersea vehicles. Autonomousundersea vehicles serve in a number ofimportant current and emerging rolesin Navy missions, including surveil-lance for anti-submarine warfare,mine countermeasures, ImprovisedExplosive Device (IED) detectionand localization, force protection (e.g.,counter-diver missions) in harbors, andriverine exploration and characteriza-tion. However, current unmanned un-dersea vehicles (UUVs) have technicalgaps in areas such as mission duration(due to power constraints), excessivenoise generation, speed, limited ma-neuverability, lack of tight integrationof sensing and maneuver, propulsionand maneuver dead zones at lowspeeds, and inability to operate in thesurf zone. Considering the capabilitiesof sea creatures for efficient propulsionover a large range of speeds, their abil-ity to operate in or even exploit ener-getic ocean environments like the surfzone, extraordinary maneuverability,and the special sensing evolved for pre-dation, schooling and navigation,there are many lessons for technolo-gists. The basic research programs inBio-Inspired Autonomous Systems atthe Office of Naval Research (ONR)have supported research using four

approaches: (1) the identification,modeling, and emulation of the bio-mechanics and fluid mechanics ofunderwater propulsion and controlin swimming organisms, (2) iden-tification and exploration of thesensorimotor control of animals andintegrated closed loop control usingbiosensing (e.g., biosonar, electro-sense, lateral line sensors, optic flow,magnetic sense), (3) the developmentof muscle-like actuators and fin de-signs that exploit these materials,and (4) design and development ofswimming prototype AutonomousUnderwater Vehicle (AUVs) as proofof principle and for performanceevaluation. These studies take theirinspiration from diverse sea creaturesand include high-performance swim-mers like bluefin tuna, squid, raysand seals; animals that thrive in thesurf zone like lobsters and crabs; andanimals with special senses likebiosonar (i.e., dolphins) and electro-sense (i.e., sharks and ribbonfish).More recently, opportunities haveemerged for microrobotic underwatersystems capable of sensing, reporting,and remediation of environmentalchemicals that exploit the conver-gence of synthetic biology, nano-technology, and electro-optic systems.This has prompted renewed interestin the propulsion biology of organ-isms that use cilia and flagella forlocomotion.

Key Science andTechnology Issues

There are a number of key issuesfor the development of bioinspiredautonomous systems.

1. Developing high-efficiencypropulsion that exceeds the capabil-ity of propellers. This is particularlyimportant for achieving long-durationmissions. Although there is a sizableNavy effort to develop new energysources for underwater vehicles, intro-duction of more efficient propulsionsystems would reduce the power re-quirements and thereby lengthen mis-sion durations. Early efforts to mimicthe caudal fins of high-performanceswimmers like tuna did not produceperformance that exceeds propellers(but recent work on the Ghostswim-merTM, discussed in Rufo and Smithers’commentary in this issue, looks prom-ising), and Bandyopadhyay (2005) hasproduced a meta-analysis showing thatanimals do not have an advantage overman-made systems in cruise, but ani-mals do show greater maneuverabilityrelative to man-made underwatervehicles. One promising new form ofpropulsion is the use of bioinspiredhigh-lift foil propulsors. High-lift pro-pulsion was introduced in the biologi-cal context in the analysis of fly winglift (Bandyopadhyay, 2009; Ellington,1984; Dickinson et al., 1999). Flywings, which pitch and heave with a90° phase difference, can achieve liftcoefficients substantially greater thanrigid foils that have a constant angleof attack, and high-lift foil propulsorsshould be capable of an order of mag-nitude greater lift than traditionalnaval propellers. High-lift foils alsoproduce substantially less noise thantraditional propellers. Bandyopadhyayet al. (2008) have produced a series ofrigid high-lift foils (roughly analogousto penguin pectoral fins), characterized


their propulsion properties, andmounted them on an undersea vehicle.This first version was called bioinspiredautonomous undersea vehicle (BAUV)and a second, lower-diameter versionwas called self-propelled line array(SPLINE) (Figures 1 and 2). High-liftflapping foils are efficient, and combin-ing six multiple foils on a vehicle canachieve extraordinary maneuverabilityand hover and exhibit low-noise emis-sion, but the flapping foil configurationis not consistent with producing highspeeds. Recently, Bandyopadhay hasdesigned a new propulsor called“Slosher” that has multiple foils withvariable angles of attack arranged radi-ally that can operate as a high-lift pro-pulsor at low speeds or, with the foilslocked, as a traditional propeller athigh speeds. This avoids the dead-band of controllability at low speeds.Moreover, hybrid vehicles, such as theRAZOR (Figure 3), have been de-veloped that combine four high-liftfoils and two props. At low speeds,the high-lift foils provide high maneu-verability and hover, but when theprops provide higher speeds for tran-sit of the vehicle, the foils become con-trol surfaces.

2. Developing adaptive control-lers for high-degree-of-freedombioinspired propulsors. Many ofthe bio-inspired propulsion systemsexhibit high degrees of freedom. Forthe case of the high-lift flappingfoils, roll, pitch, and frequency aremotion parameters to be specified toachieve efficient fin kinematics, andfor multiple fin vehicles, the phasingof the fins is critical. In addition tocommanding these parameters, for agiven vehicle maneuver, the fin con-troller must respond to perturbations.Fortunately, there are bioinspired so-lutions to such control issues. One ofthe key components of the mamma-

lian motor control system is the olivo-cerebellar system. Coupled neuronsin the inferior olive generate a 10-Hzrhythm, and animal muscle contrac-tions are initiated from particularphases of this rhythm. The inferiorolive has multiple oscillating domains,with phase shifts between these do-mains, and these domains are underthe control of the cerebellar cortex,which is a recipient of sensory inputsfrom many modalities. Llinas et al.(2004) developed a model wherebysequences of motor commands canbe generated by the olivo-cerebellarsystem. The biophysics of couplingof these neurons also promotes aphase reset property for rapid syn-chronizat ion (Kazantsev et a l . ,2004). Bandyopadhyay (2008) hasimplemented this model in analog cir-cuits and has shown it to be an effec-tive controller of power and thrust inthe high-lift foils and the ability torapidly return to normal followingperturbations. This model is able tosynchronize across multiple fins inorder to achieve optimum gaits thatminimize pitching and rolling of theBAUV. This controller enabled theBAUV to precisely hold a line loadfor 20 days. Based on these laboratoryresults, the BAUV vehicle is predictedto support a mission duration of3 weeks with current battery technol-ogy, although this assumes most ofthat time is spent in hover.

For fish and eels that use wholebody or caudal fin propulsion, theoscillations are generated by centralpattern generators. However, the mod-ulation of these neural generatorsduring locomotion to achieve desiredthrust and vectors has not been fullycharacterized for animals, and usingthis approach in artificial systems re-quires learning or genetic algorithmsto tune them to particular maneuvers.

FIGURE 1

BAUV developed by Dr. Promode Bandyopadyayat the Naval UnderseaWarfare Center, Newport,RI (NUWC-NPT). The BAUV has six high-liftfoils and a controller based on a model of theolivo-cerebellar circuits of mammals.

FIGURE 2

SPLINE. This is a refinement of the BAUV, de-signed for long-duration testing of its abilityto pull a load, maintain accurate position, andkeep a line taut that has been fixed at the dis-tal end while maneuvering (NUWC-NPT).

FIGURE 3

RAZOR vehicle. This vehicle was developed byRichard Berube and Promode Bandyopadyayat NUWC-NPT. It has four high-left foils andtwo rotating propulsors. At low speeds, ituses the foils for high maneuverability.


3. Exploitation of fish swimmingmodes for underwater vehicle pro-pulsion and maneuvers. Fish have anumber of swimming modes that areworthy of consideration for emula-tion. Fish swimming types can beclassified as either body and/or caudalfin movements vs. those that use me-dian and/or paired fin propulsion.One review of fish swimming modes(Sfakiotakis et al., 1999) singled outlunate tail propulsion (e.g., tuna), un-dulating fins (e.g., some rays), andlabriform (oscillatory pectoral fin)swimming mechanisms as having thegreatest potential for exploitation inartificial systems. ONR is currentlysupporting research to exploit allthree of these mechanisms in additionto the gymnotiform mode that in-volves undulations of a long ventralmedian fin. Batoid fishes utilize oneof two modes of locomotion, employ-ing either undulatory (passing multi-ple waves down the fin or body)or oscillatory (flapping) kinematics(Rosenberger, 2001). An ambitiouseffort to build batoid vehicles, usingpredominately oscillating kinematics,is being undertaken by a team ledby Hillary Bart-Smith (described inMoored, Fish, Kemp, & Bart-Smith,this issue) supported by an ONR pro-gram managed by Robert Brizzolara.Two prototype manta vehicles wererecently demonstrated in a studentcompetition (Pennisi, 2011).

Exploitation of fish locomotionusing lunate tail fins began withthe seminal studies of Triantafyllou(Barrett et al., 1996; Anderson et al.,1998) analyzing how bio-inspiredfoils create and exploit vortex struc-tures. His laboratory also developedthe interesting Robotuna (based onthe bluefin tuna) and Robopike proto-types, which were propelled usingcaudal fins. Triantafyllou’s group

demonstrated very high-propulsionefficiencies (up to 91%) for theRobotuna; however, the Robotunadid not achieve very high speeds.One of the creators of Robotuna,David Barrett at Olin College, is nowteamed with Boston Engineering todevelop the Ghostswimmer vehicle(described in Rufo’s commentaryin this issue). The objective of theGhostswimmer project is to exceedthe performance of current similarsize UUVs on speed, maneuver, mis-sion duration, noise, rapid responseand cost. The goal is to demonstratethe capability to conduct fully autono-mous missions.

Another active research area seeksto exploit the principles of fish pecto-ral fins for propulsion and maneuver.Fish pectoral fins are highly flexibleand enable a high degree of precisemaneuver and station keeping incurrents. Sunfish fins have flexiblerays, and attached membranes exhibita cupping motion on the forwardstroke that produces upper and lowerleading edge vortices. These fins gener-ate positive thrust throughout the finbeat, and turning involves asymmetricuse of these fins. Small prototype finswith this ray and membrane structurehave been shown in the laboratory,but no free swimming vehicles haveyet been developed (Gottlieb et al.,2010; Tangorra & Lauder, 2011, thisissue). Rigid pectoral fins have beenimplemented on a number of swim-ming vehicles, but these are mainlyused as control surfaces.

Aquatic animals that are propelledby jetting can also provide inspirationfor novel propulsion mechanisms.Mohseni (Krieg & Mohseni, 2010;Krieg et al., this issue) has character-ized the biomechanics and hydro-dynamics of squid propulsion anddeveloped new bioinspired thrusters,

and Priya and his colleagues (seeJoshi et al., this issue) have built pro-totype jellyfish that closely mimictheir swimming modes.

4. Development of muscle-likeactuators. To fully exploit animal-like locomotion and sensorimotorcontrol mechanisms, it is essential todevelop muscle-like actuators, linearactuators, adaptive compliant struc-tural materials, and elastic skins withproperties much closer to biologicalsystems than current technology.There are substantial inefficiencieswith implementing bio-inspired loco-motion using rotary motors and com-plex power transmission systems. Tworecent efforts to develop fins based onionic polymer-metal composites aredescribed in the papers by Kim et al.and Tan in this issue.

5. Closed-loop control of bio-inspired underwater vehicles. Incrustaceans, there are detailed accountsof sensorimotor reflexes for control oftail and legs, and robotic lobsters havebeen built with many levels of bio-inspiration (Ayers & Crisman, 1992;Ayers et al., this issue). However, forfish, the neural mechanisms by whichthe sensory afferent information onmotion, flows, and hydroacousticpressure is processed, integrated andrelayed to motor neurons is notknown. One reason for this is thatelectrophysiological recording inswimming fish is technically challeng-ing. One open question is whether fishcan sense vortices and maneuver ortime fin movements to exploit themor avoid them. The ONR is currentlysupporting a number of efforts inclosed-loop control and bionavigation.These projects include identifying therole of hydroacoustic receptors in thelateral line and fin and body me-chano-reception in flow and vortexsensing and tracing the connectivity


of these receptors into spinal sensori-motor circuits involved in locomotionand fin control (Green et al., 2011;Green & Hale, 2011). This project isalso developing robotic fish prototypes(see Tangorra et al., this issue; Lauderet al., this issue) with detailed pectoralfins. This work entails some basic re-search on fish neurophysiology, sincecircuit-level neurophysiology is muchmore challenging in fish than in ter-restrial vertebrates, and many of thespinal motor reflexes and circuits in-volved in limb control that were wellestablished for mammals by the1980s are still in nascent stage for thepectoral fins of fish.

6. Exploitation of special sensesof aquatic animals. A number offish species and sharks are able to nav-igate and search for prey using electro-reception; some animals also exploitgeomagnetic signals for navigation,and the biosonar of dolphins supportsa range of behaviors. The ribbon fishcan sense prey in murky waters usingelectrosense and then use its ventralribbon fin to maneuver to this prey.Ongoing ONR projects are per-forming system identification of visualand eletroreceptive feedback control oflocomotion in the ribbon fish (Rothet al., 2011; Mitchell et al., 2011) tobuild a working electrosense and elec-tromagetosense module for navigationand target localization and to build aprototype ribbon fish (“Ghostbot”)(Curet et al., 2011). A vehicle designedto sense electric and magnetic anom-alies could enable new mine counter-measures systems or enable navigationthat does not depend on GPS orexpensive Inertial Measurement Units(IMUs).

Sharks exhibit exquisite sensitivityto perturbations of electric fields andone current project seeks to developnew highly sensitive electromagnetic

and hydroacoustic sensors based onshark sensor biophysics (Kalmijn,Scripps).

Dolphins have extraordinary abili-ties to recognize objects using biosonar.ONR has supported research character-izing this ability in terms of psycho-acoustics, and several dolphin-inspiredsonars have been developed and dem-onstrated. Recently, Forsythe et al.(2008) demonstrated closed-loopcontrol of his bio-inspired autonomousunderwater vehicle using a simpledolphin-inspired biosonar in the explo-ration of acoustic targets. Dolphins,however, have the disadvantage thatone cannot analyze directly the neuralcircuits involved in their biosonar.Hence, from the earliest days of its ex-istence, ONR has supported the studyof bat biosonar. The neural circuits ofbat biosonar have been well character-ized. Recently research has focused onhow bats use multistatic biosonar forobstacle avoidance. Bats not only havean extraordinary ability to navigatearound obstacles and strike preyusing their own sonar, but they canalso perform these feats using the re-turns from calls emitted by other batsin a swarm. This has implications forthe cooperative behavior of multipleAUVs.

7. Group behaviors. The ONRScience of Autonomy program is sup-porting efforts to identify the principlesof fish schooling and applying them tomultiple AUVs conducting ocean sur-veys. Such studies have addressed theminimal sensing requirements of fishused in schooling, including vision.Additionally, for AUVs, one could con-ceivably use man-made communica-tions like acoustic modems or lasers toenhance coordinated actions, but suc-cessful emulation of sense modalitiesused in nature (i.e., fish hydroacousticsensing) might greatly simplify coor-

dination of multiple vehicles—in par-ticular when a large number of vehiclesis operating at close quarters. There isalso an effort to emulate the grouphunting behaviors of dolphins in intel-ligent control systems for cooperativeautonomous vehicles.

8. The Navy is interested in dis-tributed, persistent sensing systemsfor tasks such as anti-submarine war-fare. One bioinspired approach tothis is to develop small sensor plat-forms that float in the water columnwith limited mobility, similar to jelly-fish. An ambitious effort to exploitmultifunctional materials and buildsuch jellyfish colonies (Priya) is under-way. Additionally, the emerging con-vergence of nanotechnology, syntheticbiology and micro-optoelectronics hasfostered the nascent development ofvery small autonomous systems. Someof the newmicrorobots being designedhave components consisting of colo-nies of micro-organisms and hybridbiomolecules, with motility beingprovided by a multitude of cilia orflagella. These new systems couldfunction as distributed sensors of haz-ardous materials, but with integratedmobility, mitigation and reportingcapabilities.

In summary, there are a number ofpromising opportunities for the Navyto extend the capabilities of autono-mous undersea platforms using bio-inspired technologies in propulsion,control and sensing.

Author:Dr. Thomas McKennaOffice of Naval ResearchDivision of Human andBioengineered Systems875 N. Randolph St., Suite 1425Arlington, VA 22203-1995Email: [email protected]


ReferencesAnderson, J.M., Streitlien, K., Barrett, D.S.,

& Triantafyllou, M.S. 1998. Oscillating

foils of high propulsive efficiency. J Fluid

Mech. 360:41-72. doi: 10.1017/

S0022112097008392.

Ayers, J., & Crisman, J. 1992. The lobster as

a model for an omnidirectional robotic

ambulation control architecture. In: Biological

NeuralNetworks in InvertebrateNeuroethology

and Robotics, eds. Beer, R., Ritzmann, R.,

& McKenna, T., 287-316. San Diego:

Academic Press.

Bandyopadhyay, P.R. 2005. Trends in bio-

robotic autonomous undersea vehicles. IEEE J

Oceanic Eng. 30(1):109-39. doi: 10.1109/

JOE.2005.843748.

Bandyopadhyay, P.R. 2009. Swimming and

flying in nature—The route toward applications:

The Freeman Scholar lecture. ASME J Fluid

Eng. 131:031801-29. doi: 10.1115/1.3063687.

Bandyopadhyay, P., Beal, D.N., & Menozzi,

A. 2008. Biorobotic insights into how animals

swim. J Exp Biol. 211:206-14. doi: 10.1242/

jeb.012161.

Bandyopadhyay, P.R. 2008. Synchroniza-

tion of animal-inspired multiple high-lift fins

in an underwater vehicle using olivo-cerebellar

dynamics. IEEE J Oceanic Eng. 33(4):

563-78. doi: 10.1109/JOE.2008.2005356.

Barrett, D., Grosenbaugh,M.,&Triantafyllou,

M. 1996. The optimal control of a flexible

hull robotic undersea vehicle propelled by an

oscillating foil. In: Proc. 1996 IEEE AUV

Symp. 1-9.


& MacIver, M.A. 2011. Mechanical proper-

ties of a bio-inspired robotic knifefish with an

undulatory propulsor. Bioinspir Biomim.

6(2). doi: 10.1088/1748-3182/6/2/026004.

Dickinson, M.H., Lehman, F., & Sane, S.P.

1999. Wing Rotation and the aerodynamic

basis of insect flight. Science. 284:1954-60.

doi: 10.1126/science.284.5422.1954.

Ellington, C.P. 1984. The aerodynamics

of hovering insect flight. IV. Aerodynamic

mechanisms. Philos T R Soc Lond B.

305:79-113.

Forsythe, S.E., Leinhos, H.A., &

Bandyopadhyay, P.R. 2008. Dolphin-

inspired combined maneuvering and pinging

for short-distance echolocation. J Acoust

Soc Am. 124(4):255-61.

Gottlieb, J.R., Tangorra, J.L., Esposito,

C.J., & Lauder, G.V. 2010. A biologically

derived pectoral fin for yaw turn maneuvers.

Appl Bionics Biomech. 7:41-55.

doi: 10.1080/11762320903537782.

Green, M.H., Ho, R.K., & Hale, M.E. 2011.

Movement and function of the pectoral fins

of the zebrafish (Danio rerio) during slow

swimming. J Exp Biol. in review.

Green, M.H., & Hale, M.E. 2011. Moto-

neurons, muscles and movement: Neural

control of the pectoral fins during slow and

fast swimming. Integr Comp Biol. E51.

Kazantsev, V.B., Nekorkin, V.I., Makarenko,

V.I., & Llinas, R. 2004. Self-referential phase

reset based on inferior olive oscillator dy-

namics. P Natl Acad Sci. 101(52):18183-8.

doi: 10.1073/pnas.0407900101.

Krieg, M., & Mohseni, K. 2010. Dynamic

modeling and control of biologically inspired

vortex ring thrusters for underwater robot

locomotion. IEEE T Robot. 26(3):542-54.

doi: 10.1109/TRO.2010.2046069.

Llinas, R., Leznick, E., & Makarenko, V.I.

2004. The olivo-cerebellar circuit as a universal

motor control system. IEEE J Ocean Eng.

29(3):631-9. doi: 10.1109/JOE.2004.833212.

Mitchell, T., Fortune, E.S., & Cowan, N.J.

2011. Reweighting of vision and electrosen-

sation during locomotion in the Glass Knife-

fish. PLoS, to be submitted.

Pennisi, E. 2011. Manta machines. Science.

332:28-9. doi: 10.1126/science.332.6033.1028.

Rosenberger, L.J. 2001. Pectoral fin loco-

motion in batoid fishes: Undulation versus

oscillation. J Exp Biol. 204:379-94.

Roth, E., Zhuang, K., Stamper, S., Fortune,

E.S., & Cowan, N.J. 2011. Stimulus pre-

dictability mediates a switch in locomotor

smooth pursuit performance for Eigenmannia

virescens. J Exp Biol. 214:1170-80.

doi: 10.1242/jeb.048124.

Sfakiotakis, M., Lane, D.M., & Davies,

J.B.C. 1999. Review of fish swimming modes

for aquatic locomotion. IEEE J Oceanic Eng.

24(2):237-52. doi: 10.1109/48.757275.

Tangorra, J.L., & Lauder, G.V. 2011. Bio-

robotic models of highly deformable fins for

Studies of the mechanics and control of

fin forces. Integr Comp Biol. in press.

doi: 10.1093/icb/icr036.


C O M M E N T A R Y

GhostSwimmer™ AUV: Applying Biomimeticsto Underwater Robotics for Achievementof Tactical RelevanceA U T H O R SMichael RufoMark SmithersBoston Engineering Corporation

Introduction

I t is clear that unmanned under-water vehicles (UUVs), autonomousunderwater vehicles (AUVs), andother waterborne robots are success-fully addressing critical capabilityrequirements for many customers, in-cluding those in defense and oil andgas. There remains, however, despitethe best efforts of many entities, capa-bility gaps for these systems.

While UUVs share many of thesame navigation, power, and logisticalchallenges as their unmanned groundvehicle (UGV) and unmanned aerialvehicle (UAV) counterparts, they aresubject to additional challenges includ-ing limited communications and harshenvironments.

Developing biologically-inspired(or biomimetic) technologies can pro-vide some guidance for next generationsystems. The Advanced SystemsGroup (ASG) at Boston Engineering(Waltham, MA) is developing theGhostSwimmer™ AUV, for example,which endeavors to attack many of theproblems facing current UUVs. Theincreasing interest in the use of long-range/long-duration UUVs for littoralobservation, military surveillance, andother missions demands alternative ornew technology development. As per

the U.S. Navy UUV Master Plan(2004), the areas of autonomy, sen-sor s , and communica t ions areamong the leading areas of interest.Energy and propulsion are also mainareas of interest as per this document,where it states that “advanced energyand propulsion, in combination withother UUV technologies, will enablethe use of smaller vehicles (reducingcost) in the long term, and will pro-vide greater performance.” (TheNavy UUV Master Plan, 2004).

GhostSwimmer™ is a tactical, bio-mimetic autonomous “artificial fish”UUV that employs the mechanicsand dynamics of biological systems tocreate efficient swimming and highmaneuverability while remaining re-sponsive to the needs of current river-ine and littoral missions (among otherpossibilities). Its importance lies in itsability to provide advanced mobilityin a system that employs payloads.

Considerable work has been com-pleted on understanding the propul-sive characteristics of individual fishfins (extensive information on oscil-lating foil propulsion exists). TheGhostSwimmer™ builds upon thisand is modeled after a tuna. It is pro-pelled by a composite construction lu-nate caudal tail fin (taking its namefrom its crescent moon-like profile).In the biological tuna, almost all ofthe usable thrust is generated by theoscillating tail fin.Magnuson describesit as “a tapered hydrofoil with highaspect ratio, curved leading edge and

moderate sweep back. In cross section,the fin is shaped like a thin symmetri-cal airfoil with a rounded anterior, orleading edge, and a sharp posterior ortrailing edge” (Magnuson, 1978).

This oscillating foil has the capabil-ity of producing thrust at high efficien-cies. Triantafyllou and Triantafyllou(1993) stated “Oscillating foils areknown under certain conditions toproduce substantial thrust at high pro-pulsive efficiency. Fish and cetaceanshave developed through evolution apresumed optimal manner of propul-sion primarily through flapping ofthe posterior part of their body. It isshown that thrust production dependssignificantly on the dynamics of theunstable wake formed behind the foilor fish tail: thrust develops throughthe formation of a reverse Karmanstreet, whose preferred Strouhal num-bers are between 0.25 and 0.35. Exper-imental data from flapping airfoilsand data from fish observations con-firm the theoretical predictions.”Triantafyllou and Barrett were able toconstruct a precision oscillating foiltest apparatus with which propul-sive efficiencies in the range of 80%were repeatedly measured (Barrett,1996).

While based on this concept,GhostSwimmer™ strives to furtherthe understanding of how fish use alltheir fins to enhance propulsive perfor-mance, maneuverability, and stealthcharacteristics (all the while providinga capability for the Warfighter).


Overcoming the Challengesfor the Future AUVs

At this time, UUVs and AUVsstruggle with complex underwaterenvironments such as shallow andobstacle-filled waters. These vehiclesare subject to an intrinsic paradox;they require stable dynamic behaviorfor maintaining intended path/headingwithout changing control surfaces orpropulsion force. However, in complexenvironments, the vehicle must also beable to quickly change both path andspeed safely. Dynamic stability and re-sponse aspects such as overshoot, risetime, and peak time become critical(Azarsinal et al., 2007). Based on theirprowess in these areas, it makes sense toinvestigate biological sources of inspi-ration in the development of newman-made systems hoping to excel inthese environments.

The Boston Engineering AdvancedSystems Group’s GhostSwimmer™AUV (Figure 1) is intended to leveragethis biological inspiration. ExistingAUVs are generally propelled by rotarypropellers driven by electric motorswith energy stored in batteries. Smalldiameter propellers typically operateat low efficiencies and can suffer seri-ous lag times in transient response(Barrett, 1996). Cost effective propel-ler improvements are often limited byamaximumpractical diameter that canbe mounted on a UUV. In complexunderwater areas, rotating propellers

also present a significant snaggingrisk. Energy technology, despite recentprogress, is awaiting a breakthrough toprovide significantly larger (and safer)power densities; therefore, producinga technology that can address bothefficiency and control, independentof battery technology, has benefits.

Additional challenges that UUVsare subject to range from being as sim-ple as packaging electronics to be watertight at depth (versus splash-proof ),to using materials that are resistant toaggressive corrosion, to being able tolocalize underwater without the aid ofconventional techniques (such as theGPS or radio frequency [RF] commu-nications). As discussed later in thiscommentary, biomimetics can offerguidance in these areas as well.

Biomimetics andPragmatism

Biomimetics (from the Greek bios[life] + mimesis [to imitate]) is notnew. Mankind has been mimickingnature in products for centuries. Sim-ple everyday items such as salad tongs(based on bird beaks) through com-plex sensors such as sonar (based ondolphins and bats) have each beengenerated by mimicking biology.Even human terminology mimics na-ture, consider saw “teeth”, computer“viruses” and “worms”, or laptopsthat “hibernate.” Engineers and re-searchers have even mimicked plantsin robotic systems (turning to facethe sun for charging, for example)(Bar-Cohen, 2006). It should benoted that this discussion is notabout synthetic life (involving usingbiological components to build bio-logical systems).

Nature evolves by responding toneeds for surviving to the next gen-eration. It benefits from millions of

variations and trial and error experi-ments over the millennia. Engineerstrying to develop tactically relevanttechnologies obviously must “evolve”much faster and in a pragmatic man-ner. As such, developing biomimetictechnologies is as much about decidingwhat not to imitate as it is decidingwhat to imitate (Figure 2). This is be-cause direct and absolute mimicry isoften not appropriate or necessary.Consider that mankind flies withoutflapping wings as birds do. While un-derstanding and mimicking the con-trol surface design has direct benefitsto engineer fast and high flying aircraft,engineers diverted from nature’s de-sign with great success. However, thedefinition of the mission must alwaysbe at the core of any engineering effort.If the goal were to perch on a powerline, even mankind’s best aircraftwould fail.

Evolutionary improvements in na-ture also do not always exactly coincidewith the reasons for mimicking them.This should be considered when decid-ing whether the model is appropriateor ideal for mimicking in an engi-neered system designated for use incritical applications. Clearly, biologistsand others trying to learn about “howbiology works” are concerned withmimicking exactly how a biological

FIGURE 1

Boston Engineering’s GhostSwimmer™ PH IAUV (sponsor: ONR Code 341).

FIGURE 2

Early attempts at biomimetics often missedthe mark (Fuller).


model operates, its exact structure, itsmechanical parameters, and more.The focus of this commentary is onthe development of field-capable tech-nologies for performing unmannedtasks in relevant mission spaces. Inthis sense, it is important to recognizethat animals were evolved for survivalin a particular environment with spe-cific predators and other influences.

In the GhostSwimmer™ case, ex-tant tuna appeared roughly 60 millionyears ago (Dickson & Graham, 2004).These early tunas lived in a large cir-cumtropical waterway that encircledthe entire planet (the Tethys Sea).This waterway existed for roughly50 million years during which the de-velopment of tunas was influenced bychanges in the oceanography, inducedprimarily by tectonic activity. Thesechanges increased productivity, expandedfood webs, and opened potential niches(Dickson & Graham, 2004).

Graham and Dickson suggest thatthe appearance of these more extensiveocean areas with high productivity anddiversified food webs provided thecatalyst for tuna evolution: enhancedlocomotor performance and favoredmigratory behavior. Tuna’s specializa-tions, thunniform swimming, capacityfor regional endothermy, and an ele-vated aerobic capacity, are thought tobe based on the need for extendedswimming (Graham & Dickson,2004). Despite the fact that this is dif-ferent than the pressing missions of aUUV (such as mine counter mea-sures), this extended swimming im-plies endurance and propulsion thatis of value to a pragmatic mission.This leads to the question the robot-icist must ask: What aspects of afish’s “design” or mechanics is applica-ble to the desired mission?

Another consideration is the dif-ferent resources that are available. Evo-

lution and robotics engineers havedistinctly different tool sets, materials,actuators, and power and control sys-tems at their disposal. In certainareas, engineers have an advantage; ex-otic materials such as titanium mayallow advances that nature could notprovide. However, nature still has theadvantage (at least currently) in actua-tor power density (when consideringbandwidth and other factors), regener-ative capability, sensing, and controlprowess. For example, many biome-chanics sources have noted that muscletendon series elasticity is critical forenergetic and efficiency purposes(Paluska & Herr, 2006). Many ad-vances are in the works in the areas ofartificial muscles, neural networks, andsimilar technologies, but to date theylag behind their biological counter-parts in key areas. Some areas in needof advancement for actuation includepower to weight ratio, efficiency, fa-tigue life, and controllability.

Where the Challengesof Underwater Operationand Biomimetics Intersect

The challenges of achieving au-tonomy with unmanned systems aredetailed in many articles and pub-lications. From one perspective, au-tonomy intrinsically demands thatengineered systems act like biologicalsystems. For instance, these systemsneed obstacle detection and avoid-ance or situational awareness (“whatis happening around me?”), decision-making capacity (“should I respondaggressively or passively?”), and healthmonitoring (“am I OK?”). Interest-ing work in developing biologicallyinspired autonomy solutions basedon these principles is currently oc-curring at Jet Propulsion Lab (JPL)

(space mission systems) (Huntsberger,2001), Office of Naval Research(ONR) (sensory control, biomechan-ics) (ONR: Bio-Inspired AutonomousSystems), and Cornell (making increas-ingly complex machines) (Lipson).

The modeling and understandingof unsteady hydrodynamics is also achallenge for underwater vehicledevelopers. While analogous to aero-dynamic forces in some ways, hydro-dynamics in unsteady, obstacle-ladenenvironments is often more complexand, to date, difficult to model. Inparticular, controlled 6 degrees of free-dom movement, particularly in un-steady conditions, is challenging forcontrol, power, and propulsion sys-tems. Underwater vehicles have amore challenging environment forcontrolled mobility than most UGVsdue to this dynamic nature of theirsurroundings, and unless operating inopen seas (blue water), they can facefar more obstacles within tighter spacesthan UAVs.

Sensing and communication is an-other area where underwater systemshave very different challenges. Biolog-ical systems have remarkable and in-nate abilities to detect obstacles orprey using a variety of sensory capabil-ities from echolocation (bottlenosedolphin biosonar is “probably themost sophisticated target location andanalysis system in existence”) (Fulton,2010) to lateral lines (in fish such asthe tuna, water flow parallel to the mo-tion deflects cilium in the lateral linewhich in turn defines water velocity)(Martiny et al., 2009), to electric fields(weakly electric black ghost knife fishgenerate an electric field that causesvoltage perturbations due to the dif-ferences in electrical conductivity be-tween an object and the water, therebysensing its surroundings) (Martinyet al., 2009).


However, unmanned systems canleverage only those sensory capabilitiesdeveloped by the technology commu-nity at large. Herein lies the challenge,UAVs may have similar collisionavoidance sensory needs, but the tech-nologies for in-air sensing are currentlymore advanced and, due to relativelylow signal attenuation in air, can ef-fectively sense objects hundreds orthousands of feet away (ignoringcloud cover and other aspects for sakeof argument). The attenuation ofvarious sensing signals in salt water isdrastically more than in air, makingtechnologies that many take forgranted (WiFi and Bluetooth for ex-ample) impractical underwater. RFcommunications can work underwaterbut attenuation demands that low fre-quencies be used and the resultingtrade-off is bandwidth (a reasonablysized acoustic modem can provide140 bps-15 kbps [www.benthos.com]where even limitedWiFi [802.11 b forexample] can provide 2 Mbps “in air”[Mitchell]). To achieve real-timeor near real-time control, a UGV orUAV could use high-frequency,high-bandwidth RF communications.To date, this is a major challenge in un-derwater communicat ions. Ad-ditionally, the change in medium(from air into and through water orvice versa) makes communication dif-ficult due to refraction losses at theinterface (losses are large for electro-magnetic waves going from air intoseawater and intrinsic wavelength dif-ferences make an underwater antennafor the same radio different than itsin-air counterpart) (Butler, 2011).

Standard acoustic and light-basedsensors are greatly affected by the at-tenuation as well. Additionally, theiroperation often demands smooth,controlled motion at relatively slowspeeds. When coupling the hydro-

dynamic maneuverability challengesfor AUVs with these sensing challenges,one can only marvel at the ability offast-moving fish and marine mammalsto detect and maneuver. For these ani-mals, this detection and maneuver is asystem level process where sensingprovides exterioception and proprio-ception and enables the vorticity andoptimized flow control that is essentialto maneuverability and fast swimming(Triantafyllou et al., 2002).

The GPS is another technologytaken for granted in the 21st century.However, current AUVs must surfaceto get GPS fixes because GPS will notwork underwater; the signals from thesatellites cannot sufficiently penetratethe water. This presents the AUVwith a significant challenge in localiza-tion and navigation. Until these tech-nologies advance in a cost effectivemanner for underwater applications,one of the most effective means forovercoming the challenge includeshaving the ability to surface quickly(and often covertly) where the man-made underwater system can thenleverage in-air wireless and geo-positioning technologies. This rapidsurfacing can also benefit from bio-logically inspired techniques, as theGhostSwimmer has implemented.

Beyond Propulsion—Leveraging Biologyto Advance the Stateof the Art

Designing systems with bio-inspired control systems can open ave-nues for efficiently performing, andswitching between, necessary behaviorstates. Advances in computing tech-nology (increased memory and com-putational power in smaller, lessexpensive packages) now enables in-

corporation of newer and more ad-vanced biologically inspired controlsystems. This includes the develop-ment of intelligent control throughdistributed control in systems contain-ing their own “nervous systems,” struc-tured like their biological analog thatcan then “learn” through adaptivetechniques (Thomopoulos & Braught,1995).

The study of biology itself can pro-vide advances in the understanding ofunsteady hydrodynamics as men-tioned above. Biological systemsachieve high efficiencies and maneu-verability through the sensing, manip-ulation, and creation of optimized flowaround them. This includes the ma-nipulation of tip vortices at controland propulsive surfaces as well as theoptimization of output motion to gen-erate smooth and effective jets in theflow behind the body. The under-standing of these phenomena is di-rectly applicable to the creation ofhigher performance AUVs and evenmanned systems (Barrett, 1996).Boston Engineering’s ASG is currentlyapplying many of these principles todevelop improved control surfaces forthe U.S. Navy Sea Systems Commandas well as advanced AUVs for ONR.

Navigation is also an area that canbenefit from biomimetics. Specificareas of interest in autonomy to theNavy include path planning, behaviordevelopment, localization, on-boardmapping of environmental variability,and effective man-machine interfaceswith a limited communication capabil-ity (Wernli, 2001). Additionally, in-creasing uncertainty regarding GlobalNavigation Satellite Systems reliabilityhas led unmanned systems developersto seek valuable alternatives. Integratinginertial systems with reference maps ofGeophysical Fields of the Earth (GFE)is an area being explored by aerospace


entities that offers promise through useof the recent advancements in em-bedded micro-processing, includingmemory devices’ capability and minia-ture size. GFEs, properties of the earthitself, are already well mapped in geo-graphical system coordinates and canbe considered a reliable navigationdata source. Earth’s Magnetic Field(EMF) maps, models, and charts arecurrently in use for military and com-mercial entities (for directional infor-mation) and are available for at least98% of the earth’s surface (includingwater-covered areas) (Goldenberg,2006).

Research and behavioral experi-ments have shown that various animalssense and use the earth’s magnetic fieldfor navigation over both long and shortdistances (Johnsen & Lohmann,2005). Migratory animals, capable ofsensing variations in geomagneticfields, reference the earth’s mean fieldand its inclination at many points.Measurements of the field, includingspatial variations and temporal evolu-tions, tabulated by the U.S. GeologicalSurvey, allow the potential for geo-magnetic field sensors to mimic animalbehavior related to navigation (Zhaiet al., 2007).

Some animals, including certainbirds, sea turtles, salamanders, and lob-sters can discriminate small differencesin some of the earth’s magnetic fea-tures. They use positional informationin the earth’s field in several differentways and some actually learn the mag-netic topography of the areas they callhome. Some have postulated that ani-mals have two separate magnetosen-sory systems (Goldenberg, 2006). Acompass alone is rarely sufficient toguide animals to specific destinationsor along a long and complex migratoryroute due to currents and other error-inducing phenomena. Navigation

must be enhanced by the ability todetermine position relative to a desti-nation (human travelers use a GPS),i.e., positional information inherentin the earth’s magnetic field providesa similar, although less precise, assess-ment of location (Goldenberg, 2006).

Animals also use other cues; re-search has shown that bees navigaterelative to the sun by using the sun asa fixed point and orienting themselvesby maintaining a fixed angle betweenits line of flight and the line to thesun (www.physics.ohio-state.edu).Ocean waves combine with the earth’smagnetic field to serve as orientationcues for newly hatched turtles; whileolder turtles are following a map theylearned that enables them to establishtheir position relative to some distanttarget (Lohmann et al., 2004). Re-gardless of surface currents or otheroceanographic features, sockeye salmonuse an internal “map sense” to navigatehome after several migrations and in-duced errors. An olfactory “imprint”is made on smelts as they leave theirhome stream, but approaching theirstream from open sea demands oneother imprint. Fish are perceptive ofthe azimuth and altitude of the sun,but during overcast days, ferromag-netic mineral magnetite in their brainmay function as a biological compass.Another means for a fish to sense themagnetic field is by merely movingthrough the water (like a wire movedacross a magnetic field, electricalcurrent occurs in the wire) (Gedney,1984). It is possible that AUVscould mimic these approaches.

The electric field of ocean currentsindicate to sharks their drift relative tothe bottom or to deeper water layers.Electric current of an ocean stream in-vading a quiet bay may provide a sharkwith directional cues in familiar terri-tory. They may also explore the fields

by occasionally diving deeper or to thebottom and their orientating in uni-formDC electric fields has been provenbehaviorally. The electric sense operatesin a passive mode, whereas magneticfield detection is an active mode, allow-ing them to simultaneously sense driftwith ocean streams and magnetic head-ings. Based on this research, measuredpotentials, sensitivity and noise effects,and amplification are relatively wellknown (Kalmijn, 2000).

The sun’s movement across the skygives orientation signals that vary withtime of day. Migratory birds gain in-formation from the sun better thanhumans, detecting changing patternsof polarization. Using magnetism andcelestial rotation together can increasereliability (birds constantly cross checkthem and adjust) (www.teara.govt.nz).The sun’s position can be analyzed bylooking at the polarization pattern ofthe sky arising from sunlight scatter-ing (useful when the sun is obscured).Many insects use celestial polariza-tion for compass orientation by usingpolarization-sensitive photoreceptors.Biological research has shown thatsome underwater animals use polar-ization of light for navigation, commu-nication, and hunting. Studies havebeen performed using polarization ofscattered light underwater to improvevisibility. Polarization of light in wateris caused by refraction through the sur-face, scattering light in water, refractionby polarizing objects, and emissionfrom polarized light sources (Karpel &Schecher, 2011).

The Present andFuture of BiomimeticUnderwater Robotics

Rather than relying on the fu-ture development of a novel and


ground-breaking power source, BostonEngineering’s ASG, along with its team-mate, Olin College Intelligent VehiclesLaboratory (Needham, MA), is takinginspiration from a comparably sizedbiological system to develop an AUVoptimized for complex underwaterenvironments.

Thanks to funding from the ONRand a large internal R&D effort byboth Boston Engineering and OlinCollege, this new generation of AUVsis emerging. By leveraging past re-search, the latest technologies, and in-novative rapid prototyping techniques,the team developed an AUV that couldautonomously swim like a fish in only6 months (Figure 3). The team is cur-rently in Phase II withONR and expectsto showcase its next generation AUV inlate 2011 to demonstrate advancementsin efficiency and maneuverability whileaddressing and investigating the otherchallenges facing underwater vehiclesas described above.

While there have been many at-tempts to duplicate the swimming ac-

tions of a fish (Xianzhong DAI, 2003;Valdivia et al., 2006), a breakthroughachieved by the Boston Engineeringresearch team has been in producingthese movements efficiently within atactically relevant vehicle. After prov-ing that fish-like oscillating foil pro-pulsion could indeed have higherpropulsive efficiency and could beachieved by man-made methods, thenext challenge is to properly controland coordinate the movements. Theintended solution for this control isitself bioinspired. The result is anAUV platform that could outperformconventional technologies deployedtoday in both endurance and mobility.This work holds the potential to pro-vide a paradigm shift in underwaterAUV capability and usefulness.

The trend of mimicking biologyin engineering certainly is not new,but through the efforts of the variousparties working in the field of bio-logically inspired engineering and bio-logical mechanics, a reduction in thenumber of limitations in enablingtechnologies and knowledge is beingseen. Recent advances in the under-standing of biological hydrodynamics,sensing, navigation techniques, andcontrol coupled with low-power high-throughput computing (includingQuad Core Processors, www.intel.com), advances in material science(ionomers and self healing materials,www.bimat.org), and improved proto-typing techniques (such as direct metallaser sintering, www.morristech.com)has enabled robotics engineers asnever before. As industry and academiacontinue to make progress in demon-strating how effective biomimetic de-signs can be, the more developers ofnext generation tactical systems suchas Boston Engineering’s AdvancedSystems Group will be able to rise tothe technical challenges presented by

current adopters of unmanned tech-nology, especially in the underwaterspace.

Authors:Michael Rufo and Mark SmithersAdvanced Systems Group,Boston Engineering Corporation411 Waverley Oaks Road,Waltham, MA 02452Emails: [email protected]; [email protected]

ReferencesAzarsinal, F., Bose, N., & Seif, M.S. 2007.

An underwater vehicle maneuvering simula-

tion: Focus on turning maneuvers. J Ocean

Technol. 2(1):54-73.

Bar-Cohen, Y. 2006. Biomimetics: Biologi-

cally Inspired Technologies. Boca Raton,

FL: CRC Press. 469 pp.

Barrett, D.S. 1996. Propulsive efficiency of

a flexible hull underwater vehicle. Ph.D.

thesis, Massachusetts Institute of Technology.

Bee Navigation. http://www.physics.ohio-

state.edu/~wilkins/writing/Samples/shortmed/

fiskemedium. Accessed June 20, 2011.

Bird Migration. Navigating by the stars

and sun. http://www.teara.govt.nz/

EarthSeaAndSky/BirdsOfSeaAndShore/

BirdMigration/7/en. Accessed June 20, 2011.

Butler, L. 1987. Underwater Radio Com-

munication. http://www.qsl.net/vk5br/

UwaterComms.htm. Accessed June 20, 2011.

Dickson, K.A., & Graham, J.B. 2004.

Evolution and consequences of endothermy

in fishes. Physiol Biochem Zool. 77(6):

998-1018. doi: 10.1086/423743.

Fuller, J. Top 10 bungled attempts at one-

person flight. http://science.howstuffworks.com/

transport/flight/classic/ten-bungled-flight-

attempt4.htm. Accessed March 1, 2011.

Fulton, J.T. 2010. Dolphin biosonar echolo-

cation A case study. http://neuronresearch.net/

FIGURE 3

Boston Engineering’s GhostSwimmer™ PH IAUV in field testing.


hearing/pdf/Dolphin_biosonar_echolocation.

pdf. Accessed June 20, 2011.

Gedney, L. 1984. Do Salmon navigate by

the earth’s magnetic field? Article #691.

http://www.gi.alaska.edu/ScienceForum/

ASF6/691.html. Accessed June 20, 2011.

Goldenberg, F. 2006. Geomagnetic naviga-

tion beyond the magnetic compass. Goodrich

Corporation, Advanced Sensors Technical

Center, 14300 Judicial Road, Burnsville,

MN 55306. http://ieeexplore.ieee.org/xpl/

freeabs_all.jsp?arnumber=1650662. Accessed

June 20, 2011.

Graham, J.B., & Dickson, K.A. 2004.

Tuna comparative physiology. J Exp Biol.

207(iii):4015-24. doi: 10.1242/jeb.01267.

Huntsberger, T. 2001. Biologically inspired

autonomous rover control autonomous

robots. Auton Robot. 11:341-6.

doi: 10.1023/A:1012467829785.

Intel. Intel core quad processors. http://www.

intel.com/products/processor/core2quad/.

Accessed June 20, 2011.

Johnsen, S., & Lohmann, K. 2005. The

physics and neurobiology of magnetorecep-

tion. Nature Reviews Neuroscience. 6:703-12.

doi: 10.1038/nrn1745.

Kalmijn, J. 2000. Detection and processing

of electromagnetic and near-field acoustic

signals in elasmobranch fishes. Philos T R

Soc Lond B. 355(1401):1135-41.

doi: 10.1098/rstb.2000.0654.

Karpel, N., & Schechner, Y. 2011. Portable

polarimetric underwater imaging system with

a linear response. Dept. of Electrical Eng.,

Technion - Israel Institute of Technology,

Haifa 32000, Israel. http://webee.technion.

ac.il/~yoav/publications/polarimetric.pdf.


Lipson, H. Associate Professor. Cornell

University Mechanical and Aerospace Engi-

neering. http://www.mae.cornell.edu/lipson/.


Lohmann, K.J., Lohmann, C.M.F., Llewellyn,

M., Ehrhart, L.M., Bagley, D.A., & Swing, T.

2004. Animal behavior: Geomagnetic map

used in sea-turtle navigation. Nature.

428:909-10. doi: 10.1038/428909a.

Magnuson, J.J. 1978. Locomotion by scom-

brid fishes: hydromechanics, morphology,

and behavior. Fish Physiol. 7:239-313.

doi: 10.1016/S1546-5098(08)60166-1.

Martiny, N. 2009. Design of a lateral-line

sensor for an autonomous underwater vehicle.

In: 8th IFAC International Conference on

Maneuvering and Control of Marine Craft,

MCMC 2009. pp. 292-7. Sao Paulo, Brazil:

IFAC.

Mitchell, B. Wireless Standards-802.11b

802.11a 802.11g and 802.11n: The 802.11

family explained. http://compnetworking.

about.com/cs/wireless80211/a/aa80211standard.

htm. Accessed June 20, 2011.

Morris Technologies. Direct metal laser sin-

tering (DMLS). http://www.morristech.com/

dmls.asw. Accessed June 20, 2011.

Office of Naval Research. Bio-inspired

autonomous systems. http://www.onr.

navy.mil/en/Media-Center/Fact-Sheets/

Bio-Inspired-Autonomous-Systems.aspx.


Paluska, D., & Herr, H. 2006. Series

elasticity and actuator power output. In:

Proceedings of the 2006 IEEE International

Conference on Robotics and Automation.

Orlando, FL: IEEE.

Teledyne benthos acoustic modems.

http://www.benthos.com/acoustic-telesonar-

modem-subsea-ATM885.asp. Accessed

June 20, 2011.

The Navy Unmanned Undersea Vehicle

(UUV) Master Plan. November 9, 2004.

Thomopoulos, S., & Braught, G. 1995. A

biologically inspired mechanism of machine

perception and intelligent control for multi-

robot coordination. Penn State University.

Triantafyllou, M.S., Techet, A.H., Zhu, Q.,

Beal, D.N., Hover, F.S., & Yue, D.K.P.

2002. Vorticity control in fish-like propulsion

and maneuvering. Integr Comp Biol.

42(5):1026-31. doi: 10.1093/icb/42.5.1026.


1993. Propulsion through oscillating foils in

nature and technology. J Fluids Struct.

7:205-24. doi: 10.1006/jfls.1993.1012.

Valdivia y Alvarado, P., & Youcef-Toumi, K.

2006. Design of machines with compliant

bodies for biomimetic locomotion in liquid

environments. J Dyn Syst-T ASME.

128(1,3):11.

Wernli, R. 2001. Recent U.S. Underwater

Vehicle Projects. San Diego, CA: Space and

Naval Warfare Systems Center.

Wernli, R. 2011. Recent US Navy Under-

water Vehicle Projects. http://www.nmri.go.

jp/main/cooperation/ujnr/24ujnr_paper_us/

Ocean_Engineering_and_Resource/OERD_

Wernli.pdf. Accessed June 20, 2011.

Xianzhong, D. 2003. Some new progress of

advanced robotics in China. In: Report to JCF

2003 of IARP, Xianzhong DAI Robotics

Expert Group of National High-Tech R&D

Programme, China. Madrid, Spain.

Zhai, J. 2007. Geomagnetic sensor based on

giant magnetoelectric effect. Appl Phys Lett.

91(12):1-3. doi: 10.1063/1.2789391.


P A P E R

Autonomous Robotic Fish as Mobile SensorPlatforms: Challenges and Potential SolutionsA U T H O RXiaobo TanSmart Microsystems Laboratory,Department of Electrical andComputer Engineering,Michigan State University

A B S T R A C TWith advances in actuation and sensing materials and devices, there is a grow-

ing interest in developing underwater robots that propel and maneuver themselvesas real fish do. Such robots, often known as robotic fish, could provide an engi-neering tool for understanding fish swimming. Equipped with communicationcapabilities and sensors, they could also serve as economical, dynamic samplersof aquatic environments. In this paper we discuss some of the major challenges inrealizing adaptive, cost-effective, mobile sensor networks that are enabled byresource-constrained robotic fish. Such challenges include maneuvering in thepresence of ambient disturbances, localization with adequate precision, sustainedoperation with minimal human interference, and cooperative control and sensingunder communication constraints. We also present potential solutions and prom-ising research directions for addressing these challenges, some of which are in-spired by how fish solve similar problems.Keywords: robotic fish, adaptive sampling, mobile sensing platforms, aquaticsensor networks, water quality monitoring

Introduction

With 500 million years of evolu-tion, fish and other aquatic animalsare endowed with a variety of mor-phological and structural featuresthat enable them to move throughwater with speed, efficiency, and agil-ity (Lauder & Drucker, 2004; Fish &Lauder, 2006). The remarkable featsin biological swimming have stim-ulated extensive theoretical, experi-mental, and computational researchby biologists, mathematicians, andengineers, in an effort to understandand mimic locomotion, maneuvering,and sensing mechanisms adopted byaquatic animals.

Over the past two decades, therehas also been significant interest indeveloping underwater robots thatpropel and maneuver themselveslike real fish do (Triantafyllou &Triantafyllou, 1995; Kato, 2000;Anderson & Chhabra, 2002; Alvarado& Youcef-Toumi, 2006; Hu et al.,2006; Low, 2006; Epstein et al.,2006; Morgansen et al., 2007; Lauderet al., 2007; Chen et al., 2010; Aureliet al., 2010; Smithers, 2011). Oftentermed robotic fish, these robots pro-vide an experimental platform forstudying fish swimming and holdstrong promise for a number of under-

water applications. Instead of usingpropellers, robotic fish accomplishswimming by deforming the bodyand/or fin-like appendages, mostlyfunctioning as caudal fins and some-times as pectoral fins. Body deforma-tion and fin movements are typicallyachieved with motors. On the otherhand, advances in smart materialshave been explored to actuate roboticfish in a noiseless and compact way(Paquette & Kim, 2004; Tangorraet al., 2007; Chen et al., 2010; Aureliet al., 2010). Robotic fish producewake signatures similar to those ofreal fish and are thus less detectablethan propeller-driven underwater vehi-cles, which is an important advantagein applications requiring stealth.

Recent advances in computing,communication, electronics, andmaterials have made it possible tocreate untethered robotic fish withonboard power, control, navigation,

wireless communication, and sensingmodules, which turns these robotsinto mobile sensing platforms inaquatic environments. Schools of ro-botic fish can form wireless sensornetworks, which will have numerouspromising applications, such as moni-toring water quality, tracing oil spills,and patrolling harbors and coasts.Figure 1a shows a prototype of a ro-botic fish swimming in an inlandlake. Figure 1b shows the close-up ofanother prototype, equipped with adissolved oxygen (DO) sensor, globalpositioning system (GPS), and otherelectronic components, which hasbeen developed for monitoring theDO level in aquafarms. CollectedDO information will then be used tocontrol the aerators to maintain ahealthy environment for the aquaticanimals on the farm.

Autonomous robotic fish schoolswill provide a competitive alternative


to existing sensing technologies foraquatic and marine environments.Manual sampling, sometimes boat orship-based, is still a common practicein environmental monitoring, whichis labor-intensive with difficulty incapturing dynamic phenomena ofinterest. In-situ sensing with fixed orbuoyed sensors or vertical profilers isanother approach (Doherty et al.,1999; Reynolds-Fleming et al . ,2002). However, these sensors havelittle freedom to move laterally, and itwould require prohibitively manyunits for capturing distributed, spatiallyinhomogeneous information. The pastdecade has seen great progress in the useof robotic technology in aquatic sens-ing. Autonomous underwater vehicles(AUVs) (Bandyopadhyay, 2005), forexample, are being used for hydro-graphic survey, fishery operations, andenvironmental monitoring (Hydroid,2009). Another highly successful tech-nology is autonomous sea gliders,which has remarkable duration forcontinuous field operation becauseof highly energy-efficient design(Ericksen et al., 2001; Sherman et al.,2001; Webb et al., 2001; Rudnicket al., 2004). The downside for bothAUVs and gliders is their cost, startingat US $50,000 per unit (not including

the cost of sensors), prohibiting thedeployment of many of them forobserving with high spatial resolution,and excluding them from many appli-cations (such as aquafarm monitoring)where cost is critical. The size (meterslong) and weight (at the order of 50 kg)of these vehicles also make them cum-bersome to handle by a single person.

Small autonomous robotic fishhave the potential to address many ofthe aforementioned challenges. By asmall robotic fish, we mean one thathas length of 50 cm or less, displacesvolume of up to 5 liters, and costs nomore than US $5,000 (excluding thatof aquatic sensors to be mounted). Itslow cost, compact size, and lightweight would make it affordable andconvenient to deploy these robots ingroups for versatile applications andvarious environments, such as ponds,lakes, rivers, and even oceans. Schoolsof robotic fish could form dynamic,adaptive sensor networks and providedistributed sensing coverage with de-sired spatiotemporal resolution.

The realization of such a vision,however, is faced with a myriad ofchallenges. The size and cost consid-erations put stringent constraints onthe robot’s locomotion, battery, com-puting, and communication ca-

pacities. The wide adoption of therobotic fish-based sensing technologywill hinge on the robots’ ability to op-erate robustly in the unfriendly andoften unpredictable environment,with their limited onboard resourcesand with minimal human interven-tion. This poses challenges across awide spectrum, ranging from locomo-tion and maneuvering mechanisms, toenergy-efficient designs, localizationand communication schemes, andcontrol and coordination strategies,to name a few. In this paper, we outlinesome of the most critical challengesand discuss potential approaches oropportunities in research and technol-ogy advancement for addressing thechallenges.

Maneuvering inUncertain Environment

As a sensor platform, the roboticfish often needs to survey a givenpath or hover over a particular regionin the presence of ambient distur-bances caused by wind, waves, cur-rents, and turbulences. Regardless ofits propulsion mechanism, however, asmall robotic fish has limited actuationauthority to counteract the distur-bances. It is thus of great interest tobe able to sense the flow and react inthe most effective way under the actu-ation constraints. We can look to livefish for inspiration, because they dealwith this very problem on a regularbasis and have developed intricatesensing and actuation systems thatoffer us interesting insight.

Artificial Lateral LineMost fish use the lateral line system

as an important sensory organ to probetheir environment (Coombs, 2001).A lateral line consists of arrays of so

FIGURE 1

Prototypes of autonomous robotic fish developed by the Smart Microsystems Laboratory atMichigan State University: (a) testing in a lake and (b) prototype for dynamically monitoringthe DO level in aquafarms.


called neuromasts, each containingbundles of sensory hairs, encapsulatedin a gelatinous structure called cupula.Under an impinging flow, the hairs aredeflected, which elicits firing of thehair cell neurons and thus enables theanimal to sense the flow field, performhydrodynamic imaging, and identifynew field objects of interest. The lateralline system plays an important role invarious fish behaviors, including prey/predator detection, schooling, rheo-taxis, courtship and communication.

A lateral line-like sensory moduleor an artificial lateral line will be veryuseful for a robotic fish to improve itsmaneuverability. For example, withfeedback from the lateral line, therobot could manipulate vortices inthe flow with its actuated fins and ex-ploit the ambient flow energy for loco-motion (Beal et al., 2006) or performstation-keeping by responding appro-priately to the sensed ambient flow.Artificial lateral line systems, where ar-rays of beam or hair-like structures areused to measure flow velocities, havebeen proposed based on various phys-ical transduction principles, includinghot wire anemometry (Yang et al.,2006), piezoresistivity (Yang et al.,2010), capacitive sensing (Dagamsehet al., 2010), and encapsulated inter-face bilayers (Sarles et al., 2011).

Recently, we have exploited the in-trinsic mechanosensory property ofionic polymer-metal composites(IPMCs) to construct artificial lat-eral lines (Abdulsadda & Tan, 2011;Abdulsadda et al., 2011). As illustratedin Figure 2, an IPMC consists of threelayers, with an ion-exchange polymermembrane (e.g., Nafion) sandwichedbymetal electrodes. Inside the polymer,(negatively charged) anions covalentlyfixed to polymer chains are balancedby mobile (positively charged) cations.An applied mechanical stimulus, such

as a flow impinging on the IPMC,redistributes the cations inside andproduces a detectable electrical signal(typically open-circuit voltage orshort-circuit current) that is correlatedwith the mechanical or hydrodynamicstimulus (Chen et al., 2007). Con-versely, an applied voltage across anIPMC leads to the transport of cationsand accompanying solvent molecules,resulting in both differential swell-ing and electrostatic forces inside the

material, which cause the material tobend and hence the actuation effect(Shahinpoor & Kim, 2001). Figure 3ashows a prototype of an artificial lateralline consisting of four IPMC sensors.

While the physical construction ofrobust and sensitive artificial laterallines remains an active research area,it is of equal importance to makesense out of the data collected by thelateral line. Existing studies on biolog-ical and artificial lateral lines have

FIGURE 2

Illustration of the IPMC sensing principle.

FIGURE 3

Experimental results on localization of a dipole source with unknown location and vibration am-plitude: (a) prototype of IPMC-based lateral line, consisting of four IPMC sensors, and (b) local-ization results along three different tracks, based on solving a model-based nonlinear estimationproblem (Abdulsadda et al., 2011).


mostly focused on the problem oflocalizing a vibrating sphere, knownas a dipole, which is used to emulate pe-riodic tail beating or other appendagemovement of aquatic animals. Severalapproaches to signal processing havebeen reported, which include exploita-tion of the characteristic points (e.g.,zero-crossings, maxima, etc.) in themeasured velocity profile (Dagamsehet al., 2010), matching of the mea-sured data with preobtained templates(Pandya et al., 2006), beamformingtechniques (Yang et al., 2010), andartificial neural networks (Abdulsadda& Tan, 2011). We have further con-sidered a source localization problemwhere both the source location andits vibrating amplitude are unknown.The posed problem is interesting,since a source far away but with largevibration could produce a signal thathas similar amplitude as a signal pro-duced by a source nearby but withsmall vibration. By formulating andsolving a nonlinear estimation prob-lem based on an analytical model fordipole-generated flow, we are able toresolve both the source location andthe vibration amplitude simultaneously(Abdulsadda et al., 2011). As shown inFigure 3b, experimental results on anIPMC-based lateral line prototype(Figure 3a) have confirmed the effec-tiveness of the model-based estimationapproach.

Other than the dipole source lo-calization problem, there are a fewinteresting directions for the signalprocessing of artificial lateral lines.The first is the detection and localiza-tion of multiple, more sophisticatedmoving sources (including vortices).With the sources moving, the resulting-flow is no longer at a steady state, andthe processing algorithm needs tolocalize the sources with minimallatency. Another major problem to

consider is the information processingfor a lateral line that is mounted on arobotic fish, where the motion of therobot itself and its fins adds significant“noise” to the lateral line signal. Bio-logical fish deal with these problemseffectively through biomechanical fil-tering for enhanced signal-to-noiseratio and through dynamic filteringin the central nervous system to removethe unwanted signal components(Coombs & Braun, 2003; Bodznicket al., 2003). For example, dynamicneural mechanisms have been identi-fied for suppressing self-generatednoise (Coombs & Braun, 2003). Suchbiological insight will prove valuable indevising the mechanical, electrical, anddigital filtering mechanisms for solvingcomplex processing problems faced byartificial lateral lines.

Bioinspired FinAchieving high-maneuverability

hinges on the ability to manipulatethe fluid in a delicate manner. Fishoften use their pectoral fins to performsophisticated maneuvers (Drucker &Lauder, 2001, 2003). These maneu-vers involve complex conformationalchanges of the fins, involving cupping,twisting, and bending motions.Robotic fish fins, on the other hand,often use rigid foils (Kato, 2000;Morgansen et al., 2007). Recently,advances in soft actuation materials,e.g., IPMCs, have led to the explora-tion of these materials as flexible pro-pulsors (Paquette & Kim, 2004).However, the resulting robotic finstypically have simple deformationmodes, e.g., bending only (Chen et al.,2010; Aureli et al., 2010), and fallshort of emulating the complex de-formation of biological fins.

Understanding of the morphologyand mechanics of fish fins has spurredeffort on mimicking these features

(typically at a higher level) in designingrobotic fins (Lauder et al., 2007). Inparticular, the complex shape changeof fish fins is enabled by multiplemuscle-controlled, relatively rigid, bonyfin rays that are connected via collage-nous membrane (Lauder & Madden,2006). Coordinated movement of in-dividual fin rays results in conforma-tional changes of the fins desired inmaneuvers. On the engineering side,by patterning electrodes of IPMCmaterials, one can expect to producecomplex deformation by applying dif-ferent voltage inputs to different elec-trode areas. The patterning can beachieved with masking during electro-less plating or by selective removal ofelectrodes post-IPMC fabricationusing laser or machining (Kim et al.,2011). Inspired by the pectoral finsof bluegill sunfish, we have developeda lithography-based monolithic fab-rication process for creating IPMC ac-tuators capable of sophisticated shapechanges (Chen & Tan, 2010). Asshown in Figure 4, the fabricated sam-ple consists of multiple active IPMCregions, coupled throughmuch thinnerpassive regions. By phasing the voltageinputs to different active regions, wecan realize various deformation modesincluding bending, twisting, and cup-ping (Figure 5). For example, a peak-to-peak twisting angle of 16° is achievedwith actuation voltages of 3 V (Chen& Tan, 2010).

While the progress made in biomi-metic fins is encouraging, significantfurther advances in both materialfabrication and fin control are needed,before robotic fish are capable of ma-nipulating the flow in a manner closeto what their biological counterpartsdo. In particular, the materials needto be improved so that they can pro-duce much larger deformation withreasonable bandwidth (a few Hz). On


the control side, we need to model andunderstand the deformation and itshydrodynamic consequences of a giveninput by combining observation ofkinetic patterns of fish fin movement,nonlinear elasticity modeling, com-putational fluid dynamics modeling,and experimental flow measurementsusing digital particle image velocimetry.

Energy-EfficientSustained Operation

For the robotic fish-based sensingtechnology to gain widespread adop-tion, these robots will have to be able

to work continuously in the fieldwith minimal human intervention. Inparticular, they need to operate for atleast weeks, if not for months, beforereturning for battery recharge andother manual maintenance. Power isarguably the most crucial factor thatlimits the operational time. Whilefuel represents a potential energysource with high power and energydensity, it is unclear when fuel-basedpropulsion will become feasible forsmall underwater robots. Therefore,battery is expected to be the primarypower source for robotic fish, for atleast the next 5–10 years.

There are a number of ways onecan potentially extend the run timeof battery-powered robotic fish. Forexample, many onboard devices canbe put to the sleep mode to save en-ergy, when they are not active. Photo-voltaic films can be mounted on therobot to harvest solar energy and re-plenish the battery, when the robot ison the water surface. Wave energycould be another source to tap into,but how to harvest it on an untetheredand often goal-oriented robotic fish re-mains a challenge.

While all the aforementioned ap-proaches could stretch the mileageper battery charge to some extent,they are not game-changers. Designof energy-efficient locomotion mech-anisms will be critical in realizinglong-duration field operation, sincelocomotion is the biggest source ofenergy expenditure for autonomousrobotic fish. To this end, we are cur-rently developing a novel class of un-derwater robots, called gliding roboticfish (Figure 6a). Such a robot will rep-resent a hybrid of underwater gliderand robotic fish; for example, it willhave wings for gliding and fins for ma-neuvering and assistive propulsion.Consequently, a gliding robotic fishis expected to possess both high energyefficiency and great maneuverability.

Figure 6b further illustrates thegliding principle and why a gliding ro-botic fish will be energy-efficient.Under the combined influence of grav-ity and buoyancy, the body will expe-rience vertical (up or down) motion.When the glider is properly pitched,the lift generated during buoyancy-induced vertical motion will enablehorizontal travel. Through the controlof pitch direction and buoyancy, onecan switch between the descent/ascentglidingmotion, resulting in a sawtooth-shaped trajectory. Since buoyancy

FIGURE 5

Examples of deformation modes demonstrated by the fabricated IPMC fin: (a) bending and(b) twisting.

FIGURE 4

Monolithically fabricated IPMC sample inspired by fish fins: (a) top view and (b) SEM picture of thecross section, showing that the passive area ismuch thinner than the active area (Chen& Tan, 2010).


control and pitch control are the majorsources of energy expenditure and takeplace only during ascent/descentswitching, the motion is very energy-efficient, especially if the dive depthis relatively large.

Communication andLocalization

Robotic fish need to communicatewith a base station to receive com-mands and send back the collected en-vironmental information. They alsoneed to communicate with eachother for information relay andmotioncoordination. Underwater communi-cation, however, is particularly chal-lenging for small robotic fish thathave stringent power and size con-straints. Radio frequency (RF) signalsattenuate quickly in water, severelylimiting the achievable communica-tion range and data rate. Light com-munication is possible (Verzijlenberg& Jenkin, 2010), but again the rangeand data rate are very limited and itdoes not work in a turbid environ-ment. Acoustic and sonar communica-tion underwater has been studied formany decades and was recently ex-plored for communication among ro-botic fish (Science Daily, 2008).However, the associated power and

hardware required to achieve reason-ably large communication distanceand data rate are typically not afford-able by small robotic fish. For thesereasons, the most viable solutionwould be to communicate when therobot surfaces, in which case low-power, low-cost RF communicationprotocols such as ZigBee can be readilyused. Unlike the Bluetooth protocol,which is intended for eliminating cablesbetween electronic devices, the ZigBeeprotocol is built on top of the IEEE802.15.4 standard and it targets specifi-cally wireless sensor network applica-tions. For wider range communication,cellular networks could also be em-ployed if such networks are available.

Limiting the communication to thewater surface entails additional chal-lenges in robotic fish coordination,control, and networking. For effective

networking, we need to have a suffi-cient number of nodes on the surface.This can be achieved through jointmotion planning and control. For ex-ample, we can hold robotic fish onthe surface until a network with ade-quate density and coverage is formedand completes data transmission.

Localization is another challenge inrobotic fish-based sensor networks.For small robotic fish, having onboardlocalization capability is essential forsuccessful navigation of the robot andfor effective coordination of roboticfish networks. Accurate localization isalso critical for tagging the sensed in-formation so that the data collectedby robotic fish are associated correctlyto the physical location in water.Whilethe GPS is readily available and doesnot take upmuch space, its typical pre-cision of 5–10 m is inadequate formany applications of robotic fish dueto their small size and relatively lowspeeds (50 cm/s or less). In addition,the GPS may take a few minutes tolock satellites every time the robotemerges from underwater, which se-verely limits the networking and controlperformance. More agile and preciselocalization technology is needed.

We have developed an efficient lo-calization scheme for small robotic fish(Shatara & Tan, 2010). As illustratedin Figure 7a, the scheme is based on

FIGURE 7

Underwater acoustic ranging-based localization: (a) schematic of the ranging protocol and(b) localization performance in a pool test (Shatara & Tan, 2010).

FIGURE 6

Energy-efficient gliding robotic fish: (a) the concept of a gliding robotic fish with a hydrodynamicgliding body and a caudal fin and (b) illustration of the gliding principle.


acoustic ranging, which measures thetime it takes an acoustic signal to travelfrom one node to the other. For exam-ple, node 1 simultaneously sends anRF packet and an acoustic pulse tonode 2. When node 2 starts its on-board timer when it receives the RFpacket and then stops its timer whenit detects the acoustic pulse. Since theRF signal travels much faster thanthe acoustic signal, we can estimatethe distance between the two nodesbased on the timer reading. Thescheme involves simple hardware, abuzzer and a microphone, for eachnode. A sliding discrete Fourier trans-form algorithm, implemented on adigital signal controller, is employedfor the detection of arrival of the acous-tic signal. Figure 7b shows the resultsfrom experiments in a swimmingpool, where a small robotic fish wastowed across the deep side of thepool (about 13 m long) while its dis-tances to the two beacon nodesmounted on the pool wall were mea-sured through acoustic ranging. Theresulting localization error was lessthan 1 m for the entire tested range,which was a significant improvementover the precision of a commercialGPS.

Note that the above localizationscheme works only when the robotsurfaces, since it involves RF commu-nication. The location of a robot whenit is underwater can be inferred usingdead reckoning. The scheme inFigure 7 does require beacon nodes(whose locations are known) to obtainthe absolute location of a node. In theabsence of such beacon nodes, thescheme can be used to get relative loca-tions among nodes, which is of interestin coordinating schools of robotic fish.There are many other in-air localiza-tion schemes for wireless sensor net-works, e.g., ranging based on received

signal strength. While these schemescan be adapted for robotic fish-basedaquatic networks, care must be takento address the challenges associatedwith noises, disturbances, and signalattenuation at the air/water interface.

Autonomous Controland Coordination

With onboard communication,navigation, control, and sensing de-vices, robotic fish are desired to operateautonomously, as individuals and asschools, to carry out envisioned moni-toring tasks. A few challenges arise inthe control and coordination of theserobots. A robotic fish needs to handlemultiple functions subject to environ-mental uncertainties and resource con-straints. In particular, the functionscould include sampling the environ-ment, processing and transmittingthe measured data, maintaining net-work connectivity, and controllingits motion. There are various uncer-tainties that interfere with these func-tions, examples of which includemotion perturbations due to wavesand turbulences, imperfect sensormeasurements, and localization errorand communication packet drops.Furthermore, all of these functionscompete for limited onboard comput-ing and power resources. This is a clas-sic multi-objective, multi-constraintoptimization problem, and it demandsa systematic approach to the jointconsideration of control, networking,and sensor fusion. Evolutionary algo-rithms (Deb, 2001), which codifybasic principles of genetic evolution,can offer a promising solution to thismulti-objective optimization problem.

Another challenge lies in coordinat-ing a school of robotic fish. It is in-triguing to deploy groups of robotic

fish that cooperatively perform sensingtasks. In that case, it is often desirablenot to use centralized control, becausethe centralized paradigm would entailprohibitive cost in communication,and it would paralyze the whole net-work if the command node fails.Therefore, individual robotic fish areexpected to communicate only withtheir local neighbors and make deci-sions in a distributed manner. Animalsincluding fish often exhibit coordi-nated collective movement facilitatedby only local interactions, which hasinspired great interest from the con-trols community in analyzing and syn-thesizing control laws for groups ofunmanned vehicles. Significant prog-ress has been made in this area, evenwith some demonstrated success inadaptive sampling using under-water gliders (Leonard et al., 2007).While these accomplishments canprovide a sound starting point for thecontrol and coordination of roboticfish schools, we need to recognizemany new and subtle difficultiesfaced by the latter. For example, a ro-botic fish can only communicate withits peers when it surfaces, which ren-ders communication and feedbackintermittent and asynchronous. Thisagain points to the need to jointly con-sider control, communication, andnetworking issues.

Other ChallengesAs sensor platforms, the potential

of robotic fish in environmental sens-ing will be ultimately limited by theavailability of versatile sensors that arecompact and easy to interface with.Most commercial sensors availabletoday are not amenable to integrationinto small robotic fish, since sensormanufacturers have mostly been tar-geting handheld, fixed, or buoyed


sensors where miniaturization is notcritical. It is expected that, with the de-velopment of robotic fish and wirelessnetworking technologies, manufac-turers will see the growth opportuni-ties in robotic fish-enabled aquaticsensing and start investing in the devel-opment of compact, economical, androbust aquatic sensors.

There are other engineering chal-lenge, one example of which is bio-fouling, where microorganisms andother organisms accumulate on thesurface of robotic fish and their sen-sors, degrading movement and sens-ing performance. Periodic cleaning isan option; thanks to the mobility ofrobotic fish, access to these robots isrelatively easy. Another possibility isto apply anti-fouling coatings.

ConclusionIn this paper, we have explored the

potential of small robotic fish asmobile sensor platforms for aquaticand marine environments. Realizationof this vision poses a rich set of chal-lenges across a wide spectrum of areas,such as actuation/sensing materials,mechanism design, communication,control, and packaging. We have re-viewed some of the major challengesand discussed possible routes to over-come them. The list of challenges out-lined in this paper is by no meansexhaustive, but even partial successin addressing them could have far-reaching impact on aquatic envi-ronmental monitoring and otherengineering applications.

AcknowledgmentThis work was supported in part by

the Office of Naval Research undergrant N000140810640 (programmanager Dr. T. McKenna) and the

National Science Foundation un-der grants ECCS 0547131, CCF0820220 , EEC 0908810 , I I S0916720, DBI 0939454, ECCS1050236, and ECCS 1029683. Theauthor gratefully acknowledges thecontributions of many former andcurrent members of the Smart Micro-systems Laboratory at Michigan StateUniversity, for the results and ideaspresented in this paper.

Author:Xiaobo TanSmart Microsystems LaboratoryDepartment of Electrical andComputer EngineeringMichigan State University,East Lansing, MI 48824Email: [email protected]

ReferencesAbdulsadda, A.T., & Tan, X. 2011.

Underwater source localization using an

IPMC-based artificial lateral line. In:

Proceedings of the 2011 IEEE International

Conference on Robotics and Automation,

to appear. 2719-24. Shanghai, China: IEEE.

Abdulsadda, A.T., Zhang, F., & Tan, X.

2011. Localization of source with unknown

amplitude using IPMC sensor arrays. In:

Proceedings of Electroactive Polymer Actuators

and Devices (EAPAD) XIII. Proc. SPIE, eds.

Bar-Cohen, Y., & Carpi, F., 797627(1-11).

San Diego, CA: SPIE or Society of Photo-

Optical Instrumentation Engineers.

Alvarado, P.V., & Youcef-Toumi, K. 2006.

Design of machines with compliant bodies

for biomimetic locomotion in liquid envi-

ronments. Trans ASME J Dyn Syst Meas

Control. 128:3-13. doi: 10.1115/1.2168476.

Anderson, J.M., & Chhabra, N.K. 2002.

Maneuvering and stability performance of a

robotic tuna. Integr Comp Biol. 42:118-126.

doi: 10.1093/icb/42.1.118.

Aureli, M., Kopman, V., & Porfiri, M. 2010.

Free-locomotion of underwater vehicles

actuated by ionic polymer metal composites.

IEEE/ASME Trans Mech. 15(4):603-14.

doi: 10.1109/TMECH.2009.2030887.

Bandyopadhyay, P.R. 2005. Trends in

biorobotic autonomous undersea vehicles.

IEEE J Oceanic Eng. 30:109-39.

doi: 10.1109/JOE.2005.843748.

Beal, D.N., Hover, F.S., Triantafyllou,

M.S., Liao, J.C., & Lauder, G.V. 2006.

Passive propulsion in vortex wakes.

J Fluid Mech. 549:385-402. doi: 10.1017/

S0022112005007925.

Bodznick, D., Montgomery, J., & Tricas,

T.C. 2003. Electroreception: Extracting

behaviorally important signals from noise.

In: Sensory Processing in Aquatic Environ-

ments, eds. Collin, S.P., & Marshall, N.J.,

389-403. New York: Springer-Verlag.

Chen, Z., Shatara, S., & Tan, X. 2010.

Modeling of biomimetic robotic fish pro-

pelled by an ionic polymer-metal composite

caudal fin. IEEE/ASME Trans Mech. 15(3):

448-459. doi: 10.1109/TMECH.2009.

2027812.

Chen, Z., & Tan, X. 2010. Monolithic

fabrication of ionic polymer-metal composite

actuators capable of complex deformation.

Sensor Actuat A: Phys. 157:246-57.

doi: 10.1016/j.sna.2009.11.024.

Chen, Z, Tan, X., Will, A., & Ziel, C. 2007.

A dynamic model for ionic polymer-metal

composite sensors. Smart Mater Struct. 16:

1477-88. doi: 10.1088/0964-1726/16/4/063.

Coombs, S. 2001. Smart skins: Informa-

tion processing by lateral line flow sensors.

Auton Robot. 11:255-61. doi: 10.1023/

A:1012491007495.

Coombs, S., & Braun, C.B. 2003. Infor-

mation processing by the lateral line system.

In: Sensory Processing in Aquatic Environ-

ments, eds. Collin, S.P., & Marshall, N.J.,

122-38. New York: Springer-Verlag.

Dagamseh, A., Lammerink, T., Kolster, M.,

Bruinink, C., Wiegerink, R., & Krijnen, G.


2010. Dipole-source localization using bio-

mimetic flowsensor arrays positioned as

lateral-line system. Sensor Actuat A. 162(2):

355-60. doi: 10.1016/j.sna.2010.02.016.

Deb, K. 2001. Multi-objective Optimization

using Evolutionary Algorithms. Chichester,

UK: John Wiley & Sons.

Doherty, K.W., Frye, D.E., Liberatore,

S.P., & Toole, J.M. 1999. A moored pro-

filing instrument. J Atmos Ocean Technol.

16:1816-29. 2.0.CO;2"target=_blankdoi:-

10.1175/1520-0426(1999)016<1816:

AMPI>2.0.CO;2.

Drucker, E.G., & Lauder, G.V. 2001.

Wake dynamics and fluid forces of turning

maneuvers in sunfish. J Exp Biol. 204:431-42.


Function of pectoral fins in rainbow trout:

behavioral repertoire and hydrodynamic

forces. J Exp Biol. 206:813-26. doi: 10.1242/

jeb.00139.

Epstein, M., Colgate, J.E., & MacIver,

M.A. 2006. Generating thrust with a

biologically-inspired robotic ribbon fin.

In: Proceedings of the 2006 IEEE/RSJ

International Conference on Intelligent

Robots and Systems. 2412-17. Beijing, China:

IEEE. doi: 10.1109/IROS.2006.281681.

Ericksen, C.C., Osse, T.J., Light, R.D.,

Wen, T., Lehman, T.W., Sabin, P.L., …

Chiodi, A.M. 2001. Seaglider: A long-range

autonomous underwater vehicle for oceano-

graphic research. IEEE J Oceanic Eng. 26:

424-36. doi: 10.1109/48.972073.

Fish, F.E., & Lauder, G.V. 2006. Passive

and active flow control by swimming fishes

and mammals. Annu Rev Fluid Mech.

38:193-224. doi: 10.1146/annurev.

fluid.38.050304.092201.

Hu, H., Liu, J., Dukes, I., & Francis, G.

2006. Design of 3D swim patterns for

autonomous robotic fish. In: Proc. 2006

IEEE/RSJ Int. Conf. Intell. Robots Syst.

2406-11. Beijing, China: IEEE.

Hydroid, LLC. 2009. REMUS: Autonomous

underwater vehicles - REMUS AUVs.

Available at http://www.km.kongsberg.

com/ks/web/nokb0240.nsf/AllWeb/

D5682F98CBFBC05AC1257497002976E4?

OpenDocument/ (accessed 5 July 2011).

Kato, N. 2000. Control performance in

the horizontal plane of a fish robot with

mechanical pectoral fins. IEEE J Oceanic

Eng. 25(1):121-9. doi: 10.1109/48.820744.

Kim, K.J., Pugal, D., & Leang, K.K. 2011.

A twistable ionic polymer-metal composite

artificial muscle for marine applications.

Mar Technol Soc J. This issue.

Lauder, G.V., Anderson, E.J., Tangorra, J.,

& Madden, P.G.A. 2007. Fish biorobotics:

Kinematics and hydrodynamics of self-

propulsion. J Exp Biol. 210:2767-80.

doi: 10.1242/jeb.000265.

Lauder, G.V., & Drucker, E.G. 2004.

Morphology and experimental hydrodynamics

of fish fin control surfaces. IEEE J Oceanic

Eng. 29(3):556-71. doi: 10.1109/JOE.

2004.833219.

Lauder, G.V., & Madden, P.G.A. 2006.

Learning from fish: Kinematics and experi-

mental hydrodynamics for roboticists. Int

J Automat Comput. 4:325-35. doi: 10.1007/

s11633-006-0325-0.

Leonard, N.E., Paley, D.A., Lekien, F.,

Sepulchre, R., Fratantoni, D.M., & Davis,

R.E. 2007. Collective motion, sensor

networks, and ocean sampling. Proc IEEE.

95(1):48-74. doi: 10.1109/JPROC.2006.

887295.

Low, K.H. 2006. Locomotion and depth con-

trol of robotic fish with modular undulating

fins. Int J Automat Comput. 4:348-57.

doi: 10.1007/s11633-006-0348-6.

Morgansen, K.A., Triplett, B.I., & Klein,

D.J. 2007. Geometric methods for modeling

and control of free-swimming fin-actuated

underwater vehicles. IEEE T Robot. 23(6):

1184-99. doi: 10.1109/LED.2007.911625.

Pandya, S., Yang, Y., Jones, D.L., Engel, J.,

& Liu, C. 2006. Multisensor processing

algorithms for underwater dipole localiza-

tion and tracking using MEMS artificial

lateral-line sensors. EURASIP J Appl Signal

Processing. 2006:76593. doi: 10.1155/

ASP/2006/76593.

Paquette, J.W., & Kim, K.J. 2004. Ionomeric

electroactive polymer artificial muscle for

naval applications. IEEE J Oceanic Eng. 29(3):

729-37. doi: 10.1109/JOE.2004.833132.

Reynolds-Fleming, J.V., Fleming, J.G., &

Luettich, R.A. 2002. Portable autonomous

vertical profiler for estuarine applications.

Estuaries. 25:142-7. doi: 10.1007/BF02696058.

Rudnick, D.L., Eriksen, C.C., Fratantoni,

D.M., & Perry, M.J. 2004. Underwater gliders

for ocean research. Mar Technol Soc J. 38:

48-59. doi: 10.4031/002533204787522703.

Sarles, S.A., & Leo, D.J. 2011. Hair cell

sensing with encapsulated interface bilayers.

In: Proceedings of Bioinspiration, Biomimetics,

and Bioreplication. Proc. SPIE, eds. Martin-

Palma, R.J., & Lakhtakia, A., 797509(1-11).

San Diego, CA: SPIE.

Science Daily. 2008. School of robofish com-

municate with each other in underwater robot

teams. Science Daily. http://www.sciencedaily.

com/releases/2008/06/080606105454.htm

(accessed 5 July 2011).

Shahinpoor, M., & Kim, K. 2001. Ionic

polymer-metal composites: I. Fundamentals.

Smart Mater Struct. 10:819-33. doi: 10.1088/

0964-1726/10/4/327.

Shatara, S., & Tan, X. 2010. An efficient,

time-of-flight-based underwater acoustic

ranging system for small robotic fish. IEEE

J Oceanic Eng. 35(4):837-46. doi: 10.1109/

JOE.2010.2060810.

Sherman, J., Davis, R.E., Owens, W.B., &

Valdes, J. 2001. The autonomous under-

water glider “spray”. IEEE J Oceanic Eng.

26:437-46. doi: 10.1109/48.972076.

Smithers, M. 2011. Designing future

autonomous underwater vehicles. Defense

Tech Briefs. Available at http://www.

defensetechbriefs.com/component/content/

article/9075 (accessed 5 July 2011).

Tangorra, J., Anquetil, P., Fofonoff, T.,

Chen, A., Del Zio, M., & Hunter, I. 2007.


The application of conducting polymers to a

biorobotic fin propulsor. Bioinspir Biomim.

2:S6-17. doi: 10.1088/1748-3182/2/2/S02.



Sci Amer. 272:64-71. doi: 10.1038/


Verzijlenberg, B., & Jenkin, M. 2010.

Swimming with robots: Human robot

communication at depth. In: Proceedings of

the 2010 IEEE/RSJ International Confer-

ence on Intelligent Robots and Systems.

4023-28. Taipei, Taiwan: IEEE. doi: 10.1109/

IROS.2010.5652751.

Webb, D.C., Simonetti, P.J., & Jones, C.P.

2001. “SLOCUM”: An underwater glider

propelled by environmental energy. IEEE J

Oceanic Eng. 26:447-52. doi: 10.1109/

48.972077.

Yang, Y., Chen, J., Engel, J., Pandya, S.,

Chen, N., Tucker, C., … Liu, C. 2006.

Distant touch hydrodynamic imaging with

an artificial lateral line. Proc Nat Acad

Sci. 103(50):18891-5. doi: 10.1073/

pnas.0609274103.

Yang, Y., Nguyen, N., Chen, N., Lockwood,

M., Tucker, C., Hu, H.,… Jones, D.L. 2010.

Artificial lateral line with biomimetic neuro-

masts to emulate fish sensing. Bioinspir

Biomim. 5:016 001.


P A P E R

Robotic Models for Studying UndulatoryLocomotion in FishesA U T H O R SGeorge V. LauderJeanette LimRyan SheltonMuseum of Comparative Zoology,Harvard University

Chuck WittErik AndersonDepartment of MechanicalEngineering, Grove City College

James L. TangorraDepartment of MechanicalEngineering, Drexel University

A B S T R A C TMany fish swim using body undulations to generate thrust and maneuver in

three dimensions. The pattern of body bending during steady rectilinear locomotionhas similar general characteristics in many fishes and involves a wave of increasingamplitude passing from the head region toward the tail. While great progress hasbeen made in understanding the mechanics of undulatory propulsion in fishes, theinability to control and precisely alter individual parameters such as oscillation fre-quency, body shape, and body stiffness, and the difficulty of measuring forces onfreely swimming fishes have greatly hampered our ability to understand the funda-mental mechanics of the undulatory mode of locomotion in aquatic systems. In thispaper, we present the use of a robotic flapping foil apparatus that allows these pa-rameters to be individually altered and forces measured on self-propelling flappingflexible foils that produce a wave-like motion very similar to that of freely swimmingfishes. We use this robotic device to explore the effects of changing swimmingspeed, foil length, and foil-trailing edge shape on locomotor hydrodynamics, thecost of transport, and the shape of the undulating foil during locomotion. Wealso examine the passive swimming capabilities of a freshly dead fish body. Finally,wemodel fin-fin interactions in fishes using dual-flapping foils and show that thrustcan be enhanced under correct conditions of foil phasing and spacing as a result ofthe downstream foil making use of vortical energy released by the upstream foil.Keywords: fish, robot, swimming, biomechanics

Introduction

Fish moving through the water arecapable of using a variety of locomotormodes. Some species swim nearly ex-clusively using their fins and generatepropulsive and maneuvering forcesusing midline fins (dorsal and anal)or paired fins (pectoral and pelvic)(Drucker & Lauder, 1999, 2001;Hove et al., 2001; Standen, 2008;Standen & Lauder, 2007). Otherspecies generate thrust primarily byactivating body musculature to bendthe body and generate waves passingfrom the head toward the tail (Gillis,1996; Jayne & Lauder, 1994, 1995c;Lauder & Tytell, 2006; Rome et al.,1993). The caudal or tail fin is gener-ally considered as an extension of thebody in most analyses of fish loco-motion, but the tail also possesses asubstantial array of intrinsic musclesthat appear to stiffen the tail duringsteady forward swimming and gener-ate a variety of complex tail conforma-tions during maneuvering (Flammang

& Lauder, 2008; Flammang &Lauder, 2009).

Despite the large number of stud-ies of fish undulatory locomotionover the last 20 years, there are stillmany unanswered questions abouthow the pattern of body deformationis generated, the effect of body stiff-ness on locomotor performance, whateffect different tail shapes have on lo-comotor function, and how hydro-dynamic interactions among differentfins might influence the generationof swimming forces. Studies of freelyswimming fishes have contributedenormously to our understanding ofthe mechanics of aquatic locomotion,but such an approach is necessarilylimited to the behaviors voluntarily

executed by living fishes. And mea-suring locomotor forces on freelyswimming fishes is a challenging prop-osition. Furthermore, a wide array ofinteresting experimental manipula-tions, including changing the flexuralstiffness of the body, altering theshape of the tail, and changing thespacing between fins, are clearly im-possible to conduct in living animals.

Robotics offers a complementaryapproach to studies of living fishes byallowing manipulation of variety ofparameters such as flexural stiffness,aspect ratio, tail shape, and spacing be-tween adjacent fins. A simple roboticflapping foil device can be used to gen-erate undulatory locomotion in flexi-ble fish-like materials, and forces can


be measured and the effect of changesin body length, stiffness and tail shapecan be quantified.

The focus of this paper will be onundulatory locomotion in the waterand the use of a robotic flapping appa-ratus to produce swimming in bothflexible plastic foils and a passivefreshly dead fish body. After a briefoverview of undulatory locomotionin fishes, we discuss the design of a ro-botic controller for flapping flexiblefoils that allows measurement of self-propelled speeds (SPS), forces, andtorques and the cost of transport asso-ciated with foils of different shapes andstiffnesses. In addition, we addressan important technical issue in studiesof aquatic propulsion: the effect ofswimming at non-SPS on body wave-form, patterns of force production andwake flow patterns. Finally, we presentdata on the swimming performanceof passive fish bodies and discuss thefuture for studies of robotically con-trolled undulatory locomotion. Ouroverall aim is to introduce a numberof case studies with data that showthe utility of this approach for studyingunderwater propulsion and to presentan overview of how such studies canprovide new ideas and tests for currentviews of how fish swim.

Overview of UndulatoryPropulsion in Fishes

When fish swim using their bodiesand tail fin as the primary thrust gen-erators, they pass a wave of bendingdown the body that increases in am-plitude from the head toward the tail(Donley & Dickson, 2000; Gillis,1996; Jayne & Lauder, 1995a, 1995b;Liao, 2002; Long et al., 1994). Thisbending wave is produced by a waveof muscular activity that also moves

posteriorly and is created by spinalcord and hindbrain pattern generators(Bone et al., 1978; Fetcho & Svoboda,1993; Fetcho, 1986; Shadwick &Gemballa, 2006). Muscular powerto generate thrust is produced pri-marily by the segmented myotomalmusculature in the posterior region offish, especially during slow swimming( Johnson et al., 1994; Rome et al.,1993; Syme, 2006), and as speed in-creases more anterior body musclesare recruited to power locomotion.The division of the segmented bodymusculature into superficial “red” fi-bers and deeper and more complexlyarranged “white” fibers also plays animportant role in understanding themultiple “gaits” used by fishes; the rel-ative roles of these two types of musclefibers can change dramatically as fishchange swimming speed and executerapid locomotor behaviors such asfast-start escapes ( Jayne & Lauder,1993; Tytell & Lauder, 2002).

When fish swim slowly using bodyundulations, oscillation of the frontthird or so of the body is quite smalleven in species as diverse as eels, trout,and tuna (Donley & Dickson, 2000;Gillis, 1998; Lauder & Tytell, 2006),and a primary role of this reduced os-cillation appears to be drag reductionby minimizing the frontal area of thefish that encounters oncoming flow.As swimming speed increases, thefront region of fish shows increasinglylarge side-to-side oscillations, which isin part a reflection of the recruitmentof body musculature in this more an-terior region of the fish.

Even when fish swim by undula-tory propulsion using the productionof traveling waves, other fins oftenare used too, and it is incorrect to sug-gest that fish locomotor modes repre-sent completely distinct patterns ofmotion. For example, when trout or

bluegill sunfish swim using body un-dulations, they are also actively usingtheir dorsal and anal fins, which playa key role in balancing roll torquesand generating thrust (Drucker &Lauder, 2001, 2005; Standen &Lauder, 2005, 2007). In addition,the pelvic fins play an important rolein controlling body stability duringlocomotion in fishes (Harris, 1936,1938; Standen, 2008, 2010). Fishesalso vary in tail shape, and the distinc-tion between the externally symmetri-cal (homocercal) tail of teleost fishesand the asymmetrical tail in sharksand fish such as sturgeon (Lauder,1989, 2000) is well known. The hetero-cercal tail shape induces torques aroundthe body center of mass that requirescompensatory changes in body posi-tion to allow steady horizontal swim-ming (Liao & Lauder, 2000; Wilga& Lauder, 2002, 2004b).

Although considerable progress hasbeen made in studies of fish locomo-tion by investigating the kinematics,muscle activity, and hydrodynamicsof live fishes swimming steadily andmaneuvering, there are many limitsto studies of this kind. Perhaps thetwo greatest limitations to researchon live fishes are (1) the considerabledifficulty in measuring locomotorforces and torques produced by thebending body as fish swim freely and(2) the inability to manipulate keyvariables such as body length, aspectratio, tail shape, and stiffness. Withoutan ability to alter such key componentsthat govern locomotor dynamics, wewill be limited in our understandingof the factors that influence undula-tory propulsion in the water.

The purpose of this paper is to pre-sent a number of new case studiesusing a robotic flapping device forgenerating undulatory locomotion inengineered materials as a means of


better understanding how fish swimand the dynamics of undulatory pro-pulsion. We discuss different examplesto show the utility of this approach andhow use of simple flexible foils informsstudies of fish locomotion and pointsto new avenues of research.

Simple Robotic Models ofFishUndulatory Locomotion

We have designed a robotic appara-tus that produces controlled heave andpitch motions of flapping foils. Themost important features of this deviceare (1) that it can be set up to be self-propelling (allowing locomotion byflapping foils at their natural swim-ming speed and not only at imposedspeeds) and (2) that forces and torquescan be measured on flapping foilsduring self-propelled swimming sothat within-cycle patterns of force andtorque oscillation can be comparedamong foils with different shapesand stiffnesses at different swimmingspeeds. This apparatus was designedwith two sets of flapping foils in seriesin order to be able to model, with foils,the interactions that can occur be-tween fins of fishes that are arrangedin series such as the dorsal and analfins and the tail fin (discussed furtherin the section on fin-fin interactionsbelow).

A general description of the firstgeneration of this apparatus is pre-sented in Lauder et al. (2007). Briefly,heave and pitch motors and rotaryencoders that allow readouts of foil po-sition are mounted on a carriage abovea recirculating flow tank. This carriageis supported on low friction air bear-ings, which allow the carriage to movein response to foil thrust and drag forcesgenerated during flapping motions.The second generation version of thisapparatus has an ATI Nano-17 six-

axis force/torque sensor mounted onthe shaft supporting the foils (Fig-ure 1A). This permits measurementof three forces and three torques duringself-propulsion at sample rates thatallow quantification of within-cyclepatterns of force production even dur-ing self-propulsion. In addition, a lin-ear encoder mounted on the carriageallows a readout of carriage position,

and this is used by a Labview programto calculate SPS from data generatedby a series of swimming tests at arange of speeds. Synchronizing signalsfrom the LabView program control-ling the heave and pitch motors areused to trigger data acquisition fromthe ATI sensor and also to triggerimage acquisition from three syn-chronized Photron high-speed video

FIGURE 1

Images of a variety of flexible foils and freshly dead trout (Oncorhynchus mykiss) attached to arobotic flapping foil apparatus (see Lauder et al., 2007, for details on the basic design). (A) Flexiblefoil actuated at its leading edge in heave and pitch, suspended in a recirculating flow tank. The redarrow points to an ATI Nano-17 6-axis force/torque transducer on the foil shaft. (B, C) Flexible foilswith different trailing edge shapes are used to study the effect of tail shape on swimming performance.(D, E, F) Images of a trout held behind the head to allow imposition of heave and pitch motions tostudy passive body properties. (E) An image from a trout self-propelling under an imposed heavemotion; the body waveform produced is very similar to that generated during swimming.


cameras. Images of the flapping foils toquantify both foil motion and hydro-dynamic flow patterns are thussynchronized with force and torquemeasurements on the foil and withthe imposed heave and pitch motionson the foil.

The overall goals of using a flap-ping foil robotic device are to simplifyand control as much as possible pat-terns of motion imposed on flexibleand rigid foils that swim through thewater and to allow direct testing ofthe effect on propulsion of a varietyof key factors relevant to understand-ing fish locomotion: tail shape (Fig-ures 1B and 1C), foil flexibility, andinteractions among fins. This roboticapparatus can also be used to examinethe passive swimming capabilities of afreshly dead fish body (Figures 1D, 1E,and 1F ), and below we present data onthe ability of fish bodies to passivelypropel and generate propulsive wave-forms under imposed motions. Foilsof various kinds are attached to a stain-less steel sandwich bar (to hold theleading edge of flexible materials; Fig-ures 1A, 1B, and 1C) to a solid 8-mmshaft (for rigid NACA 0012 foils; seeLauder et al., 2007, and data shownin Figure 11), and flexible fish bodiesare mounted in a holder that is at-tached behind the head (Figures 1Dand 1F). Holding systems are designednot to flex in response to imposedheave and pitch motions and to allowforces and torques generated by swim-ming foils to be transmitted to theforce/torque sensor on the shaft. Hold-ing systems such as the sandwich barsystem do have their own drag, andthis drag is time-dependent due tothe heaving and pitching motion ofthe holding system as the foils self-propel. It is thus not possible to givea single value for the drag of the foilholding system. Since all foil compari-

sons within a single experimental typeused the same holding system andweretreated identically, we do not presentdata on the performance of the holdingapparatus alone.

Sample data from a self-propelledflapping foil actuated in heave onlyare shown in Figure 2. Monitoringfoil shaft position andmeasuring forcesin the X (upstream-downstream) andY (side to side) directions allows calcu-lation of the heave velocity and otherderived quantities such as the instan-taneous power required by the foil toswim and the coefficients of thrustand power. The cost of transport is cal-culated by measuring the foil mass anddividing the cost/meter by this valuefor each foil (see Table 1). Typicalpeak thrust coefficients for highly flex-ible foils (flexural stiffness in the rangeof 10−4 to 10−6 N m2) are in the rangeof ±0.2, lower than typical for rigidfoils, but these flexible foils none-theless are capable of self-propulsionat speeds of 10-30 cm s−1.

A key technical issue arises in stud-ies of flapping foil propulsion thathope to imitate the self-propelled con-dition achieved by swimming fishes:foils that are not self-propelling mayexhibit patterns of thrust oscillationthat are not centered around zero.Any foil that is truly self-propelling(and not being dragged through thewater at speeds slower or faster thanit would naturally move) should gener-ate thrust in an oscillatory pattern, andthrust integrated over a single flappingcycle should equal zero. Graphs in theliterature of thrust coefficients duringfoil-based locomotion that are notcentered around zero indicate thatthe foil was being towed above orbelow the SPS and do not reflect theself-propelled condition. Data fromrecordings of foil forces generatedduring self-propulsion and at speedsbelow and above the SPS are shownin Figure 3. During self-propelledswimming, the thrust coefficient hasa mean of zero over each flapping

FIGURE 2

Sample data from a self-propelling flexible plastic foil (flexural stiffness = 9.2 × 10−5 N m2) actu-ated in heave at the leading edge at 2-Hz frequency. Heave position is monitored by rotary encoders,and X andY forces are measured by a force transducer mounted on the foil shaft. From the data onfoil position, force, and velocity, we make calculations of the instantaneous power, and dimension-less thrust and power coefficients. The dashed lines indicate zero for each trace.


cycle. When foils are forced to swimabove their SPS, the thrust coefficientcurves shift above the zero baseline,and when forced swimming occurs atspeeds below the SPS, data are shiftedbelow the baseline.

Change in foil swimming speedabove and below the SPS can alsohave dramatic effects on the kinemat-ics of the foil (Lauder et al., 2011) andon the hydrodynamic wakes displayedby swimming foils. Figure 4 shows

the shape observed during swimmingof a flexible foil (flexural stiffness,3.1 × 10−6 N m2) and the hydro-dynamic wakes that result from swim-ming at, below, and above the averageSPS. During swimming at an imposedspeed below the SPS, the trailing edgeof the foil has a large amplitude andirregular motion that produces a widebifurcating wake with separate mo-mentum jets to each side (Figure 4A).At the SPS where the foil is allowed to

swim freely with no imposed con-straints on speed, the foil bends intoa regular ribbon-like pattern witha fish-like body wake (see Nauen &Lauder, 2002a, 2002b) and alternatingcenters of vorticity with a fluid jet thatmeanders in between these vorticalcenters (Figure 4B). Above the SPSwhere the external free-stream flowis increased above SPS and the foil isforced to swim against the increasedflow, the foil shape exhibits largeamplitude wave-like motion and asubstantial drag-like wake with fluidvelocities below that of the free streamin the wake (Figure 4C). These chang-ing patterns of foil kinematics andhydrodynamics during swimmingthat are not under conditions of self-propulsion are most easily seen in highlyflexible materials (flexural stiffnesses inthe range of 10−3 to 10−6 N m2) wherethe fluid-structure interaction is mostevident visually. These flexible materi-als which have flexural stiffnesses similarto those of fish (McHenry et al., 1995)show fish-like propulsion and deforminto wave-like patterns with a fish-likewake (Figure 4B) when allowed to self-propel in a low-friction system.

The structure of the wake behindflapping flexible foils has a significantthree-dimensional component dueto the finite chord and span, and we

TABLE 1

Locomotor properties of flexible plastic foils of three lengths while self-propelling.

Foil length SPS (m/s) Reynolds number Strouhal number Work/cycle (mJ) Cost/meter (mJ/m) Cost of transport (mJ/g/m)

20 0.105 21,114 0.831 2.248 42.59 106.47

25 0.101 25,250 0.775 2.257 44.69 89.39

35 0.091 31,850 0.349 2.233 49.08 70.11

The three foils were made of the same material with a flexural stiffness of 3.1 × 10−6 N m2. These highly flexible foils propel at relatively high Strouhal numbers atshorter lengths.Reynolds and Strouhal numbers are dimensionless.Foils were actuated at their leading edge with ±1 cm heave, no pitch, at 2 Hz.Foil span was 6.8 cm for all three foils.Standard errors for all parameters ranged from 0.3% to 1.5% of the mean values in each column.

FIGURE 3

Graph showing the dimensionless thrust coefficient versus time for a flexible plastic foil (flexuralstiffness = 9.2 × 10−5 N m2), 20 cm long, 6.8 cm high, actuated in heave ±1 cm at the leadingedge at 2-Hz frequency. The red curve shows data for the self-propelled condition (18.8 cm/s),during which the thrust coefficient integrated over a single flapping cycle equals zero. Note thatwhen experiments are done under non-self-propelling conditions (green and blue curves) theplots shift up or down so that the integrated coefficient over a flapping cycle is no longer zero.Data shown have been digitally filtered with a bandpass filter. Strouhal number for this exper-iment = 0.3 at the SPS (red curve). (Color versions of figures available online at: http://www.ingentaconnect.com/content/mts/mtsj/2011/00000045/00000004.)


quantified this aspect of foil locomotordynamics using the volumetric flowvisualization system described byFlammang et al. (2011a, 2011b) andTroolin and Longmire (2010). Fig-ure 5 shows how each of the centers

of vorticity in the ribbon-like patternshown in Figure 4B actually representvortical columns on each side of thefoil, which connect to each other aboveand below the foil. These inter-columnconnections occur both to upstream

and downstream columns on thesame side and also across the foil tovortical columns on the opposite side(Figure 5B). To date no other studieshave provided three-dimensionalwake snapshots during self-propulsionin highly flexible foils, but such studiesin the future may reveal interestingpatterns of wake interaction amongcomponents of the foil and may con-tribute to explaining the dynamics ofpropulsion in the flexing bodies.

These results also suggest that atleast some of the diversity of fishwakes reported in the literature resultsfrom situations in which the fisheswere not swimming steadily, eitherin still water or against imposedflows, as wakes that look like thosein Figures 4A and 4C are frequentlypresented. Fish commonly accelerateand execute small maneuvers duringlocomotion and considerable effort isneeded to ensure that kinematics andwake flow patterns are taken at mo-ments when the fish is swimmingsteadily. Even small accelerations cansubstantially change fish wake flows(Tytell, 2004), and data from theflapping foil robot here illustrate thatthese kinematic and hydrodynamic al-terations can be reproduced with flex-ible foils.

One of the most straightforwardquestions that could be asked aboutundulatory propulsion and one thatis difficult to study with live fishes isthe effect of changing length alone onlocomotor performance.We clearly can-not alter fish length experimentallyand expect reasonable swimming per-formance, and comparing fish ofdifferent lengths (while useful for stud-ies of scaling) does not account for themany other changes in the muscula-ture and skeleton that occur as fishgrow. Table 1 shows data obtainedfrom three flexible foils of different

FIGURE 4

Hydrodynamics of propulsion in a flexible foil (flexural stiffness = 3.1 × 10−6 N m2) swimmingbelow, at, and above its SPS (8.55 cm/s). The foil was actuated at the leading edge with amplitude±0.5 cm, no pitch, at 3Hz, and is 20 cm long; Strouhal number at the SPS is 0.83 (see Table 1). Thebottom margin of the foil is marked in white. Yellow arrows indicate water velocity; red color in-dicates counterclockwise vorticity; blue color indicates clockwise vorticity. The panels on the leftshow the whole foil swimming, while the matched panels on the right show a close-in view of thewake structure. The effect of swimming at a non SPS is dramatic, both on foil shape and on wakestructure. In (A) the foil swam at an imposed speed of 4.75 cm/s, and in (C) the foil swam at animposed speed of 25.7 cm/s.


lengths made of the same material andactuated in heave only at the leadingedge. Altering the length of this flexibleswimming foil from 20 to 35 cm pro-duces only minor changes in the SPS

(from 10 to about 9 cm s−1) andhence in Reynolds number but dra-matically lowers the Strouhal number(from 0.83 to 0.34; Table 1) due tochanges in foil trailing edge amplitude.

The work per cycle stays nearly con-stant, as does the cost per meter(Table 1), but the cost of transportdecreases substantially as the increasedmass of the longer foils does not resultin increased energy requirements forpropulsion. Foils such as these thatare composed of a very flexible materialself-propel at shorter lengths with aStrouhal number that is quite largerelative to most self-propelling bodies.At longer lengths, the Strouhal numberapproaches that of many swimmingfishes or rigid foils (Table 1).

Figure 6 shows changes in shapeof self-propelling foils made of thesame material (flexural stiffness, 3.1 ×10−6 N m2) that occur due to change

FIGURE 5

Volumetric flow visualization using the V3V technique (see Flammang et al., 2011a, 2011b) toimage the 3D wake structure behind the flexible foil shown in Figure 4B. The trailing edge ofthe foil is located at the −60 mm position on the x-axis, and the fluid structures shown are allin the wake of the foil. This foil was self-propelling and was actuated using the same parametersshown for Figure 4B. (A) The flexible foil achieved a ribbon-like shape and columns of vorticityextend vertically on either side of the foil. Vorticity is isosurfaced at a value of 3.1 (blue surface),and a horizontal slice through the wake is shown to correspond to that shown using 2D piv inFigure 4B (green plane with velocity vectors). (B) The vortical columns on each side of theribbon-like foil motion connect to each other across the top and bottom of the same side (yellowarrows) and opposite sides (red arrows).

FIGURE 6

Graphs to show the effect of length on theshape (amplitude envelope) of a flexible foil(flexural stiffness = 3.1 × 10−6 N m2) swim-ming at its SPS. This foil was 6.85 cm inchord, and of varying length, given in thecolor-coded legend. The leading edge was ac-tuated at 2 Hz and with amplitude of ±1 cm.Each plot shows the shape of the foil duringself-propulsion as indicated by the peak-to-peak amplitude of the sideways flapping mo-tion. (A, B) Foil shapes as the absolute distancealong the foil and as percentage of the total foillength, respectively.


in length. Longer foils show peak ex-cursions at the same locations as short-er foils (Figure 6A) but amplitudes thatare lower. Each of the foils of this ma-terial generates a ribbon-like shapewhen self-propelling with a consistentwave-like pattern down its length (seeFigure 4B). When the amplitude ofeach foil as a percentage of foil lengthis plotted (Figure 6B), changes inamplitude are more evident withside-to-side excursion amplitudes de-creasing as length increases while thewave-like pattern is retained. Theshortest foils have high amplitudesnear the trailing edge and an amplitudeenvelope that grows along the foilwhile the longer foils display a taperingamplitude envelope (compare blue andblack curves in Figure 6B). These datashow that length alone can have signif-icant effects on locomotor efficiencyand foil kinematics and that, all otherthings held constant, foil length in-creases on the order of 100% can pro-vide reduced costs of transport.

The Effect of TrailingEdge Shape onSwimming Performance

The shape of the trailing tail edgeof swimming fishes has been the sub-ject of considerable discussion in theliterature, with various authors con-sidering the advantages or disadvan-tages of fish tails with symmetrical,asymmetrical, or forked shapes (Affleck,1950; Aleev, 1969; Lauder, 1989;Plaut, 2000; Thomson, 1971, 1976;Wilga & Lauder, 2004a). One advan-tage of a robotic flapping foil approachis that the trailing edge of a flexibleflapping foil can be altered and a vari-ety of configurations constructed thatallow the effect of trailing edge shapealone on locomotor performance to

be investigated. We designed five dif-ferent flexible foils (material flexuralstiffness = 3.1 × 10−4 N m2) that con-trol for total length and foil area andallow different tail shapes to be com-pared for the effect of these changeson swimming speed (Figure 7). Foil 1corresponds to a highly abstracted“trout-like” body shape with a mostlyvertical trailing tail edge, while foils 3and 4 present a shark-like tail trailingedge shape. Many fish have forkedtails, and this shape is represented byfoil 5.

The fastest swimming foil (Fig-ure 7, foil 4) is the one with themost area near the axis of actuation,even though it has the same area asseveral of the other foils. This resultcorresponds with our previous resultsshowing that foils with higher aspectratios and hence more material nearthe actuator swim significantly faster(Lauder et al., 2011). Interestingly,the foil with the angled trailing edgeand the shark tail shape (Figure 7,foil 3) swims significantly faster (P <0.003) than a foil with the same

area but a straight trailing edge (Fig-ure 7, foil 1). The foil shape withthe significantly lowest swimmingspeed was the notched shape (Fig-ure 7, foil 5) even though it possessesthe same area as foils 1 and 3. The re-duced performance of the notchedshape may be due to bending of theupper and lower “lobes” of the tailduring the flapping motion, and kine-matic data obtained for these foilsdoes show that each lobe of this shapetwists during the flapping cycle. Twist-ing of the tail could be reduced by in-troducing stiffening elements, and thismay be one reason that fishes withhigh-performance tail shapes such astuna possess substantial stiffening ofboth the upper and lower tail lobes(Fierstine & Walters, 1968; Westneat& Wainwright, 2001).

Although the angled foil shapeswims significantly faster than a foilof the same area with a straight trailingedge (Figure 7: compare foils 1 and 3),more power is required for this foil toswim at this (self-propelled) speed(Figure 8A). The foils were made of

FIGURE 7

Propulsion by flexible foils of different shapes (material flexural stiffness = 3.1 × 10−4 N m2) swim-ming at their SPS. Each foil was actuated at ±1 cm heave at 2 Hz. Error bars are ±2 SE. Strouhalnumbers for these experiments range from 0.2 to 0.3. Foils were constructed to be of differingshapes and areas as follows: foil 1= square trailing edge, area = 131.2 cm2, dimensions = 6.85 cm ×19.15 cm; foil 2 = angled trailing edge, same length as foil 1, area = 107.71 cm2; foil 3 = angledtrailing edge, same area as foil 1; foil 4 = angled trailing edge: same length and area as foil 1; foil 5 =forked trailing edge: same area as foil 1.


the same material and have the samearea, so the cost of transport can becalculated in mJ/m. Figure 8B showsthat there is no significant difference(P = 0.07) between the foils in cost oftransport, although there is a trend inthe data toward the angled foil edgecosting less per meter resulting in themarginally non-significant difference.Further experiments on both foilswill be needed to extend this resultand to determine if in fact there is asmall difference in cost of transport be-tween these two foil types.

Undulatory Locomotionof a Passive Fish Body

Although flapping flexible foilsprovide a reasonable and simplemodel for undulatory propulsion in

fishes that minimizes the complexityof the foil and maintains constant ma-terial properties along the foil length,fish bodies are clearly different inshowing changing material propertiesfrom head to tail (McHenry et al.,1995). To better understand the loco-motor properties of the passive fishbody alone, we used freshly dead rain-bow trout (Oncorhynchus mykiss) andattached the body to the robotic flap-ping foil apparatus (Figures 1D, 1E,and 1F). We took care to ensurethat rigor mortis had not set in duringthe experiments and to ensure thatthe body had thus not stiffened dur-ing the time the flapping trials wereconducted. By actuating the passivetrout body in heave and pitch in var-ious combinations just behind thehead, we were able to construct aswimming performance surface forthe passive fish body (Figure 9).Body waveforms that are remarkablylike those occurring in live troutwere generated by the passive fishbodies. These data show that heaveamplitude overall has the greater ef-fect on swimming performance than

changes in pitch alone. At any givenheave value increases in pitch furtherincrease swimming speed, but athigher heave amplitudes near ±2 cm,changes in pitch only produce modestincreases in SPS (Figure 9). Thesedata provide an interesting compari-son to our previous results showingthe effects of heave and pitch actuationon flexible foil propulsion (Lauderet al., 2011). In that study addingpitch motions to baseline heave actu-ation for foils of varying flexural stiff-ness did not produce significantincreases in foil SPS but did allowstiffer foils to maintain a relativelyhigh swimming speed that wouldhave declined with heave actuationonly.

As driving frequency increases, thetail beat amplitude of the passive flap-ping trout body remains relativelyconstant from 0.1 to 2.0 Hz beforeincreasing steadily to a peak at3.5 Hz (Figure 10). These tail beatamplitude values are very similar tothose observed in live trout swimmingunder a variety of locomotor con-ditions (Liao et al., 2003a; Webb,

FIGURE 8

Graphs of power consumed (A) and the cost oftransport (B) of foils 1 and 3 (see Figure 7) self-propelling. Both foils have the same surfacearea (131.2 cm2) and were actuated with±1 cm heave at 2 Hz. Power values are signifi-cantly different at P = 0.003. Cost of transportvalues are not significantly different at P = 0.07.

FIGURE 9

Performance surface of a freshly dead trout (Oncorhynchus mykiss) attached to a robotic control-ler (see Figures 1D, 1E, 1F) driven at 2 Hz at a variety of heave and pitch amplitudes. This trout was25.3 cm in total length. The graph shows how SPS of the passive trout body varies with differentpitch and heave actuation parameters.


1971; Webb et al., 1984) and in-creases in tail beat amplitude of themagnitude shown here for the passivetrout body are similar to those ob-served previously as fish increaseswimming speed and alter both fre-quency (primarily) and amplitude(to a lesser extent) of the tail beat.

Although during routine undula-tory swimming fish bodies are notpassive and are certainly stiffened bybody musculature as fish swim, forall but the fastest swimming speedslocomotion is powered by red musclefibers, which can make up a rathersmal l percentage of body mass(around 1.5% in largemouth bass;see Johnson et al., 1994). The bulkof the locomotor musculature is com-posed of white fibers that are not ac-tivated until the fastest swimmingspeeds are needed (Jayne & Lauder,1994, 1995a). Thus, the red musclefibers can be considered as acting tobend a mostly passive fish body com-posed of white muscle fibers and asso-ciated skeletal tissues that may not bemuch different in flexural stiffnessfrom the passive bodies studied here.

Also, data on the propulsion of pas-sive fish bodies are relevant to fishesswimming in turbulent flows. Troutswimming in a vortex street have beenshown to greatly alter body kinematicsand to utilize vortical energy shed fromobjects in the flow (Liao, 2004; Liaoet al., 2003b). The amplitude of thecenter of mass oscillation by troutswimming in the Karman gait is up to±2 cm for a 10-cm-long trout, whichcorresponds on a length-specific basisto themiddle region of the performancesurface in Figure 9 (Liao et al., 2003a).By greatly reducing muscle activitywhen trout enter a vortex street, thebody becomes largely passive and de-forms in response to the oncomingflows. Trouts are able to maintaintheir position in such flows entirely pas-sively and allow their bodies to extractenergy from the oncoming flow andgenerate thrust. This phenomenonwas further demonstrated in a studyusing passive foils and freshly deadfish bodies to show that the passivetrout body alone in a vortex wake cangenerate sufficient thrust to maintainposition in flow (Beal et al., 2006).

Fin-Fin HydrodynamicInteractions

One of the most intriguing aspectsof fish functional design is the arrange-ment of two fins in series that couldallow enhanced locomotor efficiencythrough hydrodynamic interactionsbetween the fins. For example, thedorsal fin and the anal fin in most fishesare located upstream of the caudal fin,and the wakes shed from these finscould interact with the tail fin duringlocomotion (Drucker & Lauder, 2001,2005; Tytell, 2006). Undulatory loco-motion using the body also causes theattached median fins to oscillate fromside to side, and in the fish species stud-ied so far these fins have been shown alsoto be actively oscillated by intrinsic finmusculature (Jayne et al., 1996) andthus can generate thrust on their own.

The possibility of fin-fin hydrody-namic interactions during locomotionhas been explored in a number ofexperimental papers on living fishes(Drucker & Lauder, 2001, 2005;Standen, 2008; Standen & Lauder,2007; Tytell, 2006; Webb & Keyes,1981) as well as using computationalapproaches (Akhtar et al., 2007; Weihset al., 2006). Akhtar et al. (2007) useddata on the kinematics of the bluegilldorsal fin and tail from Drucker andLauder (2001) and performed a two-dimensional computational analysis ofthe effect of having the fins in seriesand found that vortex shedding fromthe dorsal fin can increase thrust of thetail and that the amount of this thrustincrease depends on the phasing ofdorsal fin and tail motion.

In order to experimentally evaluatehydrodynamic fin-fin interactionsusing an apparatus in which phasingand the distance between fins can beexperimentally manipulated, we usedour robotic flapping foil apparatusin the dual-foil configuration with

FIGURE 10

Graph of tail-tip amplitude versus heave actuation frequency for a freshly dead trout (Oncorhynchusmykiss) attached to a robotic controller driving the passive body at a variety of frequencies. This troutwas 25.3 cm in total length and was actuated with a constant heave (±2 cm) and pitch (±5°). Errorbars are ±1 SE of the mean.


upstream and downstream foils. Foilswere rigid aluminum plates in aNACA 0012 airfoil shape, and eachfoil could be moved in heave andpitch, and both foils were driven froma common carriage mounted on airbearings as described above. Foils weremoved according to the parametersmeasured for bluegill sunfish dorsaland caudal fins (Drucker & Lauder,2001) and used by Akhtar et al.(2007) for their computational studyof this problem: the upstream foil wasmoved with heave amplitude of2.5 cm, the downstream foil at3.5 cm heave amplitude. Pitch ampli-tude for both foils was ±20°, and thefrequency for both foils was 1.7 Hz,corresponding to the frequency of finflapping in the bluegill sunfish modelcase (Drucker & Lauder, 2001).

Increases in SPS as a result ofchanging foil phase and distance re-flect thrust enhancement beyond thethrust achieved by the two foils oper-ating separately (assessed by offsettingthe foils from each other so that thedownstream foil was no longer inthe wake of the upstream foil). A gen-eral description of the dual-foil con-figurat ion and images of flowsaround and between the foils are pre-sented in Lauder et al. (2007). Datafor SPS were plotted over a range ofphase differences between the up-stream and downstream foils andpolynomial fits to these data pointswere used to determine the changein SPS with phase (e.g., Figure 11).

Here we present the results of ex-periments measuring the SPS of dualfoil flapping. Figure 11a shows the ef-fect of changing the phase of sinu-soidal motion of the downstreamfoil relative to the upstream foil onSPS for three different spacings be-tween the foils. In each case, there isa clear peak in swimming speed, and

as the distance between the foils is in-creased the peak swimming speedshifts toward a larger phase lag ofthe downstream foil. Interestingly,the maximal swimming speed of thetwo foils together does not change sig-nificantly as the distance between foilschanges, and alterations in phasingbetween the foils can thus be usedto compensate for changes in spacing.

For each interfoil spacing, however,there are phase relationships that sig-nificantly reduce swimming perfor-mance, indicating that thrust of theentire two-foil system is sensitive tophase relationships between two foilsflapping. These data correspond verywell to the computational results ofAkhtar et al. (2007), which showedpeak thrust enhancement at a phase

FIGURE 11

Robotic model of fish fin hydrodynamic interactions. Dual NACA 0012 flapping foils in series aredriven by a robotic flapping foil apparatus. Details of this device and of the experimental setup aregiven in Lauder et al. (2007). (A) SPS plotted against the phase difference between the upstreamand downstream foils. Foils were arranged with 0 side-to-side offset (in cm) and are thus movingin line with each other with the downstream foil at varying chord length separations from theupstream foil: 0.5, 1, and 2 chord lengths separation; foil chord length was 6.85 cm. Furtherexperimental details are provided in the text. (B) Effect of changing the foil offsets (in cm) sothat the downstream foil is not moving in the direct wake of the upstream foil. The one chord lengthspacing with zero offset curve (dark blue) is the same as in panel (A) and is shown for reference.The other three plots show the effect of a 3.5 cm offset of the midline motion of the downstreamfoil relative to the upstream foil, removing it from the upstream foil wake.


of about 40°, very similar to our data(Figure 11A, blue curve) where theplateau around the SPS peak includesthe 40° value.

Figure 11B illustrates the changesin propulsion that result from movingthe downstream foil to the side by anoffset of 3.5 cm to move it out of thewake of the upstream foil. This ef-fectively creates a tandem foil con-figuration in which fluid dynamicinteractions between the foils areminimized, although not completelyeliminated. Comparison of the threeoffset curves to the zero offset curveshows that offsetting the foils reducesboth the SPS and the effect of foilphasing. Offset plots show reducedeffects of foil phasing (with lowermaxima and higher minima) and lessdistinct overall peaks in SPS, showingthat the downstream foil was not ableto improve swimming speed of thetwo foils together when removedfrom the upstream foil wake.

Computational work and prelimi-nary flow visualization (Lauder et al.,2007) data indicate that the behaviorof tandem foils can be explained interms of how the downstream foil inter-acts with fluid structures generated bythe upstream foil. The shifting of thepeak SPS with increased spacing ofthe tandem foils can be explained bythe fact that, for larger spacings, the con-vection of those structures to the down-stream foil takes longer. That time, tc, issimply the spacing between the foils, s,divided by the convection velocity, Uc.The increase in phase lag, Δϕ, neededso that the downstream foil meets thefluid structures at the right time is simply

Δϕ ¼ 2πfS2 S1ð ÞUc

where f is the frequency of flappingand s2 and s1 are two different spacings.

If we assume that Uc is close to theSPS, this equation estimates the Δϕbetween the peak SPS well for thethree spacings we used. The equationpredicts Δϕ = 0.648 radians, or 37°,for a spacing difference of 0.5 chordlengths, and 74° for a difference ofone chord length. The Δϕ from ourdata are approximately 30°and 75° forthe corresponding spacing differences.The nearness of the foils and/or higherconvection velocities at the smallerspacings may explain the lower thanpredicted Δϕ for the peaks in SPS inthese cases. Overall, the agreement isvery good considering that the peaksare somewhat broad and the equationfor Δϕ is a simplification of the fluiddynamics.

The Future ofUndulatory Biorobotics

In this paper, we use a robotic toolfor investigating a variety of phenom-ena relating to undulatory propulsionin fishes and present experimentaldata that would be difficult if not im-possible to obtain from studying liveanimals. The promise of robotic mod-els for studying the biomechanics oflocomotion in fishes has just begunto be realized (Curet et al., 2011;Long et al., 2006, 2010; Tangorraet al., 2010, 2011), and fundamentalquestions relating to the mechanics ofundulatory propulsion remain to beaddressed. In particular, key unre-solved issues are the extent to whichchanges in body stiffness during pro-pulsion affect locomotor performance(see Long & Nipper, 1996) and howactive modulation of stiffness duringan undulatory cycle and acrosschanges in swimming speed areachieved and affect propulsive speedand efficiency.

To address these questions, a newgeneration of robotic undulatory de-vices will be needed that allow forcontrolled modulation of body stiff-ness and the phasing of stiffnesschanges with undulatory cycles ofcompression and tension on thebending fish or foil body. An addi-tional arena that is key to makingprogress in understanding undulatorymechanics is the ability to perturb thelocomotor system to assess how stiff-ness of the body relates to the abilityto recover from perturbations. Therehave been very few studies of pertur-bations of undulatory locomotor sys-tems (see Webb, 2004), and yet fishesoften swim in challenging hydrody-namic environments in which theyare forced to recover from impul-sive challenges to their undulatorypattern.

AcknowledgmentsThis work was supported the Of-

fice of Naval Research grant N00014-09-1-0352 on fin neuromechanicsmonitored by Dr. Thomas McKennaand by the National Science Foun-dation grant EFRI-0938043. Wethank the members of the Lauderand Tangorra labs for many helpfuldiscussions on fish fins and flexibleflapping foil propulsion and NateJackson for his assistance with thedual-flapping foil experiments. Manythanks to Brooke Flammang, TysonStrand, and Dan Troolin for assis-tance with the V3V volumetric flowimaging experiments on the undulat-ing plastic foil.

Lead Author:George V. LauderThe Museum ofComparative Zoology


26 Oxford Street, Harvard UniversityCambridge, MA 02138Email: [email protected]

ReferencesAffleck, R.J. 1950. Some points in the

function, development, and evolution of the

tail in fishes. Proc Zool Soc Lond. 120:349-68.

doi: 10.1111/j.1096-3642.1950.tb00954.x.

Akhtar, I., Mittal, R., Lauder, G.V., &

Drucker, E. 2007. Hydrodynamics of a

biologically inspired tandem flapping foil

configuration. Theor Comp Fluid Dyn.

21:155-70. doi: 10.1007/s00162-007-0045-2.

Aleev, Y.G. 1969. Function and Gross

Morphology in Fish. Jerusalem: Keter Press.

Beal, D.N., Hover, F.S., Triantafyllou,

M.S., Liao, J., & Lauder, G.V. 2006.

Passive propulsion in vortex wakes.

J Fluid Mech. 549:385-402. doi: 10.1017/

S0022112005007925.

Bone, Q., Kiceniuk, J., & Jones, D.R. 1978.

On the role of the different fibre types in fish

myotomes at intermediate swimming speeds.

Fish Bull. 76:691-9.


&MacIver,M.A. 2011. Aquatic manoeuvering

with counter-propagating waves: A novel

locomotive strategy. J R Soc Interface.

8:1041-50. doi: 10.1098/rsif.2010.0493.

Donley, J., & Dickson, K.A. 2000. Swim-

ming kinematics of juvenile Kawakawa tuna

(Euthynnus affinis) and chubmackerel (Scomber

japonicus). J Exp Biol. 203:3103-16.


Locomotor forces on a swimming fish: Three-

dimensional vortex wake dynamics quantified

using digital particle image velocimetry.

J Exp Biol. 202:2393-412.


Locomotor function of the dorsal fin in teleost

fishes: Experimental analysis of wake forces

in sunfish. J Exp Biol. 204:2943-58.


Locomotor function of the dorsal fin in

rainbow trout: Kinematic patterns and hy-

drodynamic forces. J Exp Biol. 208:4479-94.

doi: 10.1242/jeb.01922.

Fetcho, J., & Svoboda, K.R. 1993. Fictive

swimming elicited by electrical stimulation

of the midbrain in goldfish. J Neurophysiol.

70:764-80.

Fetcho, J.R. 1986. The organization of the

motoneurons innervating the axial muscula-

ture of vertebrates: I. Goldfish (Carassius

auratus) and mudpuppies (Necturus maculosus).

J Comp Neurol. 249:521-50. doi: 10.1002/

cne.902490408.

Fierstine, H.L., & Walters, V. 1968. Studies

in locomotion and anatomy of scombroid

fishes. Mem Southern Calif Acad Sci. 6:1-31.

Flammang, B.E., & Lauder, G.V. 2008.

Speed-dependent intrinsic caudal fin muscle

recruitment during steady swimming in

bluegill sunfish, Lepomis macrochirus. J Exp

Biol. 211:587-98. doi: 10.1242/jeb.012096.

Flammang, B.E., & Lauder, G.V. 2009.

Caudal fin shape modulation and control

during acceleration, braking and backing

maneuvers in bluegill sunfish, Lepomis

macrochirus. J Exp Biol. 212:277-86.

doi: 10.1242/jeb.021360.

Flammang, B.E., Lauder, G.V., Troolin,

D.R., & Strand, T. 2011a. Volumetric

imaging of shark tail hydrodynamics reveals

a three-dimensional dual-ring vortex wake

structure. P Roy Soc B. 278. doi: 10.1098/

rspb.2011.0489.

Flammang, B.E., Lauder, G.V., Troolin,

D.R., & Strand, T.E. 2011b. Volumetric

imaging of fish locomotion. Biol Lett.

doi: 10.1098/rsbl.2011.0282.

Gillis, G.B. 1996. Undulatory locomotion

in elongate aquatic vertebrates: Anguilliform

swimming since Sir. James Gray. Am Zool.

36:656-65.

Gillis, G.B. 1998. Environmental effects on

undulatory locomotion in the American eel

Anguilla rostrata: Kinematics in water and

on land. J Exp Biol. 201:949-61.

Harris, J.E. 1936. The role of the fins in the

equilibrium of the swimming fish: I. Wind

tunnel tests on a model of Mustelus canis

(Mitchell). J Exp Biol. 13:476-93.

Harris, J.E. 1938. The role of the fins in the

equilibrium of the swimming fish: II. The role

of the pelvic fins. J Exp Biol. 16:32-47.

Hove, J.R., O’Bryan, L.M., Gordon, M.S.,

Webb, P.W., & Weihs, D. 2001. Boxfishes

(Teleostei: Ostraciidae) as a model system for

fishes swimming with many fins: Kinematics.

J Exp Biol. 204:1459-71.

Jayne, B.C., & Lauder, G.V. 1993. Red and

white muscle activity and kinematics of the

escape response of the bluegill sunfish during

swimming. J Comp Physiol A. 173:495-508.

doi: 10.1007/BF00193522.

Jayne, B.C., & Lauder, G.V. 1994. How

swimming fish use slow and fast muscle fibers:

Implications for models of vertebrate muscle

recruitment. J Comp Physiol A. 175:123-31.

doi: 10.1007/BF00217443.

Jayne, B.C., & Lauder, G.V. 1995a. Are

muscle fibers within fish myotomes activated

synchronously? Patterns of recruitment within

deep myomeric musculature during swim-

ming in largemouth bass. J Exp Biol.

198:805-15.

Jayne, B.C., & Lauder, G.V. 1995b. Red

muscle motor patterns during steady swim-

ming in largemouth bass: Effects of speed

and correlations with axial kinematics.

J Exp Biol. 198:1575-87.

Jayne, B.C., & Lauder, G.V. 1995c.

Speed effects on midline kinematics during

steady undulatory swimming of largemouth

bass, Micropterus salmoides. J Exp Biol.

198:585-602.

Jayne, B.C., Lozada, A., & Lauder, G.V.

1996. Function of the dorsal fin in bluegill

sunfish: Motor patterns during four locomotor

behaviors. J Morphol. 228:307-26.

doi:10.1002/(SICI)1097-4687(199606)

228:3<307::AID-JMOR3>3.0.CO;2-Z.

Johnson, T.P., Syme, D.A., Jayne, B.C.,

Lauder, G.V., & Bennett, A.F. 1994. Modeling

red muscle power output during steady and un-

steady swimming in largemouth bass (Micropterus

salmoides). Am J Physiol. 267:R481-8.


Lauder, G., Madden, P.G.A., Tangorra, J.,

Anderson, E., & Baker, T.V. 2011.

Bioinspiration from fish for smart material

design and function. Smart Mater Struct.

in press.

Lauder, G.V. 1989. Caudal fin locomotion

in ray-finned fishes: Historical and functional

analyses. Am Zool. 29:85-102.

Lauder, G.V. 2000. Function of the caudal

fin during locomotion in fishes: Kinematics,

flow visualization, and evolutionary patterns.

Am Zool. 40:101-22. doi: 10.1668/0003-

1569(2000)040[0101:FOTCFD]2.0.CO;2.

Lauder, G.V., Anderson, E.J., Tangorra, J.,

& Madden, P.G.A. 2007. Fish biorobotics:

Kinematics and hydrodynamics of self-

propulsion. J Exp Biol. 210:2767-80.

doi: 10.1242/jeb.000265.

Lauder, G.V., & Tytell, E.D. 2006. Hydro-

dynamics of undulatory propulsion. In:

Fish Biomechanics. Volume 23 in Fish

Physiology, eds. Shadwick, R.E., & Lauder,

G.V., 425-68. San Diego: Academic Press.

Liao, J. 2002. Swimming in needlefish

(Belonidae): Anguilliform locomotion with

fins. J Exp Biol. 205:2875-84.

Liao, J. 2004. Neuromuscular control of trout

swimming in a vortex street: Implications for

energy economy during the Karman gait.

J Exp Biol. 207:3495-506. doi: 10.1242/

jeb.01125.

Liao, J., Beal, D.N., Lauder, G.V., &

Triantafyllou, M.S. 2003a. The Kármán gait:

Novel body kinematics of rainbow trout

swimming in a vortex street. J Exp Biol.

206:1059-73. doi: 10.1242/jeb.00209.

Liao, J., & Lauder, G.V. 2000. Function

of the heterocercal tail in white sturgeon: Flow

visualization during steady swimming and

vertical maneuvering. J Exp Biol. 203:3585-94.

Liao, J.C., Beal, D.N., Lauder, G.V., &

Triantafyllou, M.S. 2003b. Fish exploiting

vortices decrease muscle activity. Science.

302:1566-9. doi: 10.1126/science.1088295.

Long, J.H., Jr., Koob, T.J., Irving, K.,

Combie, K., Engel, V., Livingston, N., …

Schumacher, J. 2006. Biomimetic evolution-

ary analysis: Testing the adaptive value of

vertebrate tail stiffness in autonomous swim-

ming robots. J Exp Biol. 209:4732-46.

doi: 10.1242/jeb.02559.

Long, J.H., McHenry, M.J., & Boetticher,

N.C. 1994. Undulatory swimming: How

traveling waves are produced and modulated

in sunfish (Lepomis gibbosus). J Exp Biol.

192:129-45.

Long, J.H., & Nipper, K.S. 1996. The

importance of body stiffness in undulatory

propulsion. Am Zool. 36:678-694.

Long, J.H., Porter, M.E., Root, R.G., &

Liew, C.W. 2010. Go reconfigure: How

fish change shape as they swim and evolve.

Integr Comp Biol. 50:1120-39. doi: 10.1093/

icb/icq066.

McHenry, M.J., Pell, C.A., & Long, J.A.

1995. Mechanical control of swimming speed:

Stiffness and axial wave form in undulating

fish models. J Exp Biol. 198:2293-305.

Nauen, J.C., & Lauder, G.V. 2002a. Hydro-

dynamics of caudal fin locomotion by chub

mackerel, Scomber japonicus (Scombridae).

J Exp Biol. 205:1709-24.

Nauen, J.C., & Lauder, G.V. 2002b.

Quantification of the wake of rainbow trout

(Oncorhynchus mykiss) using three-dimensional

stereoscopic digital particle image velocimetry.

J Exp Biol. 205:3271-9.

Plaut, I. 2000. Effects of fin size on swim-

ming performance, swimming behavior and

routine activity of zebrafish Danio rerio.

J Exp Biol. 203:813-20.

Rome, L.C., Swank, D., & Corda, D. 1993.

How fish power swimming. Science. 261:

340-3. doi: 10.1126/science.8332898.

Shadwick, R., & Gemballa, S. 2006.

Structure, kinematics, and muscle dynamics in

undulatory swimming. In: Fish Biomechanics.

Volume 23 in Fish Physiology, eds. Shadwick,

R.E., & Lauder, G.V., 241-280. San Diego:

Academic Press.

Standen, E.M. 2008. Pelvic fin locomotor

function in fishes: Three-dimensional kine-

matics in rainbow trout (Oncorhynchus mykiss).

J Exp Biol. 211:2931-42. doi: 10.1242/

jeb.018572.

Standen, E.M. 2010. Muscle activity and

hydrodynamic function of pelvic fins in

trout (Oncorhynchus mykiss). J Exp Biol.

213:831-41. doi: 10.1242/jeb.033084.

Standen, E.M., & Lauder, G.V. 2005. Dorsal

and anal fin function in bluegill sunfish

(Lepomis macrochirus): Three-dimensional

kinematics during propulsion and maneuver-

ing. J Exp Biol. 205:2753-63. doi: 10.1242/

jeb.01706.

Standen, E.M., & Lauder, G.V. 2007.

Hydrodynamic function of dorsal and anal

fins in brook trout (Salvelinus fontinalis).

J Exp Biol. 210:325-39. doi: 10.1242/

jeb.02661.

Syme, D.A. 2006. Functional properties of

skeletal muscle. In: Fish Biomechanics.

Volume 23 in Fish Physiology, eds. Shadwick,

R.E., & Lauder, G.V., 179-240. San Diego:

Academic Press.

Tangorra, J., Phelan, C., Esposito, C., &

Lauder, G. 2011. Biorobotic models of highly

deformable fins for studies of the mechanics

and control of fin forces. Integr Comp Biol.

in press. doi: 10.1093/icb/icr036.

Tangorra, J.L., Lauder, G.V., Hunter, I.,

Mittal, R., Madden, P.G., & Bozkurttas, M.

2010. The effect of fin ray flexural rigid-

ity on the propulsive forces generated

by a biorobotic fish pectoral fin. J Exp

Biol. 213:4043-54. doi: 10.1242/

jeb.048017.

Thomson, K.S. 1971. The adaptation and

evolution of early fishes. Q Rev Biol. 46:

139-66. doi: 10.1086/406831.

Thomson, K.S. 1976. On the heterocercal

tail in sharks. Paleobiology. 2:19-38.

Troolin, D., & Longmire, E. 2010. Volumetric

velocity measurements of vortex rings from

inclined exits. Exp Fluids. 48:409-20.

doi: 10.1007/s00348-009-0745-z.

Tytell, E.D. 2004. Kinematics and hydro-

dynamics of linear acceleration in eels,


Anguilla rostrata. P Roy Soc Lond B.

271:2535-40. doi: 10.1098/rspb.2004.2901.

Tytell, E.D. 2006. Median fin function

in bluegill sunfish, Lepomis macrochirus:

Streamwise vortex structure during steady

swimming. J Exp Biol. 209:1516-34.

doi: 10.1242/jeb.02154.

Tytell, E.D., & Lauder, G.V. 2002. The

c-start escape response of Polypterus senegalus:

Bilateral muscle activity and variation during

stage 1 and 2. J Exp Biol. 205:2591-603.

Webb, P.W. 1971. The swimming energetics

of trout: I. Thrust and power output at

cruising speeds. J Exp Biol. 55:489-520.

Webb, P.W. 2004. Response latencies to

postural disturbances in three species of

teleostean fishes. J Exp Biol. 207:955-61.

doi: 10.1242/jeb.00854.

Webb, P.W., & Keyes, R.S. 1981. Division

of labor between median fins in swimming

dolphin (Pisces: Coryphaenidae). Copeia.

1981:901-04. doi: 10.2307/1444198.

Webb, P.W., Kostecki, P.T., & Stevens,

E.D. 1984. The effect of size and swimming

speed on the locomotor kinematics of rainbow

trout. J Exp Biol. 109:77-95.

Weihs, D., Ringel, M., & Victor, M. 2006.

Aerodynamic interactions between adjacent

slender bodies. AIAA J. 44:481-4.

doi: 10.2514/1.18902.

Westneat, M., & Wainwright, S.A. 2001.

Mechanical design for swimming: Muscle,

tendon, and bone. In: Tuna: Physiology,

Ecology, and Evolution, eds. Block, B., &

Stevens, E.D., 271-311. San Diego: Academic

Press. doi: 10.1016/S1546-5098(01)19008-4.

Wilga, C.D., & Lauder, G.V. 2002. Function

of the heterocercal tail in sharks: Quantitative

wake dynamics during steady horizontal

swimming and vertical maneuvering.

J Exp Biol. 205:2365-74.

Wilga, C.D., & Lauder, G.V. 2004a.

Biomechanics of locomotion in sharks, rays

and chimeras. In: Biology of Sharks and Their

Relatives, eds. Carrier, J.C., Musick, J.A.,

& Heithaus, M.R., 139-64. Boca Raton:

CRC Press.

Wilga, C.D., & Lauder, G.V. 2004b.

Hydrodynamic function of the shark’s tail.

Nature. 430:850. doi: 10.1038/430850a.


P A P E R

Thrust Production in Highly Flexible PectoralFins: A Computational DissectionA U T H O R SSrinivas Ramakrishnan1

ANSYS, Inc.

Meliha BozkurttasFranklin W. OlinCollege of Engineering

Rajat MittalDepartment of Mechanical Engineering,Johns Hopkins University

George V. LauderDepartment of Organismicand Evolutionary Biology,Harvard University

A B S T R A C TBluegill sunfish pectoral fins represent a remarkable success in evolutionary

terms as a means of propulsion in challenging environments. Attempts to mimictheir design in the context of autonomous underwater vehicles have overwhelm-ingly relied on the analysis of steady swimming. Experimental observations ofmaneuvers reveal that the kinematics of fin and wake dynamics exhibit character-istics that are distinctly different from steady swimming. We present a computationalanalysis that compares, qualitatively and quantitatively, the wake hydrodynamicsand performance of the bluegill sunfish pectoral fin for two modes of swimming:steady swimming and a yaw turn maneuver. It is in this context that we comment onthe role that flexibility plays in the success of the pectoral fin as a versatile propulsor.Specifically, we assess the performance of the fin by conducting a “virtual dissec-tion” where only a portion of fin is retained. Approximately 90% of peak thrust forsteady swimming is recovered using only the dorsal half. This figure drops to 70%for the yaw turn maneuver. Our findings suggest that designs based on fin analysisthat account for various locomotion modes can lead to more robust performancethan those based solely on steady swimming.Keywords: computational fluid dynamics (CFD), immersed boundary methods(IBM), bluegill sunfish, biological locomotion

Introduction

Robust design based on naturalsystems is a significant engineeringchallenge. Evolution-based design isinherently a multi-objective optimiza-tion problem. Natural selection putspressure on organisms to producelocomotion abilities that balance com-peting requirements of speed, effi-ciency, and effectiveness. The goalsof ongoing research efforts are to eluci-date the competing requirements thathave enabled the evolution of highlymaneuverable propulsion/locomotionat low speeds. Prominent natural sys-tems of interest are flapping flight inair and aquatic locomotion and a com-mon feature among these systems isthe presence of highly compliant con-trol surfaces. Organisms that employthese models of locomotion appear toexploit the flexibility of their wings/

fins to achieve high maneuverabilityat low speeds. This paper presents theanalysis of one such control surface:the bluegill sunfish pectoral fin.

A typical sunfish pectoral finconsists of 14 fin rays as shown inFigure 1. We see the fin rays num-bered sequentially starting from thedorsal edge (ray 1) to the ventral edge(ray 14). These rays support an asym-metric planform shape for the pectoralfin. Figure 3 shows different frames ofthe sunfish executing a maneuver froma ventral view. The motion of the pec-toral fin and body are captured usingmultiple high-speed video camerassimultaneously operating at 250 ormore frames per second with a 1024 ×1024 resolution (Lauder et al., 2006).The wing surface is digitized at about300 spatial locations at several points

during the fin cycle. Thus, the kine-matics of the fin motion is acquiredfor the simulation. The collaborationwith experimentalists (biologists and

1Work presented here was done while the authorwas a postdoctoral scientist at TheGeorgeWash-ington University prior to joining Ansys Inc.

FIGURE 1

Bluegill sunfish pectoral fin consists of14 rays, which form the full planform. Thedissected planform is interpolated from rays1–8 to investigate the flow and performance.


engineers), through a multi-disciplinaryeffort (Lauder et al., 2006;Mittal et al.,2006), has enabled high-fidelity datato be used in the computational analy-sis (see Figure 2).

It is clear from looking at the finmotion during the maneuver (see Fig-ure 3) that the kinematics of the fininvolves both deformation and transla-tion. This poses severe challenges fortraditional body-fitted computationalmethods. Here, the immersed bound-ary method, with its ability to handlecomplex deforming structures, enablesus to undertake high-fidelity computa-

tional fluid dynamics (CFD) analysis ofthe pectoral fin hydrodynamics (seeComputational Methodology). It hasbeen used to gain valuable insight intopectoral fin hydrodynamics in steadyswimming (Bozkurttas et al., 2009;Dong et al., 2010). The experimentallyobtained steady swimming kinematicswas analyzed, and an efficient reconstruc-tion of the kinematics using properorthogonal decomposition (POD)was obtained. The POD modes usinga combination of the first three modes(hereafter referred to asMode 1 + 2 + 3)were successful in reproducing two

thirds of the full fin kinematics. Moresignificantly, this combination ofmodes was found to retain 92%of the thrust produced using the ac-tual kinematics (Bozkurttas, 2007;Bozkurttas et al., 2009). Further, de-tailed analysis of the pressure distribu-tion over the full fin surface (rays1-14) during steady swimming also re-vealed that most of the thrust was pro-duced by the dorsal part mainly aroundthe spanwise tip region (Bozkurttaset al., 2009; Dong et al., 2010) (see Fig-ure 5). Since different sections of thepectoral fin trace different trajectoriesduring a fin stroke, the contributionof each region of the fin to its overallperformance may not be uniform. Nat-urally, this leads us to the central themeof this paper, the idea of examining thethrust production of different sectionsof the fin. The goal is to enable a virtual“dissection” or “ablation” of the pecto-ral fin dynamics and the effect of thisablation on the fin performance. It isexpected that this will yield useful in-sight into the hydrodynamic functionof the fin in various swimming modes.

ComputationalMethodology

We present a brief description ofthe Cartesian grid-based immersedboundary method for moving bound-aries starting with the governingequations. The three-dimensionalunsteady, viscous incompressibleNavier-Stokes equations are given as

∂ui∂xi

¼ 0

∂ui∂t

þ ∂ uiuj ∂xj

¼ 1ρ∂p∂xi

þ ν∂∂xj

∂ui∂xj

ð1Þ

where i; j = 1, 2, 3, ui are the velocitycomponent, p is the pressure, and ρ

FIGURE 3

A bluegill sunfish during a maneuver: ventral (bottom) view. Images are frames from a high-speed video. Note the differential motion of the left and right side fins. Top row: t/T = 0, t/T =0.23, t/T = 0.30. Bottom row: t/T = 0.46, t/T = 0.70, t/T = 0.84.

FIGURE 2

Bioinspired design paradigm.


and ν are the fluid density and kine-matic viscosity. We have employed aconventional notation where repeatedindices imply summation.

1. Numerical MethodThe Navier-Stokes equations(Eq. 1) are discretized using a cell-centered, collocated (non-staggered)arrangement of the primitivevariables (ui, p). In addition to thecell-centered velocities (ui), theface-centered velocities, Ui, arecomputed. A second-order Adams-Bashforth scheme is employed forthe convective terms while the dif-fusion terms are discretized usingan implicit Crank-Nicolson schemewhich eliminates the viscous sta-bility constraint. The spatial de-rivatives are computed using asecond-order accurate central dif-ference scheme. The equationsare integrated in time using thefractional step method (Chorin,1967). In the first sub-step of thismethod, a modified momentumequation is solved and an interme-diate velocity u* obtained. The sec-ond sub-step requires the solutionof the pressure correction equa-tion which is solved with the con-straint that the final velocity ui

n+1

be divergence-free. This gives aPoisson equation for the pressurecorrection and a Neumann bound-ary condition imposed on this pres-sure correction at all boundaries.This Poisson equation is solvedwith a highly efficient geometricmultigrid method which employsa Gauss-Siedel line-SOR smoother.Once the pressure correction is ob-tained, the pressure and velocity areupdated (see Dong et al., 2006 andMittal et al., 2008, for additionaldetails). These separately updatedface velocities satisfy discrete massconservation to machine accuracy

and use of these velocities in estimat-ing the non-linear convective fluxleads to a more accurate and robustsolution procedure. The advantageof separately computing the face-centered velocities was initially pro-posed by Zang et al. (1994) anddiscussed in the context of theCartesian grid methods in Ye et al.(1999) and Mittal et al. (2008).2. Immersed Boundary TreatmentThe immersed boundary methodused he re employs a mul t i -dimensional ghost cell methodologyto impose the boundary conditionson the immersed boundary. Thecurrent solver is designed from thestart for fast, efficient, and accuratesolution of flows with complexthree-dimensional, moving bound-aries. Also, the current method isa “sharp“ interface method in thatthe boundary conditions on theimmersed boundary are imposedat the precise location of the im-mersed body, and there is no spuri-ous spreading of boundary forcinginto the fluid as what usually oc-curs with diffuse interface methods(Mittal & Iaccarino, 2005).3. Geometric Representationof Immersed BoundaryThe current method is designed tosimulate flows over arbitrarily com-plex 2D and 3D immersed station-ary and moving boundaries and theapproach chosen to represent theboundary surface should be flexibleenough so as not to limit the typeof geometries that can be handled.A number of different approachesare available for representing thesurface of the immersed bound-ary, including level sets (Osher &Sethian, 1988; Tran&Udaykumar,2004), and unstructured surfacegrids. In the current solver, wechoose to represent the surface of

the immersed boundary by an un-structured mesh with triangular ele-ments. This approach is very wellsuited for the wide variety of engi-neering and biological configura-tions that are of interest to us andis compatible with the immersedboundary methodology used inthe current solver.4. Boundary MotionBoundary motion can be includedinto immersed boundary formula-tion with relative ease. In advancingthe field equations from time leveln to n + 1 in the case of a movingboundary, the first step is to movefrom its current location to thenew location. This is accomplishedby moving the nodes of the surfacetriangles with a known velocity.Thus, we employ the followingequation to update the coordinates(Xi) of the surface element vertices,

X nþ1i X n

i

Δt¼ V nþ1

i ð2Þ

where Vi is the vertex velocity. Thevertex velocity can either be pre-scribed or it can be computedfrom a dynamical equation if thebody motion is coupled to thefluid. The next step is to determinethe ghost cells for this new im-mersed boundary location andrecompute interpolation weightsassociated with the ghost pointmethodology. Subsequently, theflow equations, which are writtenin Eulerian form, are advancedin time. The general frameworkdescribed above can, therefore, beconsidered as Eulerian-Lagrangian,wherein the immersed boundariesare explicitly tracked as surfaces ina Lagrangian mode, while the flowcomputations are performed on afixed Eulerian mesh. Additional


details regarding the current im-mersed boundary methodologymay be found inMittal et al. (2008).

Computational SetupAll simulations are conducted in

a rectangular computational domain.The boundary conditions on thebounding box of the domain are free-stream on the left (x direction), out-flow on the right while the remainingboundaries (top and bottom ( y direc-tion) and front and back (z direction))employ slip boundary conditions (seeFigure 4). The fin surface and fishbody are considered as no-slip bound-aries. The fins are treated as deformingmembranes while the body, where ap-plicable, is treated as rigid body under-going general motion. The Reynoldsnumber in the present work is definedas Re = ULs/ν where U, Ls , and ν arethe swimming velocity, spanwise finlength, and the kinematic viscosity of

water (ν = 1.007 × 10−6 m2 s−1 at roomtemperature), respectively.

Based on a swimming speed of1.1 body length per second, theReynolds number for the steady swim-ming is 6300. However, a compari-son of the force coefficients obtainedat Re = 1440 with those at the experi-mental Reynolds number appear to bein good agreement both quantitativelyand qualitatively (Bozkurttas, 2007).So, for computational expediency, weuse the lower Reynolds number in thesteady swimming analysis (Dong et al.,2010). As mentioned earlier, low di-mensional model performance anal-yses have shown that Mode 1 + 2 + 3gait that accounts for 67% of the finmotion still produces 92% of thethrust (Bozkurttas et al., 2009). There-fore, in lieu of the experimentallyextracted fin kinematics, this simpli-fied model has been used here. Thegrid size in these simulations is 153 ×161 × 97, which is about 2.35 million

grid points. A domain size of 3.8Ls ×4.5Ls × 1.8Ls is selected where Ls isthe span wise size of the fin. Compre-hensive studies have been carried outto assess the effect of the grid reso-lution and domain size on the salientfeatures of the flow and also to demon-strate the accuracy of the selected grid(Bozkurttas, 2007).

The Reynolds number for the turn-ing maneuver based on a freestreamvelocity of 0.5 body lengths per secondis approximately 3500. The domainsize employed for the maneuver is7.5Ls × 5Ls × 5Ls. The pectoral finsand an idealized body, immersed inthe computational grid, are shown inFigure 4. The nominal grid size usedin the current simulation is 241 ×145 × 145 (see Figure 4). Finally, thedomain size for the maneuver withjust the strongside (outside) fin is4Ls × 4Ls × 4Ls with a non-uniformgrid using 128 points in all threedimensions.

We note in passing that all the steadyswimming cases and ablated fin sim-ulations (for the maneuver) do not in-clude the fish body. This is reasonablesince we have observed that the dif-ference in the thrust coefficients withand without the body is minimal. Aswe shall see shortly, the wake dynamicsfor both steady swimming and maneu-ver are dominated by vortex structuresgenerated far from the fish body (seeFigures 5 and 8). Thus, the interactionbetween the body and the fin hydrody-namics is minimal.

The performance of the fin is eval-uated using the computed force co-efficients which are defined as,

CT ¼ 2FxρU 2

∞Afin;CL ¼

2FyρU 2

∞Afin;

CZ ¼ 2FzρU 2

∞Afinð3Þ

FIGURE 4

Cartesian grid (4.8 million grid points) and unstructured mesh employed for yaw maneuver:(a) x-y plane section, (b) x-z plane section, (c) y-z plane section (strongside fin on the leftand weakside on right of the body), and (d) unstructured surface mesh (pectoral fin only, num-ber of nodes = 10,000, number of elements = 19,602).


where Fx, Fy and Fz are the forces re-spectively in the streamwise (drag/thrust), vertical ( lift), and spanwise(lateral) directions, Afin is the nominalfin area, and ρ is the density of thefluid. U∞ is the forward swimmingvelocity. The force components arecalculated by directly integrating thecomputed pressure and shear stresson the fin surface.

ResultsSteady Swimming

A snapshot of the vortex dynamicsat the end of a steady swimming finbeat is shown in Figure 5. Note the

complex interaction of among vorticesgenerated by the path traversed by thefin tip during a fin beat. Clearly, bothadduction and abduction appear toproduce distinct vortex structures.This is in stark contrast with a simplering vortex created during the maneu-ver (see Figure 8). The time variationsof the force coefficients (CT, CL andCZ) for three fin planforms are plottedin Figure 6. Note the presence of twodistinct and comparable peaks cor-responding to the adduction and ab-duction phases. This force signaturebears the trademark of efficiency wherethe fin sustains net forward thrustthroughout its fin beat. Clearly, thechordwise and spanwise complianceof the fin allows the simultaneous for-mation and persistence of two distinctvortex structures within a single finbeat. A rigid planform would lead toa more restrictive envelope for the fintip path resulting in vortex dynamicsthat have stronger interactions detri-mental to sustained net thrust produc-tion (see Akhtar et al., 2007).

We now construct two differentablated fin models: one that containsonly the rays 1-4 and one that containsrays 1-8 (see Figure 1). The motion ofthese dissected fins is precisely the

same as that for the full fin and wecarry out flow simulations for both ofthese cases. Examining the results fromour virtual dissection, we notice thatthe dorsal half of the fin (rays 1-8) cap-tures the two main peaks of the thrustand preserves 90% of the thrust pro-duction of the full fin planform. Con-sequently, the ventral contribution ofthe fin, represented by rays 9-14 inFigure 1, to the thrust production isfound to be insignificant. Also, theplanform interpolated from rays 1-4has a similar trend in thrust variationduring the entire fin-beat cycle albeitwith smaller amplitudes. Interestingly,it has twomain peaks and even the twolocal peaks in the abduction phase as inthe full fin case. This further reinforcesthe notion that the dorsal leading edgeof the bluegill’s pectoral fin dominatesthe overall performance during steadyswimming propulsion. This planformproduces almost 40% of the thrustproduced by the fish fin while under-going Mode 1 + 2 + 3 gait. Finally,we observe similar tendencies for liftand spanwise force coefficients for thethree planforms, except the case withjust rays 1-4 where the values show at-tenuation. The key observation here isthat the dorsal half of the pectoral fin

FIGURE 6

Comparison of time variation of force coefficients for three different fin planforms (rays 1–4, rays 1–8, full planform) at Mode 1 + 2 + 3 gait:(a)streamwise force, (b) vertical force, and (c) lateral force.

FIGURE 5

The anatomy of the principal vortex dynamicsinvolved in steady swimming.


(rays 1-8) is responsible for producinga majority of the thrust. These resultsbring into question the need for theventral portion of the fin. We explorethis in detail as we consider the case ofthe yaw turn maneuver.

Yaw Turn ManeuverThe evolution of wake structure

from the strongside fin, that drivesthe maneuver, is shown from two van-tage points: lateral (Figure 8 (a,c,e))and dorsal (Figure 8 (b,d,f )). Thewell-defined vortex ring formed duringthe outstroke (abduction) produces alateral jet oriented normal to the fishbody (see Figure 9). This type of vortexring and associated lateral jet shown inFigures 8 and 9 have also been ob-served in experimental visualization(Drucker & Lauder, 2001). The peaklateral velocity is found to be greaterthan three times the freestream veloc-ity. Consequently, the lateral forcesdeveloped are several times that ob-served in forward thrust for the steadyswimming case (Bozkurttas, 2007).Preliminary estimates for stroke-aver-aged force coefficients ratio betweenlateral force in maneuvering (

―CZ =

6.1) to steady swimming thrust (―CT =

1.29) is approximately 4 ((‐) denotesaverage over stroke). This factor is inreasonable agreement with the forcesmeasured experimentally (Drucker &Lauder, 2001).

Returning to Figure 7(a), we notethat the CZ peak is reached betweent/T = 0.15 and t/T = 0.3. Shortly there-after, the CT peak occurs betweent/T = 0.3 and t/T = 0.4. As expected,the first priority in the maneuver isto evade the stimulus (an obstacle orpredator in the wild) by quickly gener-ating a strong lateral force (maximumoccurs at t/T = 0.2). Thereafter, thedrag force developed in the streamwisedirection is likely used to modulate the

direction of the resultant force as thesunfish turns away from the stimulus.The evolving vortex ring, clearly seenin Figures 8(d) and 8(f ), continues tobe oriented nearly parallel to the fishbody. Consequently, the lateral jet ori-entation ensures that the maximumlateral force continues to act normalto the fish body for the duration ofthe maneuver. Here, the inherent flex-ibility of the pectoral fin structure andthe ability to continuously alter plan-form area is likely to be very useful.

Finally, an examination of the forcehistories for the dissected fin revealsthat the peak lateral thrust developedby the dorsal part (rays 1-8) is approxi-mately 70% of the total as opposed to30% for the ventral (rays 8-14) portion(see Figure 10 and Figure 11c). Thestreamwise drag is slightly more compa-rable, although the dorsal part peak ishigher (see Figure 11a). Overall, whilethe dorsal portion contributes to the ma-jority of lateral force production, the ratioof dorsal to ventral contribution appearsto be more equitable than the steadyswimming case.

ConclusionsA comparative analysis of the pec-

toral fin performance in steady swim-

ming and yaw turn maneuver revealsthat the dorsal part of the pectoral finis responsible for the majority of forceproduction. The chordwise and span-wise flexibility of the pectoral fin andits ability to have them function eitherin concert or independently seems toenable the bluegill sunfish to achievea variety of maneuvers. The virtualdissection reveals a significant loss ofperformance with maneuvering withrespect to peak lateral thrust whenthe ventral portion is removed. Thus,a fin design using just the dorsal por-tion of the pectoral fin might performas well as the full fin in steady swim-ming but will not retain the samemaneuverability. Hence, any effectivedesign based on the pectoral fin thataims to preserve its performance overall locomotion mode needs to retaina greater portion of the fin than thatsuggested by steady swimming alone.

The pectoral fins of fishes display adiversity of shapes (e.g., Drucker &Lauder, 2002; Thorsen & Westneat,2005), and although some generalconclusions about correlations of finshape with fish ecology have been pos-sible (see Wainwright et al., 2002),there are very few data on functionalregionalization of pectoral fins and onthe role that different fin rays within

FIGURE 7

Comparison of time variation of force coefficients: (a) strongside and (b) weakside.


FIGURE 8

Formation of the vortex ring due to the strongside pectoral fin motion: (a), (c), and (e) are lateral views at t/T = 0.22, t/T = 0.49, and t/T = 0.66,respectively; (b), (d), and (f) are the corresponding dorsal views at t/T = 0.22, t/T = 0.49, and t/T = 0.66, respectively.


the pectoral fin might play in control-ling locomotor performance. Taft et al.(2008) discussed functional regional-ization during steady swimming insculpin, but the role that different finrays play during maneuvering behav-iors has not previously been analyzed.The results presented here suggest thatthe ventral region of the fin plays animportant role in modulating maneu-vering forces, and future studies on thediversity of fish pectoral fin shapescould focus on the surface area andmechanical properties of this regionof the fin in correlation with maneu-vering performance. No data are cur-rently available that would permiteven general conclusions about the di-versification of pectoral fin structure inrelation to maneuvering capability,and this represents a new and very in-teresting direction for future work thatintegrates approaches from biome-chanics and fluid dynamics with be-havioral and ecological studies of fishlocomotion.

AcknowledgmentsThis work was done while the first

three authors were at The George

FIGURE 9

The strongside lateral jet associated with the vortex structures in Figure 8 (c) at t/T = 0.49.

FIGURE 10

Formation of the vortex ring due to the strongside pectoral fin motion: (a) full, (b) dorsal, and(c) ventral portion of the fin sections.

FIGURE 11

Comparison of forces produced on the dorsal and ventral halves of the strongside fin with respect to the full fin: (a) streamwise force, (b) verticalforce, and (c) lateral force.


Washington University, and the workwas supported under ONR-MURIgrant N00014-03-1-0897 monitoredby Dr. Thomas McKenna.

Lead Author:Srinivas RamakrishnanANSYS, Inc.10 Cavendish Court,Lebanon, NH 03766Email: [email protected]

ReferencesAkhtar, I., Mittal, R., Lauder, G.V., &

Drucker, E.G. 2007. Hydrodynamics of a

biologically inspired tandem flapping foil

configuration. Theor Comp Fluid Dyn.

21:155-70. doi: 10.1007/s00162-007-0045-2.

Bozkurttas, M. 2007. Hydrodynamic perfor-

mance of fish pectoral fins with application to

autonomous underwater vehicles. Ph.D. Thesis,

The George Washington University. p. 9.

Bozkurttas, M., Mittal, R., Dong, H.,

Lauder, G.V., & Madden, P. 2009. A low-

dimensional models and performance scaling

of a highly deformable fish pectoral fin.

J Fluid Mech. 631:311-42. doi: 10.1017/

S0022112009007046.

Chorin, A.J. 1967. A numerical method

for solving incompressible viscous flow

problems. J Comput Phys. 2:1-9.

doi: 10.1016/0021-9991(67)90037-X.

Dong, H., Bozkurttas, M., Mittal, R., Lauder,

G.V., & Madden, P. 2010. A Computational

modelling and analysis of the hydrodynamics

of a highly deformable fish pectoral fin.

J Fluid Mech. 645:345-73. doi: 10.1017/

S0022112009992941.

Dong, H., Mittal, R., & Najjar, F.M. 2006.

Wake topology and hydrodynamic per-

formance of low-aspect-ratio flapping foils.

J Fluid Mech. 566:309-43. doi: 10.1017/

S002211200600190X.

Drucker, E.G., & Lauder, G.V. 2001. Wake

dynamics and fluid forces of turningmaneuvers

in sunfish. J Exp Biol. 204:431-42.

Drucker, E.G., & Lauder, G.V. 2002. Exper-

imental hydrodynamics of fish locomotion:

Functional insights from wake visualization.

Integr Comp Biol. 42:243-57. doi: 10.1093/

icb/42.2.243.

Lauder, G.V., Madden, P.G.A., Mittal, R.,

Dong, H., & Bozkurttas, M. 2006. Locomo-

tion with flexible propulsors: I. Experimental

analysis of pectoral fin swimming in sunfish.

Bioinspir Biomim. 1:35-41. doi: 10.1088/

1748-3182/1/4/S04.

Mittal, R., Dong, H., Bozkurttas, M., &

Lauder, G.V. 2006. Locomotion with flexible

propulsors: II. Computational modeling of

pectoral fin swimming in sunfish. Bioinspir

Biomim. 1:25-34. doi: 10.1088/1748-3182/

1/4/S05.

Mittal, R., Dong, H., Bozkurttas, M., Najjar,

F.M., Vargas, A., & Von Loebbecke, A. 2008.

A versatile sharp interface boundary method

for incompressible flows with complex bound-

aries. J Comput Phys. 227:1-9. doi: 10.1016/

j.jcp.2008.01.028.

Mittal, R., & Iaccarino, G. 2005. Immersed

boundary methods. Annu Rev Fluid Mech.

37:239-61. doi: 10.1146/annurev.fluid.37.

061903.175743.

Osher, S., & Sethian, J.A. 1988. Fronts

propagating with curvature-dependent.

J Comput Phys. 79:12-49. doi: 10.1016/

0021-9991(88)90002-2.

Taft, N., Lauder, G.V., & Madden, P.G.

2008. Functional regionalization of the

pectoral fin of the benthic longhorn sculpin

during station holding and swimming.

J Zool Lond. 276:159-67. doi: 10.1111/

j.1469-7998.2008.00472.x.

Thorsen, D.H., & Westneat, M. 2005.

Diversity of pectoral fin structure and function

in fishes with labriform propulsion. J Morphol.

263:133-50. doi: 10.1002/jmor.10173.

Tran, L.B., & Udaykumar, H.S. 2004. A

particle-level set-based sharp interface cartesian

grid method for impact, penetration, and

void collapse. J Comput Phys. 193:469-510.

doi: 10.1016/j.jcp.2003.07.023.

Wainwright, P., Bellwood, D.R., &Westneat,

M. 2002. Ecomorphology of locomotion in

labrid fishes. Environ Biol Fish. 65:47-62.

doi: 10.1023/A:1019671131001.

Ye, T., Mittal, R., Udaykumar, H.S., &

Shyy, W. 1999. An accurate Cartesian grid

method for simulation of viscous incompres-

sible flows with complex immersed boundaries.

J Comput Phys. 156:209-40. doi: 10.1006/

jcph.1999.6356.

Zang, Y., Streett, R.L., & Koseff, J.R. 1994.

A non-staggered fractional step method for

time-dependent incompressible Navier-

Stokes equations in curvilinear coordinates.

J Comput Phys. 114:18-33. doi: 10.1006/

jcph.1994.1146.


P A P E R

Learning From the Fins of Ray-FinnedFish for the Propulsors of UnmannedUndersea VehiclesA U T H O R SJames L. TangorraDepartment of MechanicalEngineering, Drexel University

Timo GerickeGeorge V. LauderMuseum of Comparative Zoology,Harvard University

A B S T R A C TAdvanced propulsors are required to help unmanned undersea vehicles (UUVs)

overcome major challenges associated with energy and autonomy. The fins of ray-finned fish provide an excellent model from which to develop propulsors that cancreate forces efficiently and drive a wide range of behaviors, from hover to low-speed maneuvers to high-speed travel. Although much is known about the me-chanics of fins, little is known about the fin’s sensorimotor systems or how finsare regulated in response to external disturbances. This information is crucial forimplementing propulsive and control systems that exploit the same phenomena asthe biological fins for efficiency, effectiveness, and autonomous regulation. Experi-ments were conducted to evaluate the in vivo response of the sunfish and its pec-toral fins to vortex perturbations applied directly to the fish and to the fins. The fishand the fins responded actively to perturbations that disturbed themotion of the fishbody. Surprisingly, perturbations that deformed the fins extensively did not cause areaction from either the fins or the body. These results indicate that the response of thepectoral fins to large deformations is not reflexive and that fin motions are regulatedwhen it is necessary to correct for disturbances to the motion of the fish. The resultsalso demonstrate a benefit of compliance in propulsors, in that external perturbationscan disturb the fins without having its impact be transferred to the fish body.Keywords: biorobotics, flapping fins, vortex pertubations, sensory-based control

Introduction

Military and civilian studies haveidentified that two of the most signif-icant technological obstacles to de-ploying unmanned undersea vehicles(UUVs) are energy and autonomy(Nicholson & Healey, 2008; Officeof the Secretary of Defense, 2009).The energy and the rate it is used(power) limit the duration and dis-tance of operations and bound thetype of activity that can occur evenfor short periods. Autonomy definesthe degree to which humans must su-pervise UUV operations and providesUUVs with the ability to react toexternal stimuli without human inter-vention. Among the enabling technol-ogies that are critical for solving thechallenges associated with energy andautonomy are more effective propul-sors (Office of the Secretary Defense,2009). Propulsors are required to pro-vide increased maneuverability, stealth,and endurance for the widespread rangeof missions envisioned forUUVs, fromlong duration sensing in the open oceanto mine countermeasures in very shal-low, high-energy water.

Fish are important biological mod-els from which to learn methods ofpropulsion that are effective and effi-cient over a wide range of operatingconditions. Bony fish, such as thebluegill sunfish (Lepomis macrochirus)and the swordfish (Xiphias gladius),are able to hover, swim and maneuverat low speeds, manipulate the orienta-tion of their bodies, conduct acrobaticsto escape or to attack prey, and, espe-cially for the swordfish, sustain highswimming speeds. These behaviorscan be accomplished in smooth waterand in high-energy flows and relatedirectly to the behaviors desired forUUVs. The remarkable swimming

abilities of these fish are due, in largepart, to the fish havingmultiple, highlyactuated, flexible fins that are able tocreate and to modulate large-magnitudeforces.

A great deal is known about themechanisms that contribute to theproduction of hydrodynamic forcesby flapping the fins and the fishbody. Forces are created through thedynamic interaction of the fins, thebody, and the fluid, which results inenergy being added to, or taken from,the fluid. A review of seminal workthat explains the way in which marineanimals control vorticity is presentedin Triantafyllou et al. (2002) and


Zhu et al. (2002). Numerical andexperimental studies of flexible finswith two-dimensional kinematics(heaving and pitching) include, butare in no way limited to, studiesof McHenry (1995), Liu and Bose(1997), Prempraneerach et al. (2003),Triantafyllou et al. (2005), Fish et al.(2006), Lauder et al. (2006), Mittalet al. (2006), Lauder and Madden(2007), and Zhu and Shoele (2008).Recent studies that considered de-formable fins with complex kinematicsare presented in, for example, Shoeleand Zhu (2009), Dong et al. (2010),and Tangorra et al. (2010).

In contrast to our understandingof the mechanics of fins and of hydro-dynamic forces, little is known abouthow fishes sense their interactionwith the water and use sensory infor-mation to regulate the fins. Knowledgeof fin sensorimotor control is criticalif engineered systems are to take fulladvantage of the mechanisms used byfins to create forces efficiently and toreact to changes in the environment.

The focus of this paper will be onthe pectoral fins of sunfish and, in par-ticular, on how and when sunfish alterthe use of the pectoral fins in responseto external perturbations. We beginwith an overview of pectoral fin swim-ming in sunfish and briefly present ro-botic fins that produce and modulateforces like the biological fins. A seriesof experiments where the biologicalfin is perturbed during steady swim-ming is then presented. These experi-ments address the response of the finsin the context of using the fins to con-trol and stabilize the fish body.

Ray-Finned Fish andRobotsSunfish Swimming

The ability of the sunfish to controlthe magnitude and direction of its pro-

pulsive forces is due to its ability tomodulate the kinematics, coordi-nation, and mechanical properties ofits fins and muscular tail (Tangorraet al., 2010, 2011). Hydrodynamicforces are created through an exchangeof energy between the propulsive sur-faces and the surrounding fluid. As thefish moves through the water, vorticesdevelop along the body and fins, thepropulsive structures bend and storeenergy, and the vortices are shed intothe flow along with directed jets(Triantafyllou et al., 2002; Donget al., 2010). The complex motionsthat cause this exchange of energy arethe result of driven motions of thefin rays and a dynamic interaction ofthe deformable fin surfaces with thewater. The forces created by fins are,therefore, modulated through changesto the kinematics of the fin and activeadjustments of fin’s mechanical prop-erties (Lauder et al., 2006; Mittal et al.,2006; Akhtar et al., 2007; Tangorraet al., 2010). The changes may be sub-tle, as in steady swimming where thestiffness of the fin rays is gradually in-creased with speed, but where the mo-tions of the fins are approximately thesame. Or the changes may be obvious,as when the fish interrupts a cyclicswimming pattern and uses a stiff, im-pulsive fin motion to slow the fish andturn it away from an obstacle (Gottliebet al., 2010).

Ray-Finned Robotic SystemsRobotic fins (Figure 1) have been

developed that produce motions,forces, and flows like the biologicalfins (Tangorra, Davidson et al., 2007;Phelan, Tangorra et al., 2010;Tangorra,Lauder et al., 2010). These fins weredesigned originally as physical modelswith which to conduct experimentalstudies that would have been diffi-cult to conduct with the living fish

(Tangorra, Phelan et al., 2011). Thefins are comprised of fin rays, eachwith multiple actuated degrees of free-dom (DOF), within a thin, flexiblewebbing. The geometries of the finrays were defined so that the stiffnessof the robotic fin was proportional tothat of the biological fin across thefin’s chord and span. The architectureof the robotic fin provides a great de-gree of control over the fin’s motionsand mechanical properties, which en-ables the magnitude and direction ofthe force produced by the fin to beeasily modulated (Figure 2). Grosschanges to the profile of the fin’s forcecan be made by changing the fin’s gait

FIGURE 1

Biorobotic models of the sunfish pectoral fin(A) and caudal fin (B). The pectoral fin is in-strumented with strain gages along the finrays and pressure sensors along the bodyplate in order to model distributed sensing inthe sunfish. Modified versions of the roboticpectoral and caudal fins, as well as dorsal andanal fins, are implemented on a fish robot (C).The fish robot can swim freely or be attachedto a rigid mast (shown) so that forces can bemeasured. The grooves in the side of the fishbody are used for pressure lines and ports.


pattern, for example, by switchingfrom a steady swimming gait to thepattern used by the fish for a turn ma-neuver. Smaller changes to the forceprofile can be made by changing thefrequency of the fin beat and/or bychanging phase relationships betweenfin rays (Figure 2A). Considerablechanges to the magnitude and direc-tion of the force can also be made byadjusting the mechanical propertiesof some, or all, of the fin rays. Whenthe mechanical properties of the finrays are under active control, as inthe fish, changes to the force profilecan happen very quickly since thedriven motions of the fin do not haveto be changed.

The designs of the robotic fins weremodified and the fins implementedon a freely swimming biorobotic fish(Figure 1). Modifications includedplacing actuators within the fish bodyadjacent to each fin, using a network ofmicrocontrollers to drive fin motions,

and minimizing the number of actu-ated DOF for each fin ray. The mo-tions and orientation of the roboticfish are controlled by adjusting thepropulsive forces created by five ray-finned fins. In this first implemen-tation, the forces are modulated byswitching between several fin gaitsand by making predetermined changesto fin beat frequencies and to the phaserelationships between fin rays.

Sensory-Based Control of FinsWhat is clearly missing in this ro-

botic system is the ability to automat-ically modulate the kinematics andmechanical properties of the finsbased on sensory information aboutthe fins and their interaction with thewater. The motions of the fins are ad-justed based on the forces required tocontrol the robot’s body, but sensoryinformation is not being used to ex-ploit the phenomena that are criticalto the efficient production of force(e.g., vorticity) nor to adjust behaviorsin response to changes in the flow (e.g.,speed and turbulence). This is dueto the fact that very little is knownabout the sensory-based control ofray-finned fins (Phelan et al., 2010).The fine level of control that the sun-fish has over fin motions and mechan-ical properties suggests strongly thatthere is closed-loop control of the fins.However, fundamental questionsabout the existence of sensory systemsintrinsic to fins, about the types ofstimuli that elicit responses from fins,about information in the flow thatis relevant to propulsive forces, andabout the behavior of the fins in re-sponse to external perturbations havenot yet been answered. This knowl-edge is vital for the development offin-based propulsors that take advan-tage of the phenomena used by fishto produce forces efficiently and that

automatically adjust their behavior inresponse to disturbances and changesin operating requirements.

Experimental Methodsand EquipmentExperimentation

Experiments were conducted toevaluate the response of the sunfish’spectoral fins to external perturbationsapplied to the fin and to the fish’s bodyduring steady swimming. Perturba-tions were created using a vortex gen-erator (Figure 3), which produces avortex ring that moves through thewater and imparts a short durationimpulse to the fish (Figure 4). Thestrength of the vortex was sufficientto deform the pectoral fin or to dis-place the fish laterally by several mil-limeters. The vortex is not visible, soit does not elicit a visually mediated re-sponse from the fish. The vortex does,however, produce a pressure wave thatmay be sensed by the fish.

Two bluegill sunfish, with bodylengths of 160 ± 10 mm and intactpectoral fins, were used for the experi-ments. For the experimental trials,

FIGURE 2

Thrust (horizontal) and lift (vertical) forces forpectoral fins executing normal and modifiedsteady swimming gaits. By making relativelysmall changes to fin stiffness (A) and fin mo-tions (B), the forces can be moved throughoutthe thrust-lift plane. Normal, full-fin steadyswimming for three levels of stiffness (A). Thegait wasmodified slightly by altering the phaseangle between fin rays by 30° (B, red) or byusing just the upper or lower half of the fin(B, green). The magnitude of the force canalso be altered simply by changing the fre-quency of the fin beat.

FIGURE 3

The vortex ring generator and vortex (left).Blue dye was added to the vortex generator’scavity to make the vortex visible to the nakedeye. The vortex generator comprises an orificeplate (1), two cavity plates (2), a latex mem-brane (3), and a connector plate (4), whichenables the air-line to be connected to the vor-tex generator.


a sunfish was placed in the workingarea (280 × 280 × 800 mm) of a600-l flow tank and was allowed to ac-climate for 2 h. The flow rate was set to100 mm s−1, which equates to a steadyswimming speed of approximately0.6 body lengths s−1. At this speed,sunfish generate swimming forcesusing primarily their pectoral fins. Thetail and the caudal, anal, dorsal, andpaired pelvic fins are moved very littlebut are important for stability. Thevortex generator was positioned ap-proximately 150 mm above the tankfloor and placed either perpendicularto the fish in order to perturb the fish’sbody or at a 45° angle to the fish inorder to perturb the pectoral fin duringits outstroke. A horizontal light sheet(Figure 4) used for particle imagevelocimetry (PIV) was positioned sothat it had the same height as the mid-dle of the vortex generator. The fishwas directed into the middle of thetest area and light sheet by coaxing itwith a wooden dowel. Once the fishwas positioned properly, the vortexwas launched to strike the fish. Vorti-ces impacted the fish (1) on the bodynear the tip of the left pectoral finwhile the fin rested against the bodyduring the pause between fin beatsand (2) at the tip of the left pectoralfin as the fin completed its outstroke.

High-speed (500 fps) , high-definition video (1024 × 1024 pixels)

was used to capture the motions ofthe fish and of the fish’s fins. Twocameras (Photron 1024 PCI, PhotronUSA, Inc., San Diego, CA) were syn-chronized and positioned so that theventral and posterior views of the fishwere captured.

AnalysisThe linear and rotational velocities

of the vortices were analyzed usingDaVis (LaVision GmbH, Göttingen,Germany).

The motions of the fish and of thefins were analyzed for two fin beats be-fore and two beats after the impact ofthe vortex. The coordinates of eightpoints along the fish body and pecto-ral were digitized using Matlab (TheMathworks Inc., Natick, MA) andtracked through time. Deformationsand curvatures were calculated forthe pectoral fin during the impact ofthe vortex ring. Three points along thefin were selected to characterize theshape of the fin and to define the ra-dius of curvature.

Design of the VortexRing Generator

Vortex rings are commonly gener-ated using a piston that moves withina cylindrical cavity and pushes a vol-

ume of fluid (the slug) out of the cav-ity and past an orifice with sharpedges. The movement of the pistoncauses the boundary layer that devel-ops in the cavity to separate at the or-ifice’s edge and to roll up into a vortexring that has a toroidal shape. Thespeed of the piston, the diameter ofthe orifice pate, and the ratio of cavitylength to cavity diameter influence theformation of the vortex and the speedat which the vortex travels. Excellentdiscussions of vortex generation arepresented in Gharib et al. (1998),Allen and Auvity (2002), Shusseret al. (2002), and Mohseni (2006).

The vortex generator that was de-veloped for our experiments is simi-lar to a piston based vortex generator,but the design was modified so that itwould be more appropriate for thetesting of swimming fish. Two require-ments that influenced the design were(1) the vortex generator had to be si-lent, so that the fish did not hear amechanism and anticipate the arrivalof the vortex, and (2) the system hadto be small, so that it could be placedat the side of the flow tank without in-terfering with the swimming fish. Thevortex generator consists of two acrylicplates (45 × 55 × 12 mm) in whicha cylindrical cavity is cut (Figure 3).The plates are covered by a 0.3-mmthick aluminum plate with eithera 4.0- or 7.5-mm diameter orifice. Alatex membrane is sandwiched be-tween the cavity plates and anotheracrylic plate in which a cylindricalwell is cut. This plate is connected viaa 6-mm diameter air line (PolyurethaneTubing, NewWay Air Bearings, Aston,PA) to a 60-ml syringe (BectonDickinson and Company, FranklinLakes, NJ). A fast push on the syringeplunger causes the latex membrane toexpand into the cavity and to exhaustthe fluid and create the vortex. The

FIGURE 4

Sunfish in flow tank with vortex generator (A). The sunfish kindly positioned itself in the centerof the test area and laser sheet (B). The laser sheet is used with PIV to characterize the vortex asit travels toward the fish.


effective length of the cavity can be in-creased by drawing the syringe plungerback. This draws the latex membraneback into the cylindrical well. Dyewas introduced into the chamber viaa 1.6-mm diameter hole drilled intothe acrylic plate, radial to the cavity. Asteel tube was inserted into the hole,and was connected via medical tubing(Scientific Commodities, Inc., LakeHavasu City, AZ) to a syringe filledwith food-grade dye. The vortex gen-erator was mounted to an aluminumarm (80/20 Inc., Columbia City, IN)so that it could be positioned withinthe flow tank.

The force, impulse, and linear ve-locity of 12 vortices were characterizedto better understand the propertiesof the vortex and how best to actuatethe plunger. The force generated bythe impact of the vortex was measured(Figure 5b) by shooting the vortexagainst a plate that was connected toa 2.5 g force transducer (LSB200, JRS-Beam Load Cell, Irvine, CA). Theplate was located 100 mm from the or-ifice of the vortex generator. The vor-tex was imaged using the high-speedcamera as it travelled within the 2-mmthick light sheet. Mean values for vor-tices created using a 5-mm diametercavity were: 13 mN force (0.6 mN SE),0.13 mNs impulse (0.002 mNs SE),and 0.99 m/s velocity (0.01 m/s SE).A 13-mm diameter cavity produceda more powerful but slower vortex:67 mN force (2.3 mN SE), 1.0 mNsimpulse (0.02 mNs SE), and 0.85 m/svelocity (0.01 m/s SE). These valuescompare well with estimates we havemade for the peak force and impulsecreated by a sunfish pectoral fins. Ata swimming speed of 0.5 body lengthper second, average fin forces are lessthan approximately 10 mN and theimpulse over the fin beat is less than2.5 mNs.

Response toVortex Perturbations

Perturbation experiments thatinvolved hitting the swimming fishwith a vortex ring showed that thefish did not alter the pectoral fin beatduring the time course of a single finstroke but did change the amplitudeand timing of the pectoral fin beatssubsequent to a vortex impact thatperturbed the fish’s position.

Response to Vortex PerturbationsApplied to the Body

Vortex perturbations that impactedthe side of the fish displaced the fish

laterally by several millimeters (Fig-ure 6), which is significant relative tothe thickness of the fish’s body (maxi-mum of approximately 25 mm). Thelateral displacement occurred whetherthe fish had been drifting toward oraway from the vortex generator priorto the disturbance and was not ac-companied by any obvious change tothe roll or yaw of the fish. An activeresponse of the fish to the vortex per-turbation was evident in the fishes’motion after a short delay. The soon-est the active response occurred was0.05 s, while the longest delay beforea response was evident was 0.20 s. Inthe majority of trials, the fishes actively

FIGURE 5

Evaluation of vortex ring’s velocity using PIV (A). Force from vortex ring during impact with rigidplate attached to force transducer (B).

FIGURE 6

Distance from orifice plate of the left pectoral fin (purple), the right pectoral fin (blue), and thefish at a point between the pelvic fins (green). The distance between the fish and the orifice plateis amplified relative to the fins and is measured at the scale on the right. In this trial, the fish wasmoving towards the vortex generator and was hit by the vortex at about t = 1.49 (red). The activeresponse of the fish occurred by t = 1.50 (gray). (Color versions of figures available online at:http://www.ingentaconnect.com/content/mts/mtsj/2011/00000045/00000004.)


moved away from the vortex generatorafter being hit by the vortex (Figure 7).The movement was not particularlyquick, but was always faster than thefish’s lateral velocity before the perturba-tion had occurred. In some cases (e.g.,Figure 6), the fish actively moved to-ward the vortex generator after beingpushed away from the vortex generatorby the impulse. This occurred onlywhen the fish had been drifting towardsthe vortex generator before the perturba-tion. In some trials, the fish was startledby the vortex and swam out of the testarea. The startled motions were notanalyzed quantitatively.

The motions of the pectoral finsduring the fin beat subsequent to theperturbation were significantly differ-ent from the motions of the pectoralfins prior to the perturbation. How-ever, the pectoral fins did not seem toreact quickly to the stimulus. In fact,the initial movement of the fish’s bodyin response to the vortex generallyoccurred between pectoral fin beats,while the pectoral fins were againstthe fish body (Figures 6 and 7).Thus, the active motion of the fishwas initiated by other fins, which re-acted within as little as 0.05 s. Active

movement of the pectoral fins didnot usually resume until 0.10-0.20 safter the vortex. The frequency of thepectoral fin beats did not changeconsistently after the perturbation. Inthree of the eight trials, the frequencyof the pectoral fin beat increased from,on average, 1.37 Hz (SD = 0.15) to1.83 Hz (SD = 0.28). In the otherfive trials, the frequency of the fin beatdecreased from, on average, 1.69 Hz(SD = 0.26) to 1.34 Hz (SD = 0.21).The amplitude of the pectoral fin mo-tions also changed. This altered theforce balance between the two pectoralfins and contributed to the movementof the fish body. In the beat after thevortex stimulus, the amplitude of theright pectoral fin (opposite the side ofthe impact) was consistently smallerthan before the vortex. Its motion de-creased in all eight trials, on average by18.9% (SD = 10.9%). The amplitudeof the left pectoral fin also changed,but the changes were less consistent.In four trials, the amplitude decreasedby, on average, 37.4% (SD = 22.8),while in the other four trials, theamplitude increased by, on average,8.9% (SD 7.0%). By the second finbeat after the perturbation, the mo-

tions of the left and right pectoralfins were much more similar to themotions before the fin beat, and weresimilar to each other.

Response to Vortex PerturbationsApplied to the Fin

The pectoral fins were deformedsignificantly when struck by the vortexduring the fin beat (Figures 8 and 9).The vortex made contact with the leftpectoral fin near the end of the fin’soutstroke. The vortex bent the tipsof the fin rays and progressively bentlarger portions of the fin as the vortextravelled towards the fish body. The finseemed to bend and fold as if it weremade from thin paper and exhib-ited deformations from the tip to thebase. The maximum measured cur-vature of the fin (along fin ray 6) in-creased from 0.054 mm−1 near the tipand 0.024 mm−1 near the base duringunperturbed swimming to 0.113 mm−1

near the tip and 0.029 mm−1 nearthe base when in contact with the vor-tex. The vortex remained in contactwith the fin while it travelled towardsthe fish body. This resulted in thepectoral fin being pushed back to thebody faster than during an unperturbedinstroke. Times ranged from one thirdto one half of the duration of a normalinstroke and were dependent on many

FIGURE 7

Distance from orifice plate of the left pectoral fin (purple), the right pectoral fin (blue), and thefish at a point between the pelvic fins (green). The distance between the fish and the orifice plateis amplified relative to the fins and is measured at the scale on the right. In this trial, the fish wasmoving away from the vortex generator and was hit by the vortex at about t = 1.25 (red). Theactive response of the fish occurred by t = 1.35 (gray). The fish continued to move away fromthe vortex generator until approximate t = 1.8 s.

FIGURE 8

Pectoral fin perturbed by vortex during swim-ming. The mean fin ray curvature after impactwas 14.4 mm−1 (0.05 mm SE). Reflectiveparticles are used so that the fluid movementis visible.


variables, including the speed of thevortex, how well contact was madewith the fin, and the time of impactwithin the fin beat.

Despite the severity with which thevortex changed the shape and trajec-tory of the perturbed fin, the fish didnot appear to react to the perturbationor to change its behavior subsequent tothe perturbation. During the pertur-bation, the observed motions of theunperturbed fin and of the fish bodywere not visibly different frommotionsprior to the perturbation. Subsequentto the perturbation, the perturbed finremained against the fish body untilthe unperturbed fin completed itsinstroke. Both fins then resumedwhat appeared to be a normal finbeat. Small differences in the pectoralfin beat and the use of other fins likelyoccurred to accommodate for differ-ences in propulsive forces producedduring the perturbation, but thesechanges were not visible. Nor werethere changes in the motion of thefish body, which was not observed tomove laterally or to rotate in yaw.

DiscussionThe objective of the experiments

was to determine how sunfish respond

to perturbations applied to the bodyand fins during steady swimming.These experiments provided a contex-tual understanding of sensory basedmodulation of pectoral fin function.The experiments produced a mix ofexpected and surprising results.

As expected, the fishes did alter theamplitude and timing of the pectoralfin beats subsequent to a perturbationthat disturbed the lateral position ofthe fish body. However, the pectoralfins did not respond quickly to thedisturbance, but remained against thefish body for durations that were onlyslightly different from the pausesbetween fin beats prior to the distur-bance. Active movement of the fish’sbody after the disturbance occurredwith a latency of as little as 0.05 s,which is similar to the 0.08 s latencymeasured byWebb (2004) in responseto roll disturbances. The movement ofthe fish body is believed to have beencaused by fins other than the pectoralfins, since the pectoral fins remainedagainst the body for 0.10–0.20 s afterthe perturbation. When the pectoralfins were moved, the amplitudes ofthe fins seemed to have been adjustedto help equilibrate the movement ofthe fish. By the second fin beat afterthe disturbance, the motions of the

two pectoral fins were synchronizedand had amplitudes similar to thosebefore the disturbance.

The delay in the response of thepectoral fins to the vortex and lateraldisturbance is different from the re-sponse of the fins during experimentswhere an obstacle was placed in frontof the swimming fish (Gottlieb et al.,2010). In those experiments, sun-fish altered the motions of the leftand right pectoral fins during the out-stroke of a steady swimming beat.The changes were not subtle, and thefish did not seem to wait for the nextcycle as in the present studies. The pec-toral fin on the side of the obstaclestiffened and the fin rays were movedthrough trajectories that were very dif-ferent from steady swimming. The finon the side opposite to the obstaclenearly stopped and served to stabilizethe motion of the fish. The differencein the pectoral fins’ response to theobstacle and to the vortex and lateraldisplacement may be related to thefish’s perception of the stimuli. Theobstacle may have beenmore threaten-ing than the vortex, which the fish mayhave interpreted as a common fluidicevent. The fish therefore disruptedthe steady swimming gait in order toproduce large lateral forces that turnedthe fish away from an unknown ob-stacle that may have posed a threat.In contrast, the disturbance inmotionscaused by a fluidic stimulus could beaccommodated simply by adjustingmotions of the fins within their nor-mal gaits. This would allow the centralpattern generator that drives the mo-tions of pectoral fins (Westneat et al.,2004) to continue to produce similaroutput characteristic rather than hav-ing to switch between gaits.

Most surprising was the lack of re-action to the vortex when the vortexdeformed the pectoral fin at the end

FIGURE 9

Ventral view of the fish as the left pectoral fin is hit by a vortex (no dye). The left pectoral finis hit by the vortex (1). The fin is deformed (2, 3, and 4) and is pushed to the body by the vortex.(5 and 6) The right pectoral fin continues to beat normally. The body is not deflected by the vortex.


of the fin’s outstroke and throughoutthe instroke. Nerves and free nerveendings exist throughout the fin raysand the fin webbing (experimen-tal findings, M. Hale, University ofChicago), and so it was expected thatat least one of the phenomena thatthe vortex created—pressure, impact,bending—would have elicited a sen-sory mediated response. The vortexwas in contact with the fin for over100 ms, and so the duration of thestimulus was certainly sufficient for asensory-mediated response to occur.It was also surprising that neither themotions of the body, nor subsequentbeats of the pectoral fins, were clearlydifferent from those before the vortexperturbation. It is highly likely that theleft pectoral fin, while being deformed,produced forces that were differentfrom normal. During a normal steadyswimming gait, each pectoral fin willproduce lateral forces that are similarin magnitude to thrust and lift. Sincethe fins typically beat synchronously,the lateral forces from the left andright fins balance and cancel. Thiswould not have been the case whenthe left pectoral fin was deformed,and the unbalanced forces shouldhave accelerated the fish body laterallyand/or in roll and yaw. The lack of ob-vious lateral motion and adjustment tothe pectoral fin beat may be due simplyto the fish being insensitive to lateralforces. To move the fish laterally,forces must accelerate the mass of thefish and also overcome drag forcesand the load from the mass of wateragainst which the side of the fishpushes. Thus, the loss of lateral forceduring a single fin beat can be easilytolerated because it is difficult for thefish to move sideways. So althoughstudies of biorobotic models of thepectoral fins have shown that thefin’s kinematics and mechanical prop-

erties must be controlled very care-fully to produce forces like the fish(Tangorra et al., 2007, 2010), themechanics of the fish body do notnecessarily require the careful controlof forces at all times in all directions.

ConclusionsPerturbation experiments which

involved hitting the swimming sunfishwith a vortex ring showed that thefish did not alter the pectoral fin beatduring the time course of a single finstroke but did change the amplitudeand timing of its motions in beats sub-sequent to an impact that disturbedthe fish’s position. Vortices that struckthe pectoral fin during the fin’s out-stroke deformed the fin extensively,but the perturbations did not causethe stroke of the unaffected pectoralfin to change, nor did the perturbationcause changes in the motions of thefish body or in the subsequent strokesof either pectoral fin.

These outcomes suggest that thekinematics of the pectoral fins is mod-ulated by sensory information onlywhen a perturbation results in a distur-bance to the fish body, which is the sys-tem that the fins are working to control.The pectoral fins did not react quicklywhen the vortex displaced the fish’sbody but modulated their motions tohelp stabilize the displaced fish afterother fins had already been engaged.The pectoral fins also did not react re-flexively to vortex perturbations thatdeformed the fins’ webbing and finrays. The fin did not appear to moveaway from the vortex or to resist the de-formation by stiffening. The compliantfin allowed itself to bend and perhaps toshed the load from the vortex, and thenaltered its motions during the course ofthe subsequent fin beat.

The results illustrate a benefit ofcompliant mechanisms within a highly

controllable system. The fins of ray-finned fish have the ability to controlforces precisely by altering the kine-matics and mechanical properties ofindividual fin rays. Small changes ineither the trajectories or stiffness offin rays can significantly alter theforce that is transferred to the fish(Tangorra et al., 2010). However, itis not always necessary to regulate thefins precisely. By maintaining its flex-ibility and allowing itself to be de-formed, the fin was able to be hit bythe vortex—which had sufficientforces to displace the fish—withouttransferring the full impact of the per-turbation to the fish body. Therefore,the fin’s passive mechanics made it un-necessary for the fins to be modulatedin order to restore the fish to equilibrium.However, when the fish does want tomove the quickly—as in a maneuveraway from the obstacle—the pectoralfin can be stiffened, the gait changed,and large lateral forces from the fincan be transferred to the fish body.

AcknowledgmentsThis work was supported by theOf-

fice of Naval Research grant N00014-0910352 on fin neuromechanics mon-itored by Dr. Thomas McKenna andby the National Science FoundationEFRI 0938043. We are very gratefulto the members of the Lauder and theTangorra laboratories for many helpfuldiscussions about fish fins and roboticfins and for assistance in executingexperiments and analyzing results.

Lead Author:James Tangorra3141 Chestnut St.,Randell 115Drexel University,Philadelphia, PA 19104Email: [email protected]


ReferencesAkhtar, I., Mittal, R., Lauder, G., & Drucker,

E. 2007. Hydrodynamics of a biologically

inspired tandem flapping foil configuration.

Theoret Comput Fluid Dynamics. 21(3):

155-70. doi: 10.1007/s00162-007-0045-2.

Allen, J., & Auvity, B. 2002. Interaction of a

vortex ring with a piston vortex. J Fluid Mech.

465(1):353-78.

Dong, H., Bozkurttas, M., Mittal, R.,

Madden, P., & Lauder, G. 2010. Computa-

tional modelling and analysis of the hydro-

dynamics of a highly deformable fish pectoral

fin. J Fluid Mech. 645:345-73. doi: 10.1017/

S0022112009992941.

Fish, F.E., Nusbaum, M.K., Beneski, J., &

Ketten, D. 2006. Passive cambering and

flexible propulsors: cetacean flukes. Bioinspir

Biomim. 1(4):S42-48. doi: 10.1088/

1748-3182/1/4/S06.

Gharib, M., Rambod, E., & Sharriff, K.

1998. A universal time scale for vortex ring

formation. J Fluid Mech. 360(1):121-40.

doi: 10.1017/S0022112097008410.

Gottlieb, J., Tangorra, J., Esposito, C., &

Lauder, G. 2010. A biologically derived

pectoral fin for yaw turn manoeuvres.

Appl Bionics Biomech. 7(1):41-55.

doi: 10.1080/11762320903537782.

Lauder, G.V., & Madden, P.G. 2007. Fish

locomotion: kinematics and hydrodynamics

of flexible foil-like fins. Exp Fluids. 43(5):

641-53. doi: 10.1007/s00348-007-0357-4.

Lauder, G.V., Madden, P.G., Mittal, R.,

Dong, H., & Bozkurttas, M. 2006. Locomo-

tion with flexible propulsors: I. Experimental

analysis of pectoral fin swimming in sunfish.

Bioinspir Biomim. 1:S25. doi: 10.1088/

1748-3182/1/4/S04.

Liu, P., & Bose, N. 1997. Propulsive per-

formance from oscillating propulsors with

spanwise flexibility. P Roy Soc Lond A Mat.

453(1963):1763.

McHenry, M. 1995. Mechanical control of

swimming speed: stiffness and axial wave

form in undulating fish models. J Exp Biol.

198(11):2293-305.

Mittal, R., Dong, H., Bozkurttas, M., Lauder,

G., & Madden, P.G. 2006. Locomotion

with flexible propulsors: II. Computational

modeling of pectoral fin swimming in

sunfish. Bioinspir Biomim. 1(4):S35-41.

doi: 10.1088/1748-3182/1/4/S05.

Mohseni, K. 2006. Pulsatile vortex generators

for low-speed maneuvering of small under-

water vehicles. Ocean Eng. 33(16):2209-23.

doi: 10.1016/j.oceaneng.2005.10.022.

Nicholson, J. W., & Healey, A.J. 2008. The

present state of autonomous underwater

vehicle (AUV) applications and technologies.

Mar Technol Soc. 42(1):8.

Office of the Secretary of Defense,USA. 2009.

Unmanned Systems Roadmap (2009-2034).

Phelan, C.T., Tangorra, J.L., Lauder, G.V.,

& Hale, M.E. 2010. A biorobotic model of

the sunfish pectoral fin for investigations of fin

sensorimotor control. Bioinspir Biomim. 5.

doi: 10.1088/1748-3182/5/3/035003.

Prempraneerach, P., Hover, F.S., &

Triantafyllou, M.S. 2003. The effect of

chordwise flexibility on the thrust and

efficiency of a flapping foil. Autonomous

Undersea Systems Institute. In: 13th Int.

Symp. Unmanned Untethered Submersible

Techn. Durham, NH.

Shoele, K., & Zhu, Q. 2009. Fluid–structure

interactions of skeleton-reinforced fins: per-

formance analysis of a paired fin in lift-based

propulsion. J Exp Biol. 212(16):2679.

doi: 10.1242/jeb.030023.

Shusser, M., Gharib, M., Rosenfeld, M., &

Mohseni, K. 2002. On the effect of pipe

boundary layer growth on the formation of

a laminar vortex ring generated by a piston/

cylinder arrangement. Theoret Comput Fluid

Dynamics. 15(5):303-16. doi: 10.1007/

s001620100051.

Tangorra, J., Lauder, G.V., Hunter, I.W.,

Mittal, R., Madden, P.G., & Bozkurttas, M.

2010. The effect of fin ray flexural rigidity on

the propulsive forces generated by a biorbotic

fish pectoral fin. J Exp Biol. 213:4043-54.

doi: 10.1242/jeb.048017.

Tangorra, J., Phelan, C.T., Esposito, C.J., &

Lauder, G.V. 2011. Use of biorobotic models

of highly deformable fins for studying the

mechanics and control of fin forces in fishes.

Integr Comp Biol. in press. doi: 10.1093/

icb/icr036.

Tangorra, J.L., Davidson, S.N., Hunter, I.W.,

Madden, P.G., Lauder, G.V., Dong, H., …

Mittal, R. 2007. The development of a bio-

logically inspired propulsor for unmanned

underwater vehicles. IEEE J Oceanic Eng.

32(3):533-50. doi: 10.1109/JOE.2007.903362.

Triantafyllou, M., Techet, A., Zhu, Q.,

Beal, D., Hover, F., & Yue, D. 2002.

Vorticity control in fish-like propulsion and

maneuvering. Integr Comp Biol. 42(5):1026.

doi: 10.1093/icb/42.5.1026.

Triantafyllou, M.S., Hover, F.S., Techet, A.,

& Yue, D. 2005. Review of hydrodynamic

scaling laws in aquatic locomotion and fishlike

swimming. Appl Mech Rev. 58(4):226-37.

doi: 10.1115/1.1943433.

Webb, P. 2004. Response latencies to postural

disturbances in three species of teleostean

fishes. J Exp Biol. 207(6):955-61. doi: 10.1242/

jeb.00854.

Westneat, M., Thorsen, D., Walker, J., &

Hale, M. 2004. Structure, function, and

neural control of pectoral fins in fishes. IEEE J

Oceanic Eng. 29(3):674-83. doi: 10.1109/

JOE.2004.833207.

Zhu, Q., & Shoele, K. 2008. Propulsion

performance of a skeleton-strengthened fin.

J Exp Biol. 211(13):2087. doi: 10.1242/

jeb.016279.

Zhu, Q., Wolfgang, M.J., Yue, D.K.P., &

Triantafyllou, M.S. 2002. Three-dimensional

flow structures and vorticity control in fish-

like swimming. J Fluid Mech. 468:1-28.

doi: 10.1017/S002211200200143X.


P A P E R

Bioinspired Design Process for an UnderwaterFlying and Hovering VehicleA U T H O R SJason D. GederJohn S. PalmisanoRavi RamamurtiMarius PruessnerBanahalli RatnaNaval Research Laboratory

William C. SandbergScience ApplicationsInternational Corporation

A B S T R A C TWe review here the results obtained during the past several years in a series of

computational and experimental investigations aimed at understanding the origin ofhigh-force production in the flapping wings of insects and the flapping and deform-ing fins of fish and the incorporation of that information into bioinspired vehicledesigns. We summarize the results obtained on pectoral fin force production, flap-ping and deforming fin design, and the emulation of fish pectoral fin swimming inunmanned vehicles. In particular, we discuss the main results from the computa-tional investigations of pectoral fin force production for a particular coral reef fish,the bird wrasse (Gomphosus varius), whose impressive underwater flight and hov-ering performance matches our vehicle mission requirements. We describe thetradeoffs made between performance and produceability during the bio-inspireddesign of an actively controlled curvature pectoral fin and the incorporation of itinto two underwater flight vehicles: a two-fin swimming version and four-fin swim-ming version. We describe the unique computational approach taken throughoutthe fin and vehicle design process for relating fin deformation time-histories tospecified desired vehicle dynamic behaviors. We describe the development of thevehicle controller, including hardware implementation, using actuation of the mul-tiple deforming flapping fins as the only means of propulsion and control. Finally,we review the comparisons made to date between four-fin vehicle experimentaltrajectory measurements and controller simulation predictions and discuss theincorporation of those comparisons into the controller design.Keywords: bio-inspired robotics, pectoral fin, unmanned systems, computationalfluid dynamics

Background

Biologists and zoologists havebeen studying fish swimming formany decades, and several compre-hensive texts and papers exist (Breder,1926; Lindsey, 1978; Alexander,1983; Azuma, 1992; Blake, 1983;Webb, 1975, 1984; Videler, 1993;Sfakiotakis et al., 1999). The experi-mental studies of fish swimming bio-dynamics have become increasinglyquantitative as measurement technol-ogy has improved. The use of high-speed photography shed light uponthe details of fin deformation (Gibbet al., 1994; Walker & Westneat,1997). Laser light scattering tech-niques enabled observations of notonly the fish body and fin dynam-ics but also the velocity field aboutthe fish and in the wake (Drucker &Lauder, 1999, 2002; Gharib et al.,2002; Bartol et al., 2003). Vehicle de-signers are able to draw upon such richdata sets as they embark upon designs.But where does one begin?

A bioinspired vehicle design shouldbegin, according to Webb (2004), byspecifying the performance goals of

the desired mission first and then ex-amining those living creatures whoseperformance is relevant. These crea-tures have evolved to meet all theirneeds, and the maneuvering enabledby these needs may intersect withthe performance requirements drivinga vehicle design. However, since theliving creatures selected for study aremost likely not optimized for the mo-bility characteristics that are drivingthe design, one should not copy na-ture but instead be guided by it.This point of caution to designershas been made many times by biolo-gists (Combes & Daniel, 2001;

Wainwright et al., 2002; Collar et al.,2008).

The mission selected for the NavalResearch Laboratory (NRL) swim-ming vehicle, which we describebelow, requires precise low-speed ma-neuvering and excellent hovering in acomplex near-shore environment inaddition to excellent position-keepingin tidal currents. Cost and mechanicalsimplicity constraints demanded arigid hull. This eliminated undulatoryswimming fish as a primary means ofdesign inspiration, and instead wewere lead to consider fin-based swim-ming creatures. Kato and Furushima


(1996) had already proceeded downthe path of paired fin swimming, draw-ing inspiration from the black bass todesign rigid pectoral fins and incor-porate them into a test-bed vehicle.He subsequently pursued low-speedmaneuvering using rigid pectoral fins(Kato, 2000) and then went on to de-velop a passive flexible fin and an activepneumatic actuator pectoral fin (Katoet al., 2008). Barrett and Triantafyllo(1995) on the other hand had takentheir inspiration from the undulatingbody and oscillating caudal fin of thetuna to design and build their flexible“Robotuna.” We looked for a fishwhich possessed the dynamic perfor-mance characteristics we needed andfor which experimental measurementsof swimming dynamics, including finkinematics, already existed. Experi-mental observations of pectoral finmuscle activity, kinematics, and dy-namics in coral reef wrasses (Westneat&Walker, 1997;Walker &Westneat,1997) have shown that the body isessentially held rigid during straight-line motion, thus satisfying our rigidhull constraint. Very rapid (∼10 bodylengths per second) translationalmotions were observed (Walker &Westneat, 2000) to give comparableswimming performance to that seenin body-caudal fin swimmers of com-parable size even though there was nocontribution from body undulationor caudal fin oscillation in the wrasses.This solution offered the potential offast forward swimming (or position-keeping in strong currents) whileavoiding the complexity of a flexi-ble hull. In addition, Walker andWestneat (1997) had observed highmaneuverability by these swimmersin their complex reef habitats. Thesehabitats are reasonable representationsof our vehicle’s projected operatingenvironment, hence we saw the possi-

bility of obtaining both high forwardspeed and excellent low-speed maneu-vering and hovering performance aswell.

The fish we selected to inspire ourdesign was one of the coral reef pecto-ral fin swimmers, the bird wrasse(Gomphosus varius), shown in Figure 1.

Controlled DeformationFin TechnologyDevelopment

Living creatures, such as insects,birds, and pectoral fin swimmers,generate lift and thrust by executinglarge-amplitude wing/fin flapping,often with substantial shape deforma-tion from root to tip and leading edgeto trailing edge. The flow for thesemotions is three-dimensional and un-steady, and conventional steady-stateaerodynamics is unable to correctlycompute the corresponding time his-tory of flapping-force generation.Comprehensive reports on the re-search carried out to study the fluid dy-namics of flapping fins and wings(Rozhdestvensky & Ryzhov, 2003)and of biomimetic fins for underwatervehicles (Triantafyllo et al., 2004)exist, in addition to an extensive num-ber of studies, too great to list here, onunsteady lift production by flappinginsect wings. Three-dimensional un-steady computations are necessary tocorrectly predict the lift and thrust var-

iation throughout the flapping strokecycle. Such computations, for crea-tures or vehicles with moving anddeforming surfaces, provide the time-varying pressure distribution on allsurfaces, which in turn can provide in-sights into how the flapping forces andmaneuvering moments are being gen-erated. This information can be cou-pled with computational visualizationof the time-varying flow about thefish to analyze the origin of body andfin vorticity, its growth, and eventualshedding into the wake. This is theapproach we developed for tuna cau-dal fin force production analysis(Ramamurti et al., 1996, 1999), sub-sequently validated against wrasse ex-perimental data (Ramamurti et al.,2002), and which we also followedthroughout our fin and vehicle devel-opment efforts described below.

There are 13 multiply bifurcatingfin rays in the bird wrasse pectoral finshown in Figure 2, each contributingto the fin curvature time-variationthroughout the stroke cycle. For easeof design, manufacture, actuation,and control, it is ideal to have the few-est possible number of rays (which werefer to as ribs) and have each of them

FIGURE 1

Bird wrasse (Gomphosus varius).

FIGURE 2

Bird wrasse fin structure (from Walker &Westneat, 1997).


be as simple in shape and structureas possible. But for more effective finpropulsion, it is ideal to maximize thenumber of ribs since more controlpoints result in a smoother fit todesired fin curvature time-histories.Our computational investigations(Ramamurti et al., 2004) have shownthat the loss of force magnitude overthe stroke cycle is very small if weconsiderably reduce the number ofribs, as long as we maintain the abilityto properly modify the surface curva-ture. The computations showed thatwhen the fin was made rigid by speci-fying the motion with just the leadingedge of the fin tip, the thrust producedduring the upstroke was less than halfof the peak thrust produced by theflexible fin computations. During thedownstroke, the computations forthe rigid and nearly rigid fin pro-duced no positive thrust, while thepartially and fully flexible cases pro-duced substantial thrust. In the caseof the rigid fin, there was also a sub-stantial penalty in lift during the

upstroke. An example from thesecomputational investigations is shownbelow in Figure 3.

Assessment of these findings led usto reduce the number of ribs from 13to 5. Five was selected since it enabledreduced fin complexity, thus substan-tially reducing fin size and weight,while maintaining the critically impor-tant flexural capability. Each of the fivefin rays were individually designedand constructed from compliant ABSplastic material using a 3-D printer toachieve the desired tip deflection withan achievable linear actuation force ap-plied to each individual rib at the root(Trease et al., 2003), as illustrated inFigure 4. The pushing and pulling ofthe ribs at the root of the fin is similarto how fish bend their ribs using mus-cle actuation.

The root section of the fin was se-lected to be of rectangular cross sectionwith rounded leading edges and a ta-pered trailing edge in order to accom-modate the actuators at the base of theribs as shown in Figure 5a. A translu-

cent silicone rubber membrane skin,optimized in thickness to approxi-mately 0.5 mm via finite-element anal-ysis (Palmisano et al., 2007), providedthe continuous surface covering forthe rays as shown in Figure 5b.

After several parametric 3-D un-steady computational fluid dynamics(CFD) studies had been carried out(Ramamurti & Sandberg, 2006), vari-ous fin parameters were chosen thatoptimized thrust performance, giventhe mechanical constraints of the fin.An example of these computationsshowing the variation of lift andthrust as a function of fin flexibility,stroke amplitude, and fin stroke biasis shown in Figure 6.

The angle of attack of the root sec-tion of the finwas chosen to be 20°, theamplitude of the oscillation to be 114°,the flapping frequency to be 1 Hz, andthe rib spacing to be the minimumpossible value of 1.2 cm dictated bythe size of the actuators. These specificvalues, selected for maximizing finforce production, are for a solitaryflapping and deforming fin. Incorpo-ration of the fin into a vehicle designpresents challenges to retain fin per-formance while maintaining vehiclesimplicity.

Two-Fin Test-Bed VehicleA simple vehicle was designed and

built to serve as a test-bed for evaluat-ing the performance of all aspects of

FIGURE 3

Effect of fin flexibility on the time variation of thrust forces (from Ramamurti et al., 2004).(Color versions of figures available online at: http://www.ingentaconnect.com/content/mts/mtsj/2011/00000045/00000004.)

FIGURE 4

Representative rib cross-sectional geometryand bending analysis showing a 20° ribdeflection.


the controlled curvature fin technol-ogy, including its capability for vehiclepropulsion, low-speed maneuvering,and hovering. It was, therefore, inten-tionally, a minimalist design thatserved to house the fins, actuators,and a battery. We described abovehow the parameters that govern theforce production by the fin were cho-sen. However, due to mechanicalconstraints and a desire to easilymanufacture a vehicle prototype, addi-tional 3-D unsteady CFD studies ofthe flapping fins incorporated in the

test-bed vehicle led us to modify thefin-alone values. The angle of attackof the root section of the fin was cho-sen to be 0°, and the rib spacing wasreduced to 0.8 cm. It is during con-struction tradeoffs of this type thatone relies upon what has been learnedfrom the studies of nature to meet op-erational performance goals while bal-ancing that with the desire to create avehicle that is producible at a reason-able cost. Fixing the fin root at a spe-cific angle deviates from nature, butthe construction simplicity and cost

savings are so substantial that someperformance penalty was accepted.The flapping stroke amplitude andfrequency, as well as the phasing ofthe individual fin tip deflection timehistories, were retained as controllableparameters, and we have performedCFD analyses of force productionwith this fin configuration (Ramamurtiet al., 2010).

In keeping with the review natureof this paper, we are emphasizing theprocess carried out for our specificbio-inspired vehicle design. The detailsof the controlled curvature fin design,the linear actuator design and con-struction, the isolated fin constructionand testing, the vehicle design, the ve-hicle construction, the experimentaltesting, and the validation of com-putations were previously reported(Palmisano et al., 2007, 2008; Sandberg& Ramamurti, 2008). The fin tech-nology test-bed demonstration vehicleincorporated two actively controlled de-formation fins. The two-fin vehicle isshown in Figure 7.

Four-Fin Test-Bed VehicleThe test results for the two-fin ve-

hicle demonstrated that the controlleddeformation fin force production wascapable of meeting our vehicle pro-pulsion (position-keeping in a current)requirements (Geder et al., 2008).However, by design, this test-bed fintechnology demonstration vehiclehad restricted options for sensor pay-load and did not have the fore-aftforce production capability neededfor heave-pitch control. Hence, thedesign of a larger 41-cm long four-finvehicle was initiated (Figure 8). Thefore-aft symmetry of the four-fin de-sign enables hover and higher precisionpositioning capabilities by decouplingvehicle pitch and heave control.

FIGURE 5

Mechanical fin (a) CAD image without skin and (b) actual with skin.

FIGURE 6

Mean fin generated (a) thrust and (b) lift as functions of non-dimensionalized stroke amplitudeand fin flex, or curvature. The surfaces indicate a bias in the fin stroke angle of 0° (red), 20°(green), and 40° (blue).


The current four-fin vehicle design shown here employs a water-tight cylinderfor housing the power source and electronics with a flooded space in the nose andtail for buoyancy trimming and supplemental sensors. Hardware control and allcomputations are performed by a 16-MHz ATmega2560 microcontroller. Com-putations (Ramamurti et al., 2010) and experimental tests carried out to date(Geder et al., 2011) characterized how changes in fin stroke amplitude, frequency,bias angle, and curvature affect the thrust and lift forces for zero free stream flowspeed. Further testing and computations are ongoing to fully characterize the finforces and vehicle dynamics. However, current models have the necessary fidelityto accurately predict vehicle performance as outlined in the following sections.

Four-Fin Vehicle ControlWith the vehicle state variables defined as in Figure 8, the vehicle dynamics can

be written as,

where M is a matrix of rigid body mass and inertial terms, C is a matrix of cen-tripetal and Coriolis terms,D is a matrix of hydrodynamic lift and drag terms, g is avector of hydrostatic terms, v = [u v w p q r]T, η = [x y zϕ θψ]T is the position andorientation vector in the earth-fixed frame where ϕ, θ, and ψ are roll, yaw, andpitch angles, and τ is a vector of all forces and moments external to the rigid body.The portion of the vector, τ, that is effected by the fins is represented as,

→τfins ¼

fT ;LF þ fT ;LB þ fT ;RF þ fT ;RB

0fL;LF fL;LB fL;RF fL;RB

yL fL;LF þ fL;LB yR fL;RF þ fL;RB

xF fL;LF þ fL;RF þ xB fL;LB þ fL;RB

yL fT ;LF þ fT ;LB

yR fT ;RF þ fT ;RB

26666664

37777775; ð2Þ

where fT is fin thrust and fL is fin lift.Subscripts ‘LF’, ‘LB’, ‘RF’, and ‘RB’identify the left front, left back, rightfront, and right back fins, respectively.The x-position of the center of pressureon the fins is denoted by xF for thefront fins and xB for the back fins.The y-position of the center of pressureon the fins is denoted by yL for the leftfins and yR for the right fins.

Mathematical models representingthe dynamic performance of the finshave been developed to include the ef-fects on force production of inflow ve-locities to the leading edge (or trailingedge for reverse motion) and to includethe effects of fin interactions with eachother, namely the effects of the trailingvortices off the front fins on the inflowto the back fins (Geder et al., 2011).Other fin dynamic representationsmodeled controllable parameters in-cluding fin curvature and stroke am-plitude (Ramamurti et al., 2010). Inthe earlier 3-D unsteady CFD studies,these two key parameters were foundto have a direct relationship with thrustgeneration—increasing stroke ampli-tude or fin curvature increased thrust(Ramamurti & Sandberg, 2006).However, since both stroke amplitudeand flapping frequency are limited me-chanically in the vehicle, optimal com-binations of amplitude and frequencywere experimentally found for highthrust and lift fin gaits. The best mixof these parameters for our vehiclewas determined to be 100° for strokeamplitude and 1.8 Hz flapping fre-quency, which yielded not only highforce output but also relatively lowpower consumption (Palmisano et al.,2007). These findings allowed us to fixstroke amplitude and frequency asconstants and to focus on fin curvatureas the primary thrust control parame-ter. Further, biasing the fin stroke upor down, as in Figure 9, affects fin lift

FIGURE 7

Two-fin technology test-bed demonstration vehicle: (a) exterior view and (b) interior view.

Mv→þ C v→Þv→ þ D v→ð Þv→ þ g→ η→ð Þ ¼ τ→;ð ð1Þ


generation while maintaining constantthrust (Geder et al., 2008).

In previous work we evaluated thebenefits of two vehicle control meth-ods (Geder et al., 2008). The firstmethod, called weighted gait combina-tion, used combinations of thrust-generating and lift-generating fingaits to produce vectored propulsiveforces. The second method, calledmean bulk angle bias (MBAB), usedweighted forward-reverse gait controlwith stroke bias angle control. Be-tween these two control methods,our results showed that MBAB betterdecoupled control over body-fixedthrust and lift forces and yielded bettervehicle response characteristics in sim-ulation. As such, the MBAB methodis used to control the four-fin vehicle.

The vehicle controller commandschanges to the fins to effect changesin the forces and moments impartedon the vehicle, as shown in equation 2.These fin commands are based on

errors in the vehicle dynamic states,computed as the commanded valuesminus the computed values. Statesare computed onboard the vehicleusing a suite of sensors (three axesof accelerometers, three axes of rategyros, magnetic compass, and pressuresensor) and sensor fusion and filteringschemes (Geder et al., 2009). Errors insurge motion (x-axis translation) dic-tate commands for fin thrust changesto all fins. Errors in heave motion(z-axis translation) dictate commandsfor fin lift changes to all fins. Errorsin roll motion (x-axis rotation) dictatecommands for differential lift changesbetween left and right fins. Errors in

pitch motion (y-axis rotation) dictatecommands for differential lift changesin forward and back fins. Errors in yawmotion (z-axis rotation) dictate com-mands for differential thrust changesin left and right fins. The vehicle hasno direct control over sway motion( y-axis translation), and instead com-mands yaw motion changes to movein this direction. The direction ofyaw motion depends on the swayerror. The output of a proportional-integral-derivative (PID) controllerfor each vehicle state is used to deter-mine forward and reverse fin gait per-centage and bulk angle commands.

Four-Fin VehiclePerformance

Initial measurements of the dy-namic performance of the four-fin ve-hicle have been conducted in two testfacilities, one a 6 × 2.5 × 2 foot watertank and the other a 50-foot diameterby 50-foot deep water tank (Figure 10).The experimental measurements haveserved to validate the vehicle dynamicmodel and to begin the assessment ofvehicle performance.

FIGURE 8

Four-fin technology test-bed vehicle, (a) exterior view, (b) interior view.

FIGURE 9

Vehicle images showing all four fins withstrokes (a) biased down to produce positivelift and (b) biased up to produce negative lift.

FIGURE 10

Four-fin vehicle operating in a Naval Research Laboratory test facility.


An open loop test was conducted tocharacterize vehicle heading angle re-sponse and to compare model simula-tion performance with experimentalperformance (Geder et al., 2011).The right fins were set at full reverse ki-nematics, and the left fins were set toclosely match the opposite thrust ofthe right fins. At t = 11 s, the gait weight-ing inputs were reversed. The simulatedand experimental responses show verygood agreement (Figure 11) with amax-imum difference between the two re-sponses at any given time of 5°. Bothsimulated and experimental results ex-hibit a 30°/s maximum turning rate,4 s time from zero to maximum speed,and braking angle of 35°—the amountof residual turning distance after finkinematics are reversed.

After validating the four-fin vehicledynamicmodel in yawmotion and im-plementing state feedback control,initial closed-loop experiments weredone to test heading angle control(Geder et al., 2011). In Figure 12, acomparison of experimental and simu-

lated results is given. For a simple pro-portional control algorithm (Figure 12a),we see the results match well with a30-40° amplitude and 6.5-s period. Dif-

ferences between measured experi-mental heading from onboard sensorsand simulated heading can be attrib-uted to the noise and sensitivity char-acteristics of the sensors (Geder et al.,2009). External magnetic disturbancesin our test facility cause errors up to10° in heading measurements, also fac-toring into the variation between ex-perimental and simulated headingresponses, as our calibration of the on-board compass was not perfect. Add-ing in a derivative gain to damp theheading response and a small integralgain to eliminate any steady-state er-ror, we see the response to a 180° stepcommand in heading in Figure 12b.With PID control over heading angle,the response is nearly critically dampedwith a rise time of 7 s.We also see closeagreement between measured and ac-tual angles, and again differences inthe responses can be attributed to sen-sor noise and sensitivity, as well as com-pass calibration errors.

FIGURE 11

Comparison of experimental and simulated open-loop heading angle responses (from Geder et al.,2011).

FIGURE 12

Comparison of experimental and simulated closed-loop heading angle responses with (a) propor-tional control and (b) PID control (from Geder et al., 2011). The solid curve represents the actualsimulated heading response of the vehicle based on the modeled dynamics. The dashed curverepresents the actual experimental heading response observed in vehicle testing. The squaredata represent the measured heading response of the vehicle based on the output of sensor mod-els. The circle data represent the measured experimental vehicle heading from the output of on-board sensors.


ConclusionsWe have reviewed the history

of our bioinspired vehicle design pro-cess. The process began by col-laborating with biologists in order tounderstand the dynamics of flappingand deforming fins in pectoral finswimmers. The detailed kinematicsthey measured in fish swimming ex-periments provided the time-varyingfin surface curvature data necessaryfor computing the 3-D unsteady flowabout the swimming fish. Examina-tion of the computed flow variationsabout the flapping and deformingfins and the fish body provided insightsinto the relationship between the finflows and the fin force time-historiesthroughout the stroke cycle. Parametricvariations of key stroke parameters in3-D unsteady flow computationsyielded the sensitivity of the time-varying fin forces, which in turn pro-vided the information needed tomodifyour fin design from that of the birdwrasse. It is during such computa-tions that insights from nature can beblended with design constraints toyield a range of possible bio-inspireddesigns. A controlled curvature fin uti-lizing individually designed fin ribs,each actuated at the rib root by a linearactuator was built and tested. The suc-cessful fin tests were followed by thedesign and construction of a two-fintest-bed vehicle to demonstrate themobility potential of the concept. In-corporation of the fins into a vehiclenecessitated further compromiseswhere we balanced biomimetic perfor-mance with produceability and cost.These test-bed vehicle tests were fol-lowed by the design and constructionof a more capable four-fin vehicle, in-corporating the same fins. A series ofcontrollers were developed and builtto enable assessment of vehicle propul-

sion and maneuvering performance asa function of the varying kinematics ofeach fin. Deforming fin force time his-tories, including fin-fin unsteady inter-actions, have been incorporated intothe vehicle dynamic model used forcontroller development. The prelimi-nary four-fin vehicle dynamic perfor-mance measurements indicate verygood agreement with computed per-formance. These initial results arevery encouraging and indicate thatour efforts to emulate the position-keeping, low-speed maneuvering, andhovering performance of the birdwrasse into a producible and low costvehicle are bearing fruit.

Lead Author:Jason D. GederLaboratory for ComputationalPhysics and Fluid DynamicsNaval Research LaboratoryOverlook Avenue, SW,Washington, DCEmail: [email protected]

ReferencesAlexander, R.M. 1983. The history of fish

biomechanics. In: Fish Biomechanics, eds.

Webb, P.W., & Weihs, D., 1-35.

New York: Praeger.

Azuma, A. 1992. The Biokinetics of Flying

and Swimming. Tokyo: Springer-Verlag.

265 pp.

Barrett, D.S., & Triantafyllo, M.S. 1995.

The design of a flexible hull undersea vehicle

propelled by an oscillating foil. In: Proc.

9th Int. Symp. on Unmanned Untethered

Submersible Technology. Durham, NH:

Autonomous Undersea Systems Institute

(AUSI).

Bartol, I.K., Gharib, M., Weihs, D., Webb,

P.W., Hove, J.R., & Gordon, M.S. 2003.

Hydrodynamic stability of swimming in

ostraciid fishes: role of the carapace in the

smooth trunkfish Lactophrys triqueter

(Teleostei: Ostraciidae). J Exp Biol. 206:

725-44. doi: 10.1242/jeb.00137.

Blake, R.W. 1983. Fish Locomotion.

Cambridge, UK: Cambridge University

Press. 228 pp.

Breder, C.M. 1926. The locomotion of fishes.

Zoologica. 4:159-297.

Collar, D., Wainwright, P., & Alfaro, M.

2008. Integrated diversification of locomotion

and feeding in labrid fishes. Biol Lett. 4:84-6.

doi: 10.1098/rsbl.2007.0509.

Combes, S., & Daniel, T. 2001. Shape,

flapping, and flexion: wing and fin design

for forward flight. J Exp Biol. 204:2073-85.


Locomotor forces on a swimming fish:

Three-dimensional vortex wake dynamics

quantified using digital particle image

velocimetry. J Exp Biol. 203:2393-412.


Experimental hydrodynamics of fish

locomotion: Functional insights from wake

visualization. Integr Comp Biol. 42:243-57.

doi: 10.1093/icb/42.2.243.

Geder, J.D., Palmisano, J., Ramamurti, R.,

Sandberg, W.C., & Ratna, B. 2008. Fuzzy

logic PID-based control design and perfor-

mance for pectoral fin-propelled unmanned

underwater vehicle. In: Proc. Int. Conf on

Control, Automation and Systems. Seoul,

Korea: IEEE. doi: 10.1109/ICCAS.2008.

4694526.

Geder, J.D., Ramamurti, R., Palmisano, J.,

Pruessner, M., Ratna, B., & Sandberg, W.C.

2009. Sensor data fusion and submerged

test results of a pectoral fin propelled UUV.

In: Proc. 16th Intl Symp on Unmanned

Untethered Submersible Technology. Durham,

NH: Autonomous Undersea Systems Institute

(AUSI).

Geder, J.D., Ramamurti, R., Palmisano, J.,

Pruessner, M., Sandberg, W.C., & Ratna, B.

2011. Four-fin bio-inspired UUV: modeling


and control solutions. ASME Intl Mech Eng

Congress Exposition. IMECE. 2011-64005.

Gharib, M., Pereira, F., Dabiri, D., Hove,

J.R., & Modarress, D. 2002. Quantitative

flow visualization: Toward a comprehensive

flow diagnostic tool. Integr Comp Biol.

42:964-70. doi: 10.1093/icb/42.5.964.

Gibb, A.C., Jayne, B.C., & Lauder, G.V.

1994. Kinematics of pectoral fin locomotion

in the bluegill sunfish Lepomis Macrochirus.

J Exp Biol. 189:133-61.

Kato, N. 2000. Control performance in

the horizontal plane of a fish robot with

mechanical pectoral fins. IEEE J Oceanic Eng.

25:125-9. doi: 10.1109/48.820744.

Kato, N., Ando, Y., Tomokazu, A., Suzuki, H.,

Suzumori, K., Kanda, T., & Endo, S. 2008.

Elastic pectoral fin actuators for biomimetic

underwater vehicles. In: Bio-mechanisms

of Swimming and Flying, eds. Kato, N.,

Kamimura, S., 271-82. Tokyo: Springer.

Kato, N., & Furushima, M. 1996. Pectoral

fin model for maneuver of underwater vehi-

cles. In: Proc. Symp. on Autonomous

Underwater Vehicle Technology. Monterrey,

CA: IEEE. doi: 10.1109/AUV.1996.532400.

Lindsey, C.C. 1978. Form, function, and

locomotory habits in fish. In: Fish Physiology

Vol. II, eds. Hoar, W.S., & Randall, D.J.,

1-100. New York: Academic Press.

Palmisano, J., Geder, J.G., Ramamurti, R.,

Liu, K.J., Cohen, J.J., Mengesha, T., …

Ratna, B. 2008. Design, development and

testing of flapping fin with actively-controlled

curvature for an unmanned underwater vehi-

cle. In: Bio-mechanisms of Swimming and

Flying, eds. Kato, N., & Kamimura, S.,

283-94. Tokyo: Springer. doi: 10.1007/

978-4-431-73380-5_23.

Palmisano, J., Ramamurti, R., Liu, K.J.,

Cohen, J., Sandberg, W.C., & Ratna, B.

2007. Design of a bio-mimetic controlled-

curvature robotic pectoral fin. In: Proc.

IEEE Intl. Conf. on Robotics and Automation.

Rome, Italy: IEEE. doi: 10.1109/ROBOT.

2007.363110.

Ramamurti, R., Geder, J.D., Palmisano, J.,

Ratna, B., & Sandberg, W.C. 2010. Com-

putations of flapping flow propulsion for un-

manned underwater vehicle design. AIAA J.

48(1):188-201. doi: 10.2514/1.43389.

Ramamurti, R., Lohner, R., & Sandberg,

W.C. 1996. Computation of the unsteady

flow past a tuna with caudal fin oscillation.

In: Advances in FluidMechanics, eds. Rahman,

M.,&Brebbia, C., vol. 9, 169-78. Southampton,

UK: Computational Mechanics Publishing.

Ramamurti, R., Lohner, R., & Sandberg,

W.C. 1999. Computation of the 3-D

unsteady flow past deforming geometries.

Int J Comput Fluid D. 13:83-99.

doi: 10.1080/10618569908940891.

Ramamurti, R., & Sandberg, W.C. 2006.

Computational fluid dynamics study for

optimization of a fin design. In: Proc. 24th

AIAA Applied Aerodynamics Conference.

AIAA. AIAA-2006-3658.

Ramamurti, R., Sandberg,W.C.,&Lohner, R.

2004. The influence of fin rigidity and

gusts on the force production in fish and

insects: a computational study. In: Proc. 42nd

AIAA Aerospace Sciences Meeting and Exhibit.

Am Inst Aeronautics and Astronautics.

AIAA-2004-0404.

Ramamurti, R., Sandberg, W.C., Lohner, R.,

Walker, J.A., & Westneat, M.W. 2002. Fluid

dynamics of flapping aquatic flight in the

bird wrasse: 3-D unsteady computations with

fin deformation. J Exp Biol. 205:2997-3008.

Rozhdestvensky, K.V., & Ryzhov, V.A.

2003. Aerohydrodynamics of flapping-wing

propulsors. Prog Aerosp Sci. 39:585-633.

doi: 10.1016/S0376-0421(03)00077-0.

Sandberg, W.C., & Ramamurti, R. 2008.

3-D Unsteady computations of flapping flight

in insects, fish, and unmanned vehicles.

In: Bio-mechanisms of Swimming and

Flying, eds. Kato, N., & Kamimura, S.,

205-217. Tokyo: Springer. doi: 10.1007/

978-4-431-73380-5_17.

Sfakiotakis, M., Lane, D.M., &Davies, J.B.C.

1999. Review of fish swimming modes for

aquatic locomotion. J Exp Biol. 24:241-52.

Trease, B.P., Liu, K.J., & Kota, S. 2003.

Biomimetic compliant system for actuator-

driven aquatic propulsor: preliminary results.

In: ASME Intl. Mech. Eng. Congress and

Exposition. IMECE 2003-41446.

Triantafyllo, M.S., Techet, A.H., & Hover,

F.S. 2004. Review of experimental work in

biomimetic foils. IEEE J Oceanic Eng.

29:585-94. doi: 10.1109/JOE.2004.833216.

Videler, J.J. 1993. Fish Swimming. London,

UK: Chapman and Hall. 260 pp.

Wainwright, P., Bellwood, D., & Westneat,

M.W. 2002. Ecomorphology of locomotion

in labrid fishes. Environ Biol Fish. 65:47-62.

doi: 10.1023/A:1019671131001.

Walker, J.A., & Westneat, M.W. 1997.

Labriform propulsion in fishes: Kinematics

of flapping aquatic flight in the bird wrasse

Gomphosus varius (Labridae). J Exp Biol.

200:1549-69.


Mechanical performance of aquatic rowing

and flying. P Roy Soc Lond B Bio. 267:

1875-81. doi: 10.1098/rspb.2000.1224.

Webb, P.W. 1975. Hydrodynamics and

energetics of fish propulsion. Bull Fish Res Bd

Can. 190:1-158.

Webb, P.W. 1984. Form and function in fish

swimming. Sci Amer. 251:58-68.

Webb, P.W. 2004. Maneuverability—

General issues. IEEE J Oceanic Eng.

29:547-53. doi: 10.1109/JOE.2004.833220.

Westneat, M.W., & Walker, J.A. 1997.

Motor patterns of underwater flight: an

electromyographic study of the pectoral

muscles in the bird wrasse, Gomphosus varius.

J Exp Biol. 200:1881-93.


P A P E R

A Twistable Ionic Polymer-Metal CompositeArtificial Muscle for Marine ApplicationsA U T H O R SKwang J. KimDavid PugalActive Materials and ProcessingLaboratory, Department ofMechanical Engineering,University of Nevada-Reno

Kam K. LeangElectroactive Systems and ControlsLaboratory, Department ofMechanical Engineering,University of Nevada-Reno

A B S T R A C TIonic polymer-metal composite (IPMC) artificial muscles (AMs), due to their low

driving voltage (<5 V), large strain, soft and flexible structure, and ability to operatein an aqueous environment, are suited for creating artificial fish-like propulsors thatcan mimic the undulatory, flapping, and complex motions of fish fins. Herein, anewly developed IPMC AM fin with patterned electrodes is introduced for realizingmultiple degrees-of-freedom motion, such as bending and twisting. Also, by care-fully creating isolated patterns of electrodes on the surface of the polymer-metalcomposite, sections of the composite can function as an actuator, while otherareas can be used for sensing fin deformation and responses to external stimula-tion. The manufacturing, modeling, and characterization of a twistable AM fin arediscussed. The sectored electrode pattern on the AM fin is created using two tech-niques: masking and surface machining. Using first principles, detailed models aredeveloped to describe the electromechanical transduction for the IPMC AM fin.These models can be used to guide the development of more complex AM fingeometries and electrode patterns. The bending and twisting performance of aprototype twistable AM fin is evaluated and compared to the models. Experimentalresults demonstrate good twisting response for a prototype fin. Technical designchallenges and performance limitations are also discussed.Keywords: ionic polymer-metal composites, underwater propulsion, robotics

Introduction

I onic polymer-metal composite(IPMC) material is one of the mostpromising active (smart) materialsfor developing novel soft biomimeticactuators and sensors, preferably forunderwater applications (Shahinpooret al., 1998; Shahinpoor & Kim, 2001;Kim&Shahinpoor, 2003; Shahinpoor& Kim, 2005). The advantages ofthe IPMC include low driving voltage(<5 V), relatively large strain, softand flexible structure, and the abilityto operate in an aqueous environment(such as water). Over the past twodecades, researchers have made signif-icant advances in the manufacturing,modeling, design, control, and appli-cation of IPMCs for both actuationand sensing (for instance, see theworks by Alici et al., 2008; Bhat &Kim, 2004; Chen & Tan, 2008;Nemat-Nasser & Jiang, 2000; Kanget al., 2007; Kim & Shahinpoor,2003; Krishen, 2009; Lavu et al.,2005; Leo et al., 2005; Tadokoroet al., 2000; Pugal et al., 2010a;

Richardson, et al., 2003). The com-posite material has even been studiedfor energy harvesting applications inair and aqueous environments (Aureliet al., 2010b; Brufau-Penella et al.,2008; Tiwari et al., 2008). Morecomplex actuation patterns of IPMChave also been studied ( Jeon et al.,2007; Jeon & Oh, 2009; Chenet al., 2011). Herein, a newly devel-oped IPMC artificial muscle (AM)fin with patterned electrodes is intro-duced to create bending and twistingmotion. Similar concept for PPymaterials was developed by Smelaet al. (1999). Specifically, the ‘twistable’IPMC AM can be applied to developpropulsors that can mimic, for exam-ple, the flapping (pitch and heaving)

motion and complex behavior of realpectoral and caudal fish fins. Also,with careful design the electrodes oncertain areas of the AM fin can bepatterned to create a highly deformablecontrol surface, while at the same timeother regions of the AM fin can bepattern for sensing fin deformationand responses to external stimulation(Kruusamae et al., 2009). The finalresult is a compact control surfacewith integrated sensing for multifunc-tional applications in a wide spectrumof micro-autonomous robots andmarine systems.

An IPMC consists of a neutralizedionomeric membrane sandwichedbetween noble metallic electrodes (seeFigure 1). When the composite is


saturated in a polar solvent (such aswater) and then an electric field is ap-plied across the electrodes, the com-posite bends. The bending is causedby induced swelling on the cathodeside of the composite and shrinkingon the anode side [see Figure 1(b)]due to a sudden flux of cations andpolar solvent (such as water). An oppo-sitely applied voltage causes bending inthe opposite direction. Conversely,when an IPMC is mechanically de-formed, charges develop on the elec-trodes and thus IPMCs can functionas current or voltage sensor (Aliciet al., 2008; Pugal et al., 2010a).

The electromechanical behavior ofIPMCs have many noteworthy appli-cations. Due to their biocompatibility,IPMC actuators show great promise inbiomedical devices such as active endo-scopes (Yoon et al., 2007) and smartcatheters (Fang et al., 2007). Strips ofIPMCs can be used as sensors inhand prostheses (Biddiss & Chau,2006). But one of the most promisingapplications is innovative propulsionsystems for underwater autonomoussystems (Fish et al., 2008; Kamamichiet a l . , 2006; Kim et al . , 2005,

2007; Lauder, 2007). Specifically,IPMCs can replace or enhance thedesign of propulsors for underwaterwalking and swimming machines thatare currently based on traditional actu-ators such as DC motors (Ayers &Witting, 2007; Bozkurttas et al., 2008;Buchholz et al., 2008; Kato, 1998;Krieg & Mohseni, 2008; Tangorraet al., 2007), pneumatic actuators(Cai et al., 2010), and magnetic actua-tors (Tortora et al., 2010). In fact,strips of IPMCs have been used to con-struct artificial tentacles for a jellyfish-like robot (Guo et al., 2007). Thewalking speed of the jellyfish robotwas controlled through the frequencyof the input voltage applied to theIPMC-based legs. Likewise, an artifi-cial caudal fin to propel a robotic fishwas created from an IPMC actuator(Chen et al., 2010; Guo et al., 2006;Aureli et al., 2010a). The achievablepeak swimming speed of the roboticfish in (Chen et al., 2010) was reportedat 22 mm/s. In terms of performance,the maximum (stall) torque to weightratio between a prototype twistableIPMC AM with dimensions 50 mm ×25 mm × 1 mm is comparable to a

small ungeared DCmotor as illustratedin the comparison shown in Figure 2.The comparable performance ofIPMCs and traditional actuators sug-gests that IPMCs can play a criticalrole in the development of highlymaneuverable and efficient marinesystems, e.g., the system described inMenozzi et al. (2008).

Due to the nature of the materialdeformation caused by cation andsolvent flux, bending motion is themost commonly studied and appliedfor IPMCs (Kim et al., 2007). Particu-larly, when an IPMC strip is mountedin the cantilever configuration, withone end fixed and the other free, anapplied electric field to the IPMC asillustrated causes the actuator to bendas illustrated in Figure 3(a). This singledegree-of-freedom bending motionhas wide applications, such as a single-link (Aureli et al., 2010a) and multi-link (Yim et al., 2007) oscillatorypropulsor. However, IPMCs withmultiple degrees of freedom are highlydesirable to create control surfaces,which can undergo complex motionand deformation, for both stationkeeping (Bandyopadhyay et al. ,2008) as well as propulsion and ma-neuvering. It has been observed thatthe propulsion and maneuveringcapabilities of the Bluegill (Lepomismarcrochirus) sunfish is primarily dueto its highly deformable pectoral fin(Bozkurttas et al., 2008). Thus, multi-ple degrees-of-freedom IPMC actuatortechnology offers many possibilities formimicking such behavior to createmore efficient and maneuverableunderwater systems (Chen & Tan,2010). As depicted in Figure 3(b),twisting motion can be achieved bypatterning sectored electrodes com-bined with independent control eachisolated region. By carefully creat-ing electrodes on the surface of the

FIGURE 1

(a) Scanning electron microscope image of the cross-section of a Nafion-based IPMC. (b) Illus-trative movement of cations and water molecules inside of an IPMC.


polymer with proper electrical isola-tion between adjacent units, sectionsof the AM fin can be independentlycontrolled to achieve complex shapes

and deformations. Additionally, iso-lated electrodes can be patterned forsensing motion and deformation ofthe AM fin (Kruusamae et al., 2009).

IPMC ManufacturingBasic Structure of IPMC andNafion Membrane Fabrication

A basic IPMC consists of an ion ex-change polymer, such perfluorinatedalkenes or styrene/divinylbenzene-based polymers, sandwiched betweentwo noble metallic electrodes asshown in Figure 1(a). The conductingmedia can be palladium, silver, gold,carbon, graphite, and even nanotubes;however, platinum is the most com-monly used. The metal electrodes areoften chemically deposited on thepolymer’s surface through a reductionprocess (Kim & Shahinpoor, 2003).Conductive paint can also be appliedto the surface of the membrane toserve as an electrode; however, this ap-proach is not as robust and effective aselectrochemical plating.

The commonly used ion exchangemembrane Nafion (Dupont) for man-ufacturing IPMCs is easily availablefrom distributors. In the case ofNafion, the typical chemical structureis shown in Figure 4, and it consists offluorocarbons, oxygen, sulfonategroups, and a mobile cation, whichare typically either hydrogen, sodium,or lithium. Commercially availableNafion membrane such as Nafion115, Nafion 117, and Nafion 1110for fabricating IPMCs have nominaldry thicknesses of 127, 178, and254 μm, respectively (Aureli et al.,2010a). Methods to enhance the per-formance of IPMCs include boostingthe capacitance of the composite(Akle et al., 2005; Aureli et al.,2009), where this is motivated by stud-ies that correlate actuation and sensingperformance with capacitance (Akleet al., 2005). Enhancements in perfor-mance and blocking force have alsobeen made by incorporating nano-particulates into the polymer matrix

FIGURE 2

Comparison of torque-to-weight ratios for traditional motors to twistable IPMC AM fin. IPMC AM finwas fixed on one end, then 5-V DC input was applied to each electrode, with opposing polarity, andthe blocking torque of the free end was measured with an ATI Nano17 force/torque sensor.

FIGURE 3

IPMC AM fin motion: (a) bending and (b) twisting.


(Nam et al., 2003;Nguyen et al., 2007).In these studies, the nanocomposite-based IPMCs were observed to havehigher water uptake and slower waterloss, thus leading to larger bendingdisplacement and blocking force. Itwas also found that by using a dis-persing agent in the reduction processto form fine platinum polycrystalswhich subsequently lead to deeperpenetration of the platinum layer, theblocking force was increased signifi-cantly (Shahinpoor & Kim, 2001).More recent work to improve outputforce is by increasing the thickness ofthe Nafion membrane (Kim &Shahinpoor, 2002). Since relativelythick Nafion membrane is not readilyavailable, researchers have explored thesolution casting process (Kim &Shahinpoor, 2002; Kim et al., 2003;Pak et al., 2004; Shan & Leang,2009) and the hot pressing technique(Lee et al., 2006). The output forceenhancement for thicker IPMCs is ev-ident by considering two IPMC stripactuators, both having the same lengthand width, but each have a differentthicknesses, such as t and 2t (twice asthick). Assuming that for both actua-tors the same tip displacement is re-quired, then the required strain forthe thick actuator is 2ɛ. As the stresstensor for linear beam with thickness

of t, width b, and length L can be ex-pressed as (Kim & Shahinpoor, 2002)

σ t ¼ 6F tLbt2

; ð1Þ

the ratio of the stresses is

σ2t

σt≈ 2 ¼ F2t

22Ft; ð2Þ

hence F2t = 8Ft. Therefore, a thickerIPMC will produce a larger blockingforce, motivating the need to createthicker membranes for fabricatingIPMCs. At the same time, the mea-surements show that actuation speedreduces with the increase of the thick-ness. Therefore, further study is neces-

sary to find an optimal thickness of themembrane for marine applications.

Pretreatment and PlatinumPlating Process

The manufacturing of IPMCs be-gins with the pretreatment of the ionexchange membrane (Nafion) andthe platinum plating process asoutlined in the flowchart shown inFigure 5. First, the surface of the mem-brane is either mechanically roughenedor chemically etched (Yoon et al.,2007) to either enhance the capaci-tance or to improve adhesion of themetal electrode to the surface. Then,organic and metallic impurities onthe bare Nafion membrane are re-moved through a pretreatment processby initially chemically cleaning theNafion membrane in 3% hydrogenperoxide (H2O2). Next, the cleanedmembrane is rinsed in 0.5 M sulfuricacid (H2SO4) at 80 °C. Afterwards,the pretreated and cleaned Nafionmembrane is immersed in an ap-propriate metal salt solution such astetraamineplatinum (II) chloridemo-nohydrate [(Pt(NH3)4) Cl2H2O] for2 h, followed by several washings inde-ionized water. Platinum particlesare metalized on the surface of theNafion membrane by reducing the

FIGURE 4

The chemical structure of Nafion, which includes fluorocarbons, oxygen, sulfonate groups, anda mobile cation X + that can be hydrogen, sodium, or lithium. The K is usually 5–11, and the L isusually 1 (Shahinpoor & Kim, 2001).

FIGURE 5

Left: The process flow for pretreating the Nafion membrane and applying platinum electrodes.Right: fabricated IPMCs from commercially available Nafion membrane.


membrane in a sodium borohydride(NaBH4) or lithium borohydride(LiBH4) solution for 3 h. Finally, theplatinum-plating process is repeated toachieve at least three layers of platinumon the surface of the Nafion mem-brane to enhance surface conductivityand overall performance. The plati-num particulate layer is often buried1-20 μm within the IPMC surfaceand is highly dispersed.

Electrode Patterning ProcessThe sectored electrodes for the

twistable IPMC AM fin can be createdby masking, surface machining, or ab-lating the electrode using a high-powerlaser. The first approach involves theuse of a mask to cover areas of themembrane that should not be exposedto the platinum-plating process,subsequently creating an isolationregion between adjacent electrodes.The second approach uses a precisioncomputer-controlled milling machineto mechanically remove the platinumelectrodes from the surface of an IPMCmembrane to create the isolation region.

Masking Technique: The mask-ing technique to create IPMCs withsectored electrodes involves the useof UHMW (Ultra-high-molecular-weight) polyethylene tape (3M). Thetape is used to cover specific regionsof the bare Nafionmembrane to inhib-it the plating of platinum on the sur-face of the membrane. For example,the tape is applied to the bare Nafionmembrane, then the taped Nafionmembrane is processed using the pre-treatment and plating process describedabove and outlined in Figure 5. SampleIPMCs with sectored electrodes areshown in Figure 6(a), and an outlineof the process for the masking tech-nique is illustrated in Figure 7(a).

Machining Technique: Thesurface machining method utilizes a

computer-controlled milling machine,such as an automated circuit boardrouter (e.g., ProtoMat S42, LPKF) tomechanically remove the plated plati-nummetal on the surface of the Nafionmembrane. The surface machiningprocess is outlined in Figure 7(b) anddescribed as follows:1. Create the machining path to create

the electrode pattern using CADsoftware (e.g., Solidworks) or a cir-cuit board layout program (e.g.,Eagle). The result of this step is aCAD/CAM file, which is ran bythe milling machine.

2. Attach an IPMC sample (with plat-inum electrodes) to the workingsurface of the milling machineusing an adhesive layer, for exampledouble-sided tape (see Figure 7(b))or a vacuum system. Air bubblestrapped underneath the IPMCsample should be removed. Addi-tionally, the locations of the cornersof the IPMC on the working sur-face are marked for aligning thesample.

3. Load the CAD/CAM file onto themilling machine and start the mill-ing process.

4. Remove the machined IPMC sam-ple, then flip it over and attach thesample to the working surface mak-ing sure the corners are aligned withthe markings created in Step 2.Repeat Steps 2 and 3.

5. Remove the machined IPMC sam-ple and trim away excess material.Sample IPMCs with sectored

electrodes created by the surfacemachining process are shown in Fig-ures 6(b1)-6(b4). During the ma-chining process, care is taken toavoid removing too much materialfor thin IPMC membranes. In general,the machine should be set to removeonly the platinum material (approxi-mately 25-50 μm deep). Comparing

themasking and surface-machining re-sults, the machining process allowsbetter control of the shape and patternof the electrode. One of the majorchallenges of the masking process isensuring that the tape adheres to theIPMCmembrane’s surface all throughthe plating process. During machining,if the depth of cut is notwell-controlled,removal of excess material affects themechanical properties, for example,stiffness, of the IPMC actuator.

ModelingElectromechanicalTransduction

Two approaches are discussed tomodel the electromechanical bendingand twisting response of the sectored-electrode IPMCs: (1) a simplifiedfinite-element analysis (FEA) modelderived from the piezoelectric effectand (2) a more comprehensive physics-based model. The former model hasthe advantage of being easy to imple-ment due to fact that software packagesare available and specifically tailored tosolve the problem. Additionally, theestablished algorithms are computa-tionally efficient. However, the firstmethod lacks consideration of thephysical processes that governs the be-havior of the IPMC. The latter model,on the other hand, considers the un-derlying physics that includes electro-static forces, osmotic pressure, chargeimbalance and the effects of local strains(Kim & Shahinpoor, 2003; Nemat-Nasser & Jiang, 2000; Tadokoroet al., 2000; Chen & Tan, 2008; Leoet al., 2005). This type of model offersvaluable insight on the physical be-havior of the composite material andthe model can be used to help guidethe development of the material onmany levels. Despite being morerea l i s t ic—and somet imes moreaccurate—the physics-based model


is more computationally demandingto solve. Simulation results com-parison with experimental data arepresented in Performance Character-ization and Discussion

Simplified FEA Model forElectromechanical Response

A simplified finite-element modelis created to predict and gain insight

on the bending and twisting capabili-ties of the sectored-electrode IPMCAM fin. Such a model can also beapplied to optimize the design of anAM fin for specific applications. Thekey feature of this simple model isthe deformation is estimated using anequivalent bimorph beam model anal-ogous to a piezoelectric actuator (Kim& Tadokoro, 2007). As a result the

model only captures the basic electro-mechanical behavior and is thus easy toimplement and computationally effi-cient. Figure 8 shows the boundary con-ditions for the finite-element model,where the ‘clamped area’ (25 mm ×5 mm) is considered fixed. The bimorphfinite-element model consists of twoelectro-mechanical actuation layers,layers (i) and (iii), as shown in Figure 8.

FIGURE 6

Fabricated IPMCs with sectored electrodes: samples created using (a) the masking technique and (b1)-(b4) the surface-machining approach.Dimensions are in millimeters (mm).


The material separating patterned areasof the material are modeled by layer (ii).The IPMC material is treated as ahomogenous material (uniform stiffnessand density). The standard stiffnessmatrix k, piezoelectric strain matrix d,and permittivity matrix ɛ are used inthe modeling (Moheimani & Fleming,2006). The material properties are asfollows: density is 2930 kg/m3;Poisson’s ratio is 0.49; E = 1.16 GPa;d31 = d32 = 4.11 × 10-6 m/V (bending);

d31 = d32 = 2.67 × 10−7 m/V (twisting).

The finite-element model is createdusing ANSYS software (Canonsburg,PA). A SOLID98 tetrahedral coupledfield solid element is chosen for theelectro-mechanical material, while aSOLID187 tetrahedral structural solidelement is chosen for the electrode sepa-ration material. Both material propertiesare assigned a tetrahedral solid to com-pensate for the irregular mess interfacecaused by large aspect ratio between the

IPMCs thickness and length. By selectingthese elements, a uniform tetrahedralmesh throughout the entire model isachieved.

Physics-Based Model forElectromechanical Transduction

When a voltage is applied to theelectrodes of an IPMC, the freely mov-able cations inside the polymer startmigrating due to the imposed electric

FIGURE 7

Electrode patterning process: (a) the masking and (b) the surface-machining technique.

FIGURE 8

Finite-element model structure. Layers (i) and (iii) are the bimorph (electro-mechanical) layers, and layer (ii) is assumed to be a nonconductingelectrode separation layer (absent of electro-mechanical properties).


field. However, the attached anions do not migrate nor diffuse. In case of water-based IPMCs, migrating cations drag the water molecules along, causingosmotic pressure changes and therefore swelling near the cathode and contractionof the polymer near the anode. This in turn results in bending of the materialtowards the anode side. Furthermore, the migrated cations cause charge imbal-ance near the electrodes and this possibly results in the electrostatic force con-tribution to the local strain. In the following, the basic underlying equationsare presented and how to apply the calculated charge imbalance near theelectrodes in modeling the IPMC actuation in a 3-D domain is discussed(see Figure 9).

First, the ionicmigration and diffusion are described in the polymer domain byNernst-Planck and Poisson equations,

∂C∂t

þ∇· D∇C zμFC∇ϕÞ ¼ 0;ð ð3Þ

∇2ϕ ¼ F C C 0ð Þɛ

; ð4Þ

where C is the cation concentration, μ is the mobility of counter ions,D is the dif-fusion constant, F is the Faraday constant, z is the charge number, ϕ is the electricpotential, andC0 is the anion concentration with initial value ofC0 = 1200mol/m3.For the electrode, Ohm’s law for the current density is

σ∇V ¼ →j; ð5Þ

where σ is electric conductivity, V is the voltage, and→j is the current density in the

electrode. It must be noted that the electric potential ϕ inside the polymer and theelectric potential V in the electrode are different. To relate the variables ofthe electrodes V and

→j to the ionic flux inside the polymer, the Ramo-Shockley

theorem is used (Ramo, 1939; Shockley,1938), i.e.,

I ¼ 1φ∑n qn

→W →rnð Þ · →vn; ð6Þ

where →rn;→vn, and

→qn are the position vec-tor, instantaneous velocity, and chargeof cation n, respectively. The electricfield that would be produced by 1 V ap-plied potential without any charges,neither mobile nor fixed, is denoted by→W . The current in the external circuit isrepresented by I, and ϕ is a constantwith value of 1 V. Equation (6) can besimplified for a 2-D domain with paral-lel electrodes

j ¼ 1h∫ f dl ; ð7Þ

where f is an ionic flux in the electrodedirection and has a unit of Cm2s and j islocal current density at the inner bound-ary of an electrode. The term dl is theintegration element along the pathwhere the particle moves, which in thiscase is assumed to be from one electrodeto other.

The Equations (3)-(7) are used tocalculate the time dependent cationconcentration C and correspondingcharge density ρ = C − C0 in the poly-mer of IPMC in response to an appliedvoltage. To relate the charge densityρ to the physical bending, the forcecoupling similar to the one shown inPugal et al. (2008) is used. A set ofcontinuum mechanics equations wereimplemented for the polymer do-main. Normal and shear strain are bydefinition

ɛ i ¼ ∂ui∂xi

; ɛ ij ¼ 12

∂ui∂xj

þ ∂uj∂xi

; ð8Þ

where u is the displacement vector,x denotes a coordinate and indices i

FIGURE 9

Conceptual physics-based model. Calculations are done in three domains—the polymer domainand two electrode domains.


and j are in the range of 1-3 and denotecomponents correspondingly to x, y, orz direction. The stress-strain relation-ship is

σ f ¼ Dɛ ; ð9Þ

where D is a 6 × 6 elasticity matrix,consisting of components of Young’smodulus and Poisson’s ratio. The sys-tem is in equilibrium, if the relation

∇ ·σ f ¼→F ð10Þ

is satisfied. This is the Navier’s equa-tion for displacement (Heinbockel,2001). The body force and charge cou-pling is defined as

→F ¼ Aρ x; ð11Þ

where A is a parametrically determinedconstant and x is IPMC’s longitudinaldirection. This approach allows calcu-lating deformation that is in the typicalIPMC actuation range. When a verylarge deformation is expected, for in-stance, in case of a very thin mem-brane, geometric nonlinearity andmore precise force coupling must beused in the model. The conceptualmodel with the variables is illustratedin Figure 9.

The finite element method is usedto solve Equations (3), (4), and (5)with Equation (7) as the boundarycondition between the electrode andpolymer domain and Equation (10).The equations are implemented inComsol Multiphysics software pack-age (Multiphysics, 2011). The high as-pect ratio of the domain representingan IPMC and the nonlinear natureof the problem make it difficult todirectly solve the equations in afull-scale 3-D domain. Namely, theproblem size would be very large, in

the range of hundreds of thousandsof degrees of freedom. The challengesare described in more detail in Pugalet al. (2010b). To reduce the problemsize without significant loss in thecalculation precision, the followingmodeling approach for 3-D actuationof IPMC is developed.

Firstly, the cation concentrationC (from which the charge density ρcan be directly calculated) and voltageϕ are calculated in a 2-D domain.As the 2-D domain is scaled in the lon-gitudinal (x) direction of an IPMC, theelectrode conductivity value is alsolinearly scaled. The electrode currentsand corresponding voltage gradient inthe electrodes are taken into account inthe 2-D model to obtain more precisecation concentration. See, for instance,Figure 10—there the molar flux f

and electric current j are depicted. Asa result of the calculations, spatialand temporal C and ϕ in 2-D arefound and stored. This is done foreach patterned electrode that is sub-jected to a different voltage. Due tothe small Debye screening length ofthe charges near the electrodes (Porfiri,2009), the 2-D model calculations aredone on a mesh that is extremely finenear the boundaries.

Secondly, the calculatedC andϕ areextruded onto slightly coarser mesh inthe 3-D domain as illustrated inFigure 11. For instance, Figure 12shows the extruded voltage ϕ valueson the 3-D domain boundary. Itshould be noted that in the currentmodel, it is not necessary to extrudethe values of ϕ, but it was done forillustration purposes. Finally, the

FIGURE 10

Electric current streamlines in the electrodes and total ionic flux streamlines in the polymerdomain. The color depicts the total current density A

m2

in the electrodes. (Color versions of

figures available online at: http://www.ingentaconnect.com/content/mts/mtsj/2011/00000045/00000004.)


coupling Equations (11) and (10) arecarried out to calculate the time-dependent bending of an IPMC in 3-D.

PerformanceCharacterizationand Discussion

The bending and twisting perfor-mance of the fabricated IPMC AMfins are characterized and then com-pared to the electromechanical models.The experimental setup consists ofcontrol electronics connected to theIPMC AM fin and a measurement sys-tem (laser displacement sensors anddata acquisition system) for collectingthe output response. Each set of elec-trodes are independently controlledby a separate custom-designed voltageamplifier as depicted in Figure 13. Acomplete description of the experi-mental setup is described in Riddleet al. (2010). All measurements aretaken while the subject IPMC is ac-tuated in de-ionized water.

Measured Bending andTwisting Performance

Themeasured response for a selectedmasked and machined electrodeIPMC, Figure 6(b2), are shown in Fig-ure 14. The results in Figure 14 showthat both types of actuators providedapproximately the same degree of

twist. Furthermore, the actuators alsoexhibi ted non-smooth twist ingmotion, which may have been causedby the effects of the fabrication pro-cess. A peak twist angle of 7.3° is mea-sured for the machined IPMC [seeFigures 14(c) and 14(d)]. The fre-quency of the actuation of 1 Hz is

FIGURE 11

Extruding the cation concentration C that has been calculated for eachtime step in 2-D into a 3-D domain.

FIGURE 12

Extruded data from 2-D domain. The surface depicts the extruded voltagevalues on the electrodes. Notice the voltage gradient in the longitudinalx direction.

FIGURE 13

Experimental setup for measuring bending and twisting response using two non-contact lasersensors (Micro-Epsilon, optoNCDT 1402). Voltage amplifier gain is A.


sufficient for underwater propulsion(Chen et al., 2010). The bendingangle could be enhanced in variousways, depending on the applicationof interest. For instance, rigid exten-sions with different shapes can beused (Anton, 2008).

Simulated ResponsesThe bending and twisting response

of the IPMC, Figure 6(b2), producedby the simple FEA model for aninput voltage of 2 V are shown inFigures 15(a) and 15(b). For bending,the tip displacement is predicted atapproximately 2.5 mm. For twisting,the angle of twist is estimated usingθ ¼ tan1 a

b

, where a and b are

shown in Figure 15(c). The predictedmaximum angle of twist is approxi-mately 0.7° for a 2-V input. In Fig-ures 15(d) and 15(e), the FEA resultsand the response for the bending andtwisting motion measured along thelength of the IPMC actuator arecompared. For bending, the predictedand measured results agreed well,where the maximum root-mean-squared (RMS) error is approximately0.2 mm. For bending, however, theRMS error increased significantly for3 V and beyond. It is noted that thefinite-element approach assumes thematerial is isotropic and operatingwithin the linear-elastic range. Due tothe complex nature of the IPMCmaterial, the simplified FEA may belimited in its ability to accuratelypredict the performance for largeelectric fields. For example, as shownin Figure 14, the measured twistangle is approximately 7° for an inputvoltage with magnitude of 5 V. Thisresult indicates that the IPMC’s re-sponse can be highly nonlinear forlarge input voltages. Therefore, oneof the challenges is developing detailed

enough models to aid in designing the electrode patterns to meet a specificapplication.

To adequately model the twisting deformation with the physics-based electro-mechanical model presented in Physics-Based Model for ElectromechanicalTransduction, the sinusoidal voltages

u1 tð Þ ¼ 5 V½ sin 2πtð Þ; u2 tð Þ ¼ 5 V½ sin 2πtð Þ; ð12Þ

are set as time-dependent boundary conditions to the electrodes 1, 2 and 3, 4,respectively (see Figure 13). The calculated displacement fields at two differenttimes are shown in Figure 16. The model is validated against measured time re-sponse [Figure 14(c)]. The model estimation versus measurements are shown inFigure 17. It can be seen that the model predicts the twisting deformation well.Slight discrepancies in the peak values can be attributed to the fact that the model

FIGURE 14

Measured IPMC twisting response for 5 V sinusoidal input at 1 Hz: (a) input voltage applied toelectrodes; (b) twisting response for masked electrode IPMC; (c) twisting response for machinedelectrode IPMC; (d) sequence images of machined electrode IPMC showing twisting behavior.


does not take into account electro-chemical currents and therefore doesnot provide correct voltage gradient onthe electrodes for higher applied volt-ages. It must be noted that the hydro-dynamic effects are not considered inthe calculations due to the low frequencyand rather small twisting amplitude.

ConclusionsIPMC AMs are suited for creating

artificial fish-like propulsors that can

mimic the undulation, flapping, andcomplex motions of fish fins. A newlydeveloped IPMC AM fin with pat-terned electrodes was introduced forrealizing multiple degrees-of-freedommotion, such as bending and twisting.These twistable AM fins are suited forcreating artificial fish-like propulsorsthat can mimic the undulatory, flap-ping, and complex motions of fishfins as well as novel control surfacesfor applications in a wide spectrum ofmicro-autonomous robots and marine

systems. The masking and surface ma-chining fabrication process are viableapproaches to create a twistable AMfin. Experimental characterizationshowed that peak twisting for thesectored-electrode IPMC exceeded 7°.A finite-element bimorph beammodel was used to predict the bendingand twisting behavior of a selectedIPMC actuator, where good agreementbetween the measured response andmodel output was found for lowelectric fields (2 V). A full-scale 3-D

FIGURE 15

Finite-element modeling results for (a) bending and (b) twisting for 2-V input. (c) Definition of twist angle θ. Comparison of finite-element modelresults and measured response: (d) bending and (e) twisting response for different applied voltages.


physics-based model to simulate elec-tromechanical twisting transductionof IPMC was developed. Comparisonbetween the experimental results andmodel output for the electromechani-cal transduction indicated that the

model predicts the twisting deforma-tion well. The future work will explorecomplex electrode patterns, integratedsensing electrodes, an alternativeapproach to create a twistable fin struc-ture using a soft boot, and the develop-

ment of a prototype autonomousmarine system powered by the twista-ble IPMC AM fin.

AcknowledgmentsThe authors gratefully thank

the financial support from the U.S.Office of Naval Research (grantN0001409102183) and Dr. TomMcKenna. The authors also thankS.M. Kim, Y.S. Jung, S. Song,R. Riddle, and Y. Shan for their helpwith the experiments. KJK expresseshis special thanks to Dr. PromodeBandyopadhyay of the Naval Under-sea Warfare Center (NUWC) inNewport, RI, for his thoughtfulencouragement.

Lead Authors:Kwang J. KimActive Materials andProcessing LaboratoryDepartment of MechanicalEngineering, Universityof Nevada-RenoEmail: [email protected]

FIGURE 17

Comparison of experimental (solid line) and simulated twisting angle (dash line).

FIGURE 16

Calculated twisting of IPMC at t = 0.25 s (left) and at t = 0.75 s (right). The color shows y-directional displacement.


Kam K. LeangElectroactive Systems andControls LaboratoryDepartment of MechanicalEngineering, Universityof Nevada-RenoEmail: [email protected]

ReferencesAnton, M. 2008. Mechanical modeling

of IPMC actuators at large deformations.

Ph.D. Thesis, University of Tartu.

Akle, B.J., Leo, D.J., Hickner, M. A., &

McGrath, J.E. 2005. Correlation of capaci-

tance and actuation in ionomeric polymer

transducers. J Mater Sci. 40:3715-24.

doi: 10.1007/s10853-005-3312-x.

Alici, G., Spinks, G.M., Madden, J.D.,

Wu, Y., & Wallace, G.G. 2008. Response

characterization of electroactive polymers as

mechanical sensors. IEEE/ASME T Mech.

13(2):187-96.

Aureli, M., Kopman, V., & Porfiri, M.

2010a. Free-locomotion of underwater

vehicles actuated by ionic polymer metal

composites. IEEE/ASME T Mech.

15(4):603-14.

Aureli, M., Lin, W., & Porfiri, M. 2009.

On the capacitance-boost of ionic polymer

metal composites due to electroless plating:

Theory and experiments. J Appl Phys.

105:104911. doi: 10.1063/1.3129503.

Aureli, M., Prince, C., Porfiri, M., &

Peterson, S.D. 2010b. Energy harvesting from

base excitation of ionic polymer metal

composites in fluid environments. Smart

Mater Struct. 19:015003. doi: 10.1088/

0964-1726/19/1/015003.

Ayers, J., & Witting, J. 2007. Biomimetic

approaches to the control of underwater

walking machines. Philos T R Soc A.

365:273-95. doi: 10.1098/rsta.2006.1910.

Bandyopadhyay, P.R., Beal, D.N., &

Menozzi, A. 2008. Biorobotic insights into

how animals swim. J Exp Biol. 211:206-14.

doi: 10.1242/jeb.012161.

Bhat, N.D., & Kim, W.-J. 2004. Precision

force and position control of ionic polymer-metal

composite. J Syst Contr Eng. 218(6):421-32.

Biddiss, E., & Chau, T. 2006. Electroactive

polymeric sensors in hand prostheses: Bending

response of an ionic polymer metal composite.

Med Eng Phys. 28(6):568-78. doi: 10.1016/

j.medengphy.2005.09.009.

Bozkurttas, M., Tangorra, J., Lauder, G., &

Mittal, R. 2008. Understanding the hydro-

dynamics of swimming: From fish fins to

flexible propulsors for autonomous under-

water vehicles. Adv Sci Tech. 58:193-202.

doi: 10.4028/www.scientific.net/AST.58.193.

Brufau-Penella, J., Puig-Vidal, M., Giannone,

P., Graziani, S., & Strazzeri, S. 2008. Charac-

terization of the harvesting capabilities of

an ionic polymer metal composite device.

Smart Mater Struct. 17(1):015009.

doi: 10.1088/0964-1726/17/01/015009.

Buchholz, J.H.J., Clark, R.P., & Smits, A.J.

2008. Thrust performance of unsteady pro-

pulsors using a novel measurement system,

and corresponding wake patterns. Exp

Fluids. 45:461-72. doi: 10.1007/s00348-

008-0489-1.

Cai, Y., Bi, S., & Zheng, L. 2010. Design

and experiments of a robotic fish imitating

wow-nosed ray. J Bionic Eng. 7:120-126.

doi: 10.1016/S1672-6529(09)60204-3.

Chen, Z., Shatara, S., & Tan, X. 2010.

Modeling of biomimetic robotic fish propelled

by an ionic polymer-metal composite caudal

fin. IEEE/ASME T Mech. 15:448-59.

Chen, Z., & Tan, X. 2008. A control-

oriented and physics-based model for ionic

polymer-metal composite actuators. IEEE/

ASME T Mech. 13(5):519-29.

Chen, Z., & Tan, X. 2010. Monolithic

fabrication of ionic polymer-metal composite

actuators capable of complex deformation.

Sensor Actuat A-Phys. 157:246-57.

doi: 10.1016/j.sna.2009.11.024.

Chen, Z., Um, T., & Bart-Smith, H. 2011.

A novel fabrication of ionic polymer-metal

composite actuator capable of 3-dimensional

kinematic motions. Sensor Actuat A-Phys.

168(1):131-9. doi: 10.1016/j.

sna.2011.02.034.

Fang, B.-K., Ju, M.-S., & Lin, C.-C. K. 2007.

A new approach to develop ionic polymer-

metal composites (IPMC) actuator: Fabrication

and control for active catheter systems. Sensor

Actuat A-Phys. 137(2):321-9. doi: 10.1016/

j.sna.2007.03.024.

Fish, F.E., Howle, L.E., & Murray, M.M.

2008. Hydrodynamic flow control in marine

mammals. Integr Comp Biol. 48:1-13.

doi: 10.1093/icb/icn029.

Guo, S., Ge, Y., Li, L., & Liu, S. 2006.

Underwater swimming micro robot using

IPMC actuator. In: Proceedings of the

2006 IEEE International Conference on

Mechatronics and Automation, June 25-28.

pp. 249-54. Piscataway, NJ: IEEE.

doi: 10.1109/ICMA.2006.257505.

Guo, S., Shi, L., Ye, X., & Li, L. 2007. A

new jellyfish type of underwater microrobot.

In: Proc. IEEE International Conference on

Mechatronics and Automation. pp. 509-14.

Piscataway, NJ: IEEE.

Heinbockel, J. 2001. Introduction to

Tensor Calculus and Continuum Mechanics.

Bloomington, IN: Trafford Publishing.

Jeon, J.H., & Oh, I.K. 2009. Selective growth

of platinum electrodes for MDOF IPMC

actuators. Thin Solid Films. 517:5288-92.

doi: 10.1016/j.tsf.2009.03.111.

Jeon, J.H., Oh, I.K., Han, J.H., & Lee,

S.W. 2007. Development of bio-mimetic

patterned IPMC actuators with multiple

electrodes. Key Eng Mat. 334:1005-8.

doi: 10.4028/www.scientific.net/KEM.

334-335.1005.

Kamamichi, N., Yamakita, M., Asaka, K.,

& Luo, Z.-W. 2006. A snake-like swim-

ming robot using IPMC actuator/sensor.

In: IEEE International Conference on

Robotics and Automation. pp. 1812-7.


Kang, S., Shin, J., Kim, S.J., Kim, H.J., &

Kim, Y.H. 2007. Robust control of ionic

polymer-metal composites. Smart Mater


Struct. 16(6):2457-63. doi: 10.1088/

0964-1726/16/6/049.

Kato, N. 1998. Locomotion by mechanical

pectoral fins. J Mar Sci Technol. 3:113-21.

doi: 10.1007/BF02492918.

Kim, B., Kim, B.M., Ryu, J., Oh, I.-H.,

Lee, S.-K., Cha, S.-E., & Pak, J. 2003.

Analysis of mechanical characteristics of the

ionic polymer metal composite (IPMC)

actuator using cast ion-exchange film. In:

Proceeding of the SPIE Smart Structures

and Materials: Electroactive Polymer Actua-

tors and Devices (EAPAD). pp. 486-95.


Kim, B., Kim, D.-H., Jung, J.J., & Park,

J.-O. 2005. A biomimetic undulatory tadpole

robot using ionic polymer-metal composite

actuators. Smart Mater Struct. 14:1579-85.

Kim, K.J., & Shahinpoor, M. 2002. A novel

method of manufacturing three-dimensional

ionic polymer-metal composites (ipmcs)

biomimetic sensors, actuators and artificial

muscles. Polymer. 43(3):797-802.

doi: 10.1016/S0032-3861(01)00648-6.

Kim, K.J., & Shahinpoor, M. 2003. Ionic

polymer-metal composites: II. Manufacturing

techniques. Smart Mater Struct. 12(1):65-79.

doi: 10.1088/0964-1726/12/1/308.

Kim, K.J., & Tadokoro, S. 2007. Electroac-

tive Polymers robotics Applications: Artificial

Muscles and Sensors. London: Springer-

Verlag.

Kim, K.J., Yim,W., Paquette, J.W.,&Kim,D.

2007. Ionic polymer-metal composites

for underwater operation. J Intel Mat Syst

Str. 18(2):123-31. doi: 10.1177/

1045389X06063468.

Krieg, M., & Mohseni, K. 2008. Thrust

characterization of a bioinspired vortex ring

thruster for locomotion of underwater

robots. IEEE J Oceanic Eng. 33(2):123-32.

doi: 10.1109/JOE.2008.920171.

Krishen, K. 2009. Space applications for ionic

polymer-metal composite sensors, actuators,

and artificial muscles. Acta Astronaut.

64(11-12):1160-6.

Kruusamae, K., Brunetto, P., Graziani, S.,

Punning, A., Di Pasquale, G., & Aabloo, A.

2009. Self-sensing ionic polymer-metal

composite actuating device with patterned

surface electrodes. Polym Int. 59(3):300-4.

doi: 10.1002/pi.2752.

Lauder, G.V. 2007. How fish swim: Flexible

fin thrusters as an EAP platform. In: Pro-

ceedings of the SPIE Smart Structures and

Materials: Electroactive Polymer Actuators

and Devices (EAPAD). pp. 652402. San

Diego, CA: SPIE.

Lavu, B.C., Schoen, M.P., & Mahajan, A.

2005. Adaptive intelligent control of ionic

polymer metal composites. Smart Mater

Struct. 14(4):466-74. doi: 10.1088/

0964-1726/14/4/002.

Lee, S.J., Han, M.J., Kim, S.J., Jho, J.Y., Lee,

H.Y., & Kim, Y.H. 2006. A new fabrication

method for ipmc actuators and application to

artificial fingers. Smart Mater Struct. 15:

1217-24. doi: 10.1088/0964-1726/15/5/008.

Leo, D.J., Farinholt, K., &Wallmersperger, T.

2005. Computational models of ionic

transport and electromechanical transduction

in ionomeric polymer transducers. In: Pro-

ceedings of the SPIE Smart Structures and

Materials: Electroactive Polymer Actuators

and Devices (EAPAD). pp. 170-81.

Menozzi, A., Leinhos, H.A., Beal, D.N., &

Bandyopadhyay, P.R. 2008. Open-loop con-

trol of a multifin biorobotic rigid underwater

vehicle. IEEE J Oceanic Eng. 33(2):59-68.

doi: 10.1109/JOE.2008.918687.

Moheimani, S.O.R., & Fleming, A.J. 2006.

Piezoelectric Transducers for Vibration Con-

trol and Damping (Advances in Industrial

Control). London: Springer.

Multiphysics, C. 2011. User’s Guide,

version 4.1 a. COMSOL AB. Los Angeles, CA.

Nam, J.D., Choi, H.R., Tak, Y.S., & Kim,

K.J. 2003. Novel electroactive, silicate

nanocomposites prepared to be used as

actuators and artificial muscles. Sensor

Actuat A-Phys. 105:83-90.

Nemat-Nasser, S., & Jiang, Y.L. 2000.

Electromechanical response of ionic

polymer-metal composites. J Appl Phys.

87(7):3321-31. doi: 10.1063/1.372343.

Nguyen, V.K., Lee, J.W., & Yoo, Y. 2007.

Characteristics and performance of ionic

polymer-metal composite actuators based

on nafion/layered silicate and nafion/silica

nanocomposites. Sensor Actuat B. 120:

529-37. doi: 10.1016/j.snb.2006.03.015.

Pak, J.J., Kim, J., Oh, S.W., Son, J.H.,

Cho, S.H., Lee, S.-K., … Kim, B. 2004.

Fabrication of ionic-polymer-metal-composite

(IPMC) micropump using a commercial

nafion. In: Proceedings of the SPIE Smart

Structures and Materials: Electroactive Polymer

Actuators and Devices (EAPAD). pp. 272-80.


Porfiri, M. 2009. Influence of electrode

surface roughness and steric effects on the

nonlinear electromechanical behavior of

ionic polymer metal composites. Phys Rev E.

79(4):041503. doi: 10.1103/

PhysRevE.79.041503.

Pugal, D., Jung, K., Aabloo, A., & Kim,

K.J. 2010a. Ionic polymer metal composite

mechanoelectrical transduction: Review and

perspectives. Polym Int. 59(3):279-89.

doi: 10.1002/pi.2759.

Pugal, D., Kim, K.J., & Aabloo, A. 2010b.

Modeling the transduction of IPMC in

3D configurations. In: Proceedings of the

SPIE Smart Structures and Materials and

Nondestructive Evaluation and Health

Monitoring. pp. 76441T. San Diego, CA:

SPIE.

Pugal, D., Kim, K.J., Punning, A., Kasemagi,

H., Kruusmaa, M., & Aabloo, A. 2008.

A self-oscillating ionic polymer-metal com-

posite bending actuator. J Appl Phys.

103(8):084908-6. doi: 10.1063/1.2903478.

Ramo, S. 1939. Currents induced by

electron motion. Proc IRE. 27(9):584-5.

doi: 10.1109/JRPROC.1939.228757.

Richardson, R.C., Levesley, M.C., Brown,

M.D., Hawkes, J.A., Watterson, K., &

Walker, P.G. 2003. Control of ionic polymer

metal composites. IEEE/ASME T Mech.

8(2):245-3.


Riddle, R.O., Jung, Y., Kim, S.-M., Song, S.,

Stolpman, B., Kim, K.J., & Leang, K.K.

2010. Sectored-electrode IPMC actuator for

bending and twisting motion. In: Proceedings

of the SPIE Smart Structures and Materials

and Nondestructive Evaluation and Health

Monitoring. San Diego, CA: SPIE.

Shahinpoor, M., Bar-Cohen, Y., Simpson,

J.O., & Smith, J. 1998. Ionic polymer-metal

composites (IPMCs) as biomimetic sensors,

actuators and artificial muscles—a review.

Smart Mater Struct. 7(6):R15-30.

doi: 10.1088/0964-1726/7/6/001.

Shahinpoor, M., & Kim, K.J. 2001. Ionic

polymer-metal composites: I. Fundamentals.

Smart Mater Struct. 10:819-33.

doi: 10.1088/0964-1726/10/4/327.

Shahinpoor, M., & Kim, K.J. 2005. Ionic

polymer-metal composites: IV. Industrial and

medical applications. Smart Mater Struct.

14(1):197-214. doi: 10.1088/0964-

1726/14/1/020.

Shan, Y., & Leang, K.K. 2009. Frequency-

weighted feedforward control for dynamic

compensation in ionic polymer-metal

composite actuators. Smart Mater Struct.

18(12):125016. doi: 10.1088/0964-1726/18/

12/125016.

Shockley, W. 1938. Currents to conductors

induced by a moving point charge. J Appl

Phys. 9(10):635-6. doi: 10.1063/1.1710367.

Smela, E., Kallenbach, M., & Holdenried, J.

1999. Electrochemically driven polypyrrole

bilayers for moving and positioning bulk

micromachined silicon plates. IEEE/ASME

T Microelectromech Syst. 8(4):373-83. doi:

10.1109/84.809051.

Tadokoro, S., Yamagami, S., & Takamori, T.

2000. An actuator model of ICPF for robotic

applications on the basis of physicochemical

hypotheses. In: IEEE International Conference

on Robotics and Automation. pp. 1340-6.


Tangorra, J.L., Davidson, S.N., Hunter,

I.W., Madden, P.G.A., Lauder, G.V., Dong,

H., … Mittal, R. 2007. The development

of a biologically inspired propulsor for

unmanned underwater vehicles. IEEE

J Oceanic Eng. 32(3):533-50.

doi: 10.1109/JOE.2007.903362.

Tiwari, R., Kim, K.J., & Kim, S. 2008. Ionic

polymer-metal composite as energy harvesters.

Smart Struct Syst. 4(5):549-63.

Tortora, G., Caccavaro, S., Valdastri, P.,

Menciassi, A., & Dario, P. 2010. Design of an

autonomous swimming miniature robot based

on a novel concept of magnetic actuation.

In: IEEE International Conference on

Robotics and Automation. pp. 1592-7.

Piscataway, NJ: IEEE. doi: 10.1109/

ROBOT.2010.5509599.

Yim, W., Lee, J., & Kim, K.J. 2007. An

artificial muscle actuator for biomimetic

underwater propulsors. Bioinspir Biomim.

2:S31-41. doi: 10.1088/1748-3182/2/2/S04.

Yoon, W.J., Reinhall, P.G., & Seibel, E.J.

2007. Analysis of electro-active polymer

bending: A component in a low cost

ultrathin scanning endoscope. Sensor Actuat

A-Phys. 133(2):506-17. doi: 10.1016/

j.sna.2006.04.037.


P A P E R

Batoid Fishes: Inspiration for the NextGeneration of Underwater RobotsA U T H O R SKeith W. MooredMechanical and AerospaceEngineering Department,Princeton University

Frank E. FishDepartment of Biology,University of West Chester

Trevor H. KempHilary Bart-SmithMechanical and AerospaceEngineering Department,University of Virginia

A B S T R A C TFor millions of years, aquatic species have utilized the principles of unsteady

hydrodynamics for propulsion and maneuvering. They have evolved high-enduranceswimming that can outperform current underwater vehicle technology in the areasof stealth, maneuverability and control authority. Batoid fishes, including the mantaray, Manta birostris, the cownose ray, Rhinoptera bonasus, and the Atlantic sting-ray, Dasyatis sabina, have been identified as a high-performing species due to theirability to migrate long distances, maneuver in spaces the size of their tip-to-tipwing span, produce enough thrust to leap out of the water, populate many under-water regions, and attain sustained swimming speeds of 2.8 m/s with low flapping/undulating frequencies. These characteristics make batoid fishes an ideal platformto emulate in the design of a bio-inspired autonomous underwater vehicle. Theenlarged pectoral fins of each ray undergoes complexmotions that couple spanwisecurvature with a chordwise traveling wave to produce thrust and to maneuver. Re-searchers are investigating these amazing species to understand the biological prin-ciples for locomotion. The continuum of swimming motions—from undulatory tooscillatory—demonstrates the range of capabilities, environments, and behaviorsexhibited by these fishes. Direct comparisons between observed swimming mo-tions and the underlying cartilage structure of the pectoral fin have been made. Asimple yet powerful analytical model to describe the swimming motions of batoidfishes has been developed and is being used to quantify their hydrodynamic perfor-mance. This model is also being used as the design target for artificial pectoralfin design. Various strategies have been employed to replicate pectoral fin motion.Active tensegrity structures, electro-active polymers, and fluid muscles are threestructure/actuator approaches that have successfully demonstrated pectoral-fin-like motions. This paper explores these recent studies to understand the relation-ship between form and swimming function of batoid fishes and describes attemptsto emulate their abilities in the next generation of bio-inspired underwater vehicles.Keywords: biomimicry, bioinspired, autonomous underwater vehicle, manta ray,tensegrity structures

Introduction

There has been an explosion of ac-tivity in the area of biomimicry andbioinspired engineering research. Bio-mimicry directly emulates the formand function of species to illuminatethe physical principles behind nature’sdesigns. Bioinspired engineering takesadvantage of the knowledge gainedthrough biomimetic studies to judi-ciously apply novel physical principlesto develop solutions with added func-tionality over conventional engineer-ing approaches. Biology, biomimeticsand bioinspired engineering are inti-mately linked. Thus it comes as no sur-prise that biologists and engineers arecollaborating by developing bioroboticdevices to: (1) elucidate key insightsinto biological form and function and(2) develop bioinspired autonomousunderwater vehicles (BAUVs) to im-prove functionality of autonomousunderwater vehicles (AUVs).

Aquatic species outperform con-ventional AUVs in the areas of ma-

neuverability and control authority(Bandyopadhyay, 2005), while havinga low-noise signature that blends intothe background and high swimmingefficiencies. Batoid rays excel in allof these areas, giving them an abun-dance of recent attention. The focusof this paper is to present the growingbody of work being done to under-stand and quantify the swimming per-

formance of batoid fishes (i.e., skates,sting rays, manta rays) and the state-of-the-art in robotic mimicry. Of par-ticular interest are the mechanismsassociated with the swimming ofthese fishes, which employ flattenedpectoral fins to propel and maneuverin the ocean and in rivers.

Understanding Biological Founda-tion will discuss our current biological


understanding of batoid rays. Ratio-nale for Mimicking Batoid Rays willpresent compelling reasons for sci-entists and engineers to study batoidsrays, highlighting their swimmingcharacteristics that would be desiredin an underwater vehicle. BioinspiredRobotics delves into the expandingworld of bio-inspired underwater ve-hicles, with particular emphasis onray-like platforms. Concluding Com-ments and Future Directions concludeswith a discussion on critical areas thatneed to be addressed in order for thenext generation of underwater vehiclesto be truly bioinspired.

Understanding BiologicalFoundation

Fish swim by imparting momen-tum to water from the movementsof a variety of propulsors, which caninclude the body, median fins, andpaired fins (Sfakiotakis et al., 1999).Although primitive batoid fishes usethe body and caudal fin to swim,more advanced batoids have becomespecialized to swim with enlarged pec-toral fins. It is emphasized that pecto-ral fin locomotion can have significantadvantages inmaneuvering and station-keeping (Sfakiotakis et al., 1999). Inrecent years, more attention is beingpaid to pectoral fin hydrodynamics astheir importance is being realized in notonly steady-state locomotion but also intransient maneuvers (Bandyopadhyay,

2005; Lauder et al., 2002; Combes &Daniel, 2001; Palmisano et al., 2007).

Batoid rays take pectoral fin lo-comotion to an evolutionary extreme(Figure 1). Rays have a dorso-ventrallyflattened body with enlarged pectoralfins that are seamlessly merged withtheir body to form a biological blendedwing-body configuration. Propulsivewaves are passed through the fins byserial contraction of the appendicularmusculature. The waves have theirgreatest amplitude toward the periph-ery of the fin.

Even though among rays there issimilar morphology, their locomotorstrategies can be very different. Un-dulatory motion, defined as havinggreater than one or more waves presenton a fin (Rosenberger, 2001), is oneextreme of kinematic motion and wastermed ‘rajiform’ by Breder (1926).These fishes swim just over the oceanfloor. The other extreme is oscillatorymotion, defined as having less thanhalf of a wave present (flapping) on afin, and was coined ‘mobuliform’ byWebb (1994). The mobuliform swim-ming mode appears as a wing-like flap-pingmotion and is associated with raysthat have a more pelagic existence. Thevarious species of batoids exhibit acontinuum of kinematic motions be-tween the two extremes of undulationand oscillation (Rosenberger, 2001).Myliobatoids, i.e., the mid-waterrays, including manta, eagle, bat, and

cownose rays, nearly exclusively utilizeoscillatory motion (Klausewitz, 1964;Sasko et al., 2006; Heine, 1992).Some research has been done to char-acterize the biology and behavior ofmyliobatoids (Schaefer & Summers,2005; Summers, 2000). Heine (1992)studied the kinematics of the cownoseray by videotaping live rays swimmingin a flow tank. Rosenberger (2001)compared the kinematics of manybatoid rays spanning the undulation-oscillation continuum and suggestedthat oscillatory rays have evolved tohave efficient locomotion. Klausewitz(1964) describes the kinematic mo-tions of the manta ray while Mooredet al. (Moored, 2010; Moored et al.,2011b) developed a simple yet power-ful analytical model to quantify the ki-nematics of different species of batoidrays (such as the manta ray, Atlanticstingray, and the cownose ray). Thismodel is used as a target deformationfield for a bio-inspired fin (Mooredet al., 2011b) and to calculate theswimming performance of differentbatoid ray species (Moored et al.,2011b; Pederzani et al., 2011).

MorphometricsThe greatly enlarged pectoral fins

form wide lateral extensions of thebody that range in morphology froma circular disc to triangular, wing-likeplanforms. Species of batoids showover a 90-fold range in size with thelargest being the manta (Manta biros-tris). Rays that swim by undulationsof the fin in the rajiform mode havefin shapes with relatively low aspectratios (the ratio of span to chord). Os-cillatory swimmers, using the mobuli-form mode, possess higher aspect ratiofins with longer spans.

The cross-sectional geometry ofbatoid rays has a streamlined appear-ance. Rajiform swimmers have a

FIGURE 1

(a) Image of a manta ray. (b) Image of a cownose ray. Both rays are part of the batoid family andswim via an oscillatory motion.


body and pectoral fins with a flattenedventral side and low vaulted dorsum,giving a design similar to a camberedwing. Although the central portion ofthe body shows a slight asymmetrywith a flattened ventral surface andconvex dorsal surface, the pectoralfins of mobuliform swimmers dis-play symmetrical cross-sectional pro-files reminiscent of engineered foils(Abbott & von Doenhoff, 1949).

The internal skeleton of the pecto-ral fins of batoids are composed of nu-merous short, cylindrical cartilaginouselements (Heine, 1992; Schaefer &Summers, 2005). These cartilaginouselements are the supportive radials ofthe fin. The radials are stacked end toend. The radial cartilages are mineral-ized to varying degree depending onthe species of batoid, where the mi-neralization is found on the exteriorof the cartilaginous element with thecore being unmineralized (Schaefer &Summers, 2005). Rajiform swimmingrays display joint staggering with lit-tle calcification of the joints, whereasthe skeleton of oscillatory swimmersshows cross-bracing and calcification.The skeleton is moved by long thinmuscles that run from the expandedpectoral girdle along each fin ray toevery radial. The range of motion ofthe articulated radials is small (∼15°),but the large number of componentsin the pectoral fins permits sufficientspanwise and chordwise flexibilityfor propulsion and maneuvering(Rosenberger, 2001; Klausewitz,1964; Heine, 1992; Schaefer &Summers, 2005).

Rationale for MimickingBatoid Rays

With respect to pursuing bio-inspired engineering, a key questionto be answered is to explain/justify

why a particular species is a good can-didate to emulate. These reasons canbe very diverse and are motivated bythe particular application envisioned.In the case of AUVs, compelling rea-sons to consider biology as a startingpoint for the development of the nextgeneration vehicle are (1) a stealthysignature, (2) high efficiency and econ-omy, (3) expanded working environ-ment, and (4) scalability/payloadcapacity. Additionally, a key justifica-tion for this approach is that there aretangible improvements that can bemade over current AUV technologies.The pool of species to emulate is vast.However, recent studies of batoid rayshave demonstrated significant swim-ming abilities that would be desirablein an underwater vehicle.

Stealth means either quiet opera-tion or the ability to blend into thebackground noise. A biomimetic ap-proach naturally fulfills these require-ments by creating a vehicle whoseminimal noise signature blends inwith the environment. The noise sig-nature of a fish is very different tothat of a propeller (Bandyopadhyay,2005). Even biomimetic sensor arrayssuch as an artificial lateral line (Yanget al., 2006) or artificial seal whiskers(Stocking et al., 2010) that are sensi-tive to hydrodynamic wake signatureswould presumably delineate a flappingfin wake as an animal and a propellerwake as a man-made device. Given re-cent advances in underwater imagingtechnology including LIDAR systems( Jaffe et al., 2001), synthetic aper-ture sonar (Kocak & Caimi, 2005) andbiomimetic sonar systems (Dobbins,2007), the shape and movements ofan AUV are becoming increasingly im-portant. By mimicking the body formof an aquatic species, the identificationof such a stealthy vehicle as a man-made device becomes difficult.

Batoid rays offer an intriguing de-sign solution for a high-endurance ve-hicle. Pelagic rays, such as the mantaray or cownose ray, migrate thousandsof miles a year. This suggests that thesespecies have evolved to become high-endurance swimmers. As discussedpreviously, myliobatoid rays have adorso-ventrally flattened body withenlarged pectoral fins, forming a natu-ral gliding morphology. In terms ofa vehicle, a batoid-inspired UV is anadvance on current underwater glidertechnology. This bioinspired platformwould enable a vehicle to have high-endurance capabilities like current un-derwater gliders, such as the SlocumAUV (Webb & Simonetti, 1999). Ad-ditionally, this system has the potentialto transition to a faster, more maneu-verable vehicle that can operate in dy-namic environments such as thelittoral zone or areas with large cur-rents and high wave action. Observa-tions of various rays show them to behighly maneuverable and adaptableto local conditions—for example, tostation keep and even swim back-wards. Their ability to control theirstability via the pectoral fins, especiallywhen compensating for challengingenvironments, must also be consideredas a desirable characteristic to emulatein an underwater vehicle.

Scalability is an attractive featurein any artificial system. With respectto bioinspired underwater vehicles, ba-toids display an extraordinary range ofdimensions, growing in excess of9 m tip-to-tip in the case of mantarays. Thus, the size and speed thatbatoids perform at are equivalent tothe operation range of marine vehicles.The size of the vehicle will very muchdepend on the mission requirements,but by using the batoid as the foun-dation, it is feasible to produce a varietyof sized vehicles that can explore and


traverse a wide range of ocean spacewhile performing a wide range of tasks.Moreover, a batoid-inspired vehiclewould have a large planform surfacearea, making this platform an excellentcandidate for flexible solar cells to ex-tend its range (Dennler et al., 2008),similar to the solar powered SAUV IIvehicle ( Jalbert et al., 2003). In ad-dition, the rigid body of batoids per-mits space for control systems, sensorydevices and increased payload.

Bioinspired RoboticsThere has been growing use of

AUVs in recent years with over240 different AUVplatforms developedand used in the field (Bandyopadhyay,2005). These AUVs typically are builtfor reconnaissance/surveying and wereoriginally designed for endurance(Blidberg, 2001). This gave rise tothe design of underwater glidersusing conventional design principles(i.e., steady-state hydrodynamics)that have high endurance but little ma-neuverability (Webb & Simonetti,1999). From another perspective,biology has created thousands of swim-ming p l a t f o rms tha t c an ou t -maneuver the best AUVs while stillhaving highly efficient, high-speedand high-endurance performance.Moreover, many of these biologicalsystems can also hover in place with noforward locomotion, generate largeenough forces to hold station underadverse environmental conditions,burst with incredible acceleration andhave a significantly reduced noise sig-nature compared to man-made AUVs(Fish & Lauder, 2006). In an attemptto bridge the performance gap be-tween conventional AUVs and bio-logical systems, engineers have beenshifting focus to BAUVs, which is ahighly multi-disciplinary research area

(Bandyopadhyay, 2005; Colgate &Lynch, 2004). To reach some of thesegoals, there is a spectrum of first gen-eration BAUVs that have been devel-oped. Some form an exotic collectionmimicking lamprey (Ayers et al.,2000; Crespi et al., 2004), tuna(Barrett et al., 1996; Yu et al., 2004;Anderson &Chhabra, 2002), and dol-phins (Yu et al., 2007), while othersare more conventional style AUVdesigns outfitted with bioinspired flap-ping propulsors (Fish et al., 2003; Low& Willy, 2006; Listak et al., 2005;Borgen et al., 2003; Mojarrad, 2000;Licht et al., 2004). These differentBAUV designs were made possiblepartly from advances in our under-standing of unsteady hydrodynamicsand the biology of nektonic (swim-ming) organisms. However, this BAUVtechnology still has a long way to gobefore the performance gap is bridged.

Recently, researchers have turnedto pectoral fin locomotion for inspira-tion. Pectoral fin motions utilized bysunfish, perch, bass and bird wrassefor low-speed swimming and maneu-vering have been studied (Gibb et al.,1994; Drucker & Jensen, 1997;Lauder & Jayne, 1996; Walker &Westneat, 1997). To understand theforces and moments produced bythese pectoral fins, biorobotic solu-tions began with paddle-like fins thatmimic the bulk kinematics of labri-form swimming (Kato, 1998; Kato &

Furushima, 2002). In recent years, de-vices have been constructed to moreaccurately replicate the kinematicsutilized by the fish with the advent ofactively flexible fins that can pro-duce chordwise undulations as well asspanwise curvature (Yan et al., 2010;Palmisano et al., 2008; Kato et al.,2008; Tangorra et al., 2007). Activelyflexible fins deform due to the presenceof actuators instead of undergoingrigid body motions, like the heaveand pitch of oscillating airfoils. Themotion of actively flexible fins is fullyprescribed. In contrast, passively flexi-ble fins deform under fluid loadingsuch that their motion is not fully pre-scribed, but a function of the forces ap-plied to the fin. Tangorra et al. (2008a,2008b) advanced their artificial pecto-ral fin (Figure 2) by not only matchingthe kinematics of the sunfish but alsoreplicating the internal fin structureand material properties, which allowedthe fin to have a greater degree of pas-sive flexibility. This fin was used tofully characterize how sunfish produceand manipulate fluid forces to pro-pel themselves and maneuver. Withequivalent passive flexibility as thefins of the sunfish, the artificial finwas able to produce thrust on boththe outstroke and instroke of its finbeat, as observed of the animal.

An excellent example demonstrat-ing the link between biology, biomi-metics, and bioinspired engineering

FIGURE 2

A biomimetic sunfish pectoral fin (Tangorra et al., 2008a, 2008b).


is in the development of an artificialghost knifefish (Curet et al., 2010).Through observation of biology, abiorobotic device was developed andused to understand the locomotionstrategies of this fish. Particle imagevelocimetry, in conjunction with com-putational fluid dynamics, were em-ployed to explore the propulsive andstation-keeping characteristics of thisfascinating fish.

Batoid-Inspired DevicesThere have also been attempts by

researchers to develop batoid-inspiredfins and AUVs. These devices mimicboth the undulatory swimming seenin benthic rays (similar to the loco-motion of the ghost knifefish) as wellas the oscillatory swimming seen in pe-lagic rays, such as the manta. Many ofthese batoid-inspired devices are usedas a platform for exploring actuationtechnologies.

Motors and servomotors are usedin ray-like devices due to their simplecontrollability, high-speed operationand repeatability. Some motor-drivendevices mimic undulatory rays (Low& Willy, 2006; V. y Alvarado et al.,2010), while others mimic oscillatoryrays (Yang et al., 2009; Zhou &Low, 2010; Gao et al., 2007). Re-searchers have developed oscillatoryray-like vehicles based on pneumaticpectoral fins (Brower, 2006; Caiet al., 2010; Suzumori et al., 2007).Sfakiotakis et al. (2005) also usedpneumatically driven “fin rays” to pro-duce an undulatory ray-like device.Festo (2008) has built a BAUV calledAquaRay, utilizing fluidic muscles.This robot uses an oscillatory flap-ping motion to swim, but no quantita-tive data on the performance is given.Takagi et al. (2007) utilized ionicpolymer-metal composites (IPMCs)actuators to develop a stingray-like

device that could achieve a swimmingspeed of 0.24 BL/s. Chen et al. (2011)developed a novel fabrication methodto produce IPMCs that can deformwith complex three-dimensional kine-matics. This fabrication technologywas used to produce a manta ray-likedevice. Shape memory alloys havealso been employed in the design of ar-tificial pectoral fins (Yong-hua et al.,2007; Wang et al., 2008). Wanget al. (2008) presented a robotic squidutilizing a rajiform mode of swim-ming to achieve 0.24 BL/s swimmingspeed. The best swimming speed per-formance of these actuator platformswas 1.4 BL/s achieved by the servo-motor driven devices; however, theassociated power cost is not given.

These studies showcase the pleth-ora of actuator technologies that canbe utilized to produce deformationssimilar to that of rays. One concernwith this approach is that the actuatorchoice is directly coupled with the fintechnology. An alternative approach isto start with a fin design that is actua-tor independent and so the choice ofactuator is dependent on the applica-tion of the vehicle. For instance, if ac-tuator efficiency is not a concern butnoiseless operation is of prime impor-tance, an SMA actuator could be cho-sen. Also, this approach opens up thepossibility of replacing current actua-tors with new technologies that maybe superior. Solutions like this are dis-cussed next.

In a study to increase our under-standing of the hydrodynamics ofbatoid locomotion, Clark and Smits(2006) designed and built an activeartificial oscillating fin that was inde-pendent of the actuator. They quan-tified the performance of the fin bymeasuring the efficiency and thrustproduction, as well as determining anoptimal traveling wave wavelength.

Furthermore, by using dye flow visual-ization, they characterized the wakestructure as a series of interacting trail-ing edge vortices forming a three-dimensional reverse von Kármánvortex street. In free swimming tests(Moored et al., 2011a), a swimmingspeed of 2 BL/s and a swimming econ-omy, ζ (ζ = U = Pf , where U is theswimming speed and Pf is the powerconsumption), of 0:132 BL/J wasreached for an actively flexible singlefin. When some passive flexibilitywas introduced, the swimming speeddropped to 1:7 BL/s while the econ-omy rose to 0:18 BL/J at the same flap-ping frequency of 2 Hz. This workhas also highlighted the prime impor-tance of the traveling wave in ray-likepropulsion.

Another actuator-independent ap-proach has been developed using activetensegrity structures. Tensegrity struc-tures are truss-like structures wheresome of the rigid elements have beenreplaced by cable elements (Figure 3).

The cable elements must be in astate of tension for the structure tohave integrity, giving rise to the con-traction of “tensional-integrity” totensegrity. Tensegrity structures actas a “skeleton-tendon” foundationthat can use any actuator type to sup-port the generation of large loads,match the kinematics of batoid rays,and perform with minimal actuationenergy (Moored et al., 2011b; Moored& Bart-Smith, 2009).

Various tensegrity actuation strate-gies are explored that are capable ofmatching the key kinematic featuresof batoid-propulsion: a chordwisetraveling wave coupled with a largeamplitude curved spanwise defor-mation. The strategies involve eitherembedding the actuators into the ten-segrity structure (embedded actuation)or migrating the actuators outside of


the structure (remote actuation). Withrespect to embedded actuation, opti-mal solutions have been calculatedthat give the location and actuationstrain necessary to match a target dis-placement field (Moored & Bart-Smith, 2007). However, embeddedactuation is problematic, as it requiresmany actuators to match the complexray kinematics, adds mass to theactive structure (thereby requiringmore power to flap) and limits thescalability of solutions to the size ofthe actuator. Remote actuation over-comes these limitations by placingthe actuators outside of the active re-gion and connecting to the structurevia a routed cable. A general numericalmodel––applicable to any topologyand any actuation strategy––hasbeen derived (Moored & Bart-Smith,2009).

Moored et al. (2011) derive analyti-cal solutions for active planar tensegritybeam structures. These solutions cou-pled with the numerical solution areutilized to identify optimal stiffness-to-mass and strength-to-mass strate-gies. Structural performance metricswere calculated showing that the finstructure can closely match the kine-matics of the manta ray, under externalloading, using open-loop actuation offour actuators remotely located outsideof the active structure (Moored et al.,2011b). In an attempt to simplify theexperimental design of an artificialfin, actuated via remote actuation, asingle tensegrity beam was builtand placed within an elastomer fin.Figure 4a shows images of a singletensegrity beam as it is actuated. Thebeam enables leading-edge actuationof the artificial fin (Figure 4b). This

fin was then tested in a flow tank toobserve the influence of frequency onthe wake topology. Figures 4c and 4dshow the actuating fin in water fromthe side and below. The black linessuperimposed on these images repre-sent the kinematical model for the ki-nematics of a cownose ray—note theexcellent agreement between experi-ment and theory, especially with thepassive response of the elastomer fin.This approach costs minimal powerconsumption and shows the simple de-sign of a high-performance tensegrity-based artificial pectoral fin.

Concluding Commentsand Future Directions

The idea to look to nature for inspi-ration is not new, and this rich arena

FIGURE 3

(a) Three-dimensional tensegrity structures (three, four, and six strut prismatic structures). (b) Tensegrity-based fin concept. The fin can deformwith coupled curved spanwise motion and chordwise undulation to mimic the kinematics of the manta ray and the batoid family in general. Thetensegrity deforms when active elements contract or expand.


continues to be a source for engineersto aid in solving challenging problems.From Leonardo Da Vinci’s flying ma-chine, Helical Air Screw, to leadingedge whale-like tubercles on wind tur-bine blades to improve efficiencies(Fish et al., 2011). The opportunitiesto learn from nature and emulate itsunique approach to overcoming chal-lenges seem endless. In this paper, wehave touched upon the challenge ofcreating efficient, economic, and ma-neuverable underwater swimmingplatforms. We have focused on batoidfishes for inspiration in the design ofthe next generation of bioinspired un-derwater vehicles, as emulation of itsswimming characteristics—efficiency,maneuverability, stealth, workingenvironments, scalability—has thepotential to significantly improveupon current state-of-the-art in AUVtechnologies.

The development of a batoid-inspired underwater vehicle can beclassified in terms of the approachtaken to achieve ray-like swimming.The first is developing a batoid-inspired AUV that performs as a plat-form to test the capabilities of a varietyof traditional and novel actuating tech-nologies. The motivation here is todemonstrate the capabilities of suchdevices—usually in terms of forceand stroke—and is not necessarily adesire to truly replicate the biologi-cal system. These actuators includeelectroactive polymers, fluidic mus-cles, shape memory alloys, motorsand servomotors. Of particular interestin testing these actuators has been thechallenge of quantifying the swim-ming performance of the particularvehicle, many of which do not neces-sarily mimic the kinematics of batoidfishes. As biology has limitations due

to the materials available to constructa body and the evolutionary processthat produces an organism, possibleimprovements to the basic body plancan perhaps be engineered to enhanceperformance beyond the capabilities ofnature.

The second approach considers fun-damental questions associated withbiology’s solutions to propulsion,maneuverability, stability, and stealth.Technology is used in this case to rep-licate the biology to help answer thesequestions. Using the underlying bi-ology as the basis for inquiry, themechanisms that dictate batoid swim-ming performance are explored. This isbeing done through the design and de-velopment of artificial systems—eitherreal or virtual—that can achieve nearidentical kinematic motions of the raysbeing studied. Specifically, researchersare working to elucidate the dominantmechanisms in batoid swimming thatdictate efficiency and maneuverability.A key outcome of this work is to fullyexplore nature’s design space andbeyond. By mimicking biology, weattempt to elucidate the key featuresthat control and optimize function.Nature evolves solutions that satisfymultiple constraints; engineers andscientists can design for a single desiredoutcome. By identifying and quantify-ing the key features/characteristics thatdictate optimality, we can judiciouslychoose to build these into an artificialsystem, depending on the requiredfunctionality of the device. For exam-ple, one may desire a vehicle thatcan swim for as long as possible oras fast as possible—two different solu-tions may be necessary for these tworequirements.

As mentioned in UnderstandingBiological Foundation, there hasbeen an extensive study of the under-lying cartilage structure of various

FIGURE 4

Example of a tensegrity-based actuating fin. (a) Photos of a tensegrity beams employing remoteactuation. (b) Dorsal view of elastomer fin with leading edge actuation via tensegrity beam.(c) Posterior view of actuating fin shown in (b). (d) Lateral view of actuating fin. Note the linesin (c) and (d) represent the mathematical model derived to describe kinematic motions of a mantaray pectoral fin.


batoid rays (Schaefer & Summers,2005). Biomechanical studies of thecartilage arrangement have been car-ried out to examine the relationshipbetween the form of the underlyingstructure and its impact on the func-tion (Russo et al., 2011). In thiswork, Russo et al. have taken thecartilage architecture and developeda numerical model to study the ki-nematic function of this form. Thisinitial study is beginning to explorethe relationship between form andfunction.

One of the most exciting develop-ments in the creation of these bio-inspired devices and vehicles is in thedevelopment of rapid prototyping fab-rication (Figure 5). This has openedthe possibility to build a cartilagestructure that uses the same designprinciples observed in nature, as de-scribed by Schaefer & Summers(2005). In this physical model, carti-

lage elements are connected in thespanwise direction with cross-bracingin the chordwise direction to mimicthe architecture of the Atlantic sting-ray (www.bartsmithlabs.com). Thistechnology enables the design to bequickly and easily varied so as to an-swer questions regarding the influ-ence of the architecture on kinematicperformance. The images in Figure 5are compelling, as they demonstratekinematic capabilities and possibili-ties of such a structure. By mimickingthe underlying structure of biology,we can explore the capabilities ofthese species and potentially expandupon them.

Significant progress has beenmade,but there is still much to be done. In-formation on the kinematics of swim-ming is being generated, but there isstill much to be learned, especiallywith respect to some of the more finemotor skills observed. Also, not much

is known about the hydroacousticproperties of the biology. Materialproperties of the constituent parts ofthe pectoral fin are needed to improvethe fidelity biomechanical models thatdescribe form and function. Lastly,more investigation of the sensing andcontrol strategies of batoids is needed.This improved understanding willprovide valuable insight when moresophisticated vehicles are developed.With regard to engineering a batoid-inspired AUV, there are huge opportuni-ties in actuator design and development.Structural and material design and se-lection also are areas that need to beaddressed. For example, how do we de-sign a skin that can accommodate boththe out-of-place hydrodynamic forcesand the potential in-plane stretchingexperienced during actuation. Howdo the properties of the artificial sys-tem scale with the biological proper-ties? Actuation technology is an areathat has the potential to revolutionizethe field of biomimicry and bio-inspired engineering.

In this paper, we have focused ona small subset of bio-inspired under-water vehicles. We have presented areview of work related to the develop-ment of a bio-inspired underwaterrobot—actuation technology inte-grated into a batoid-like vehicle andunderstanding the biological founda-tion to explore the full design space.It is clear though that these two cate-gories are very closely related. Withoutactuator development, it may not bepossible to achieve anything close towhat biology achieves. But a clear pic-ture of biology function is neededso that actuator requirements can bequantified. Synergy between biology,biomimicry, and bio-inspired en-gineering is essential if we want todevelop the next generation of under-water vehicles.

FIGURE 5

Example of artificial cartilage structure design using rapid-prototyping technology. The individualelements represent the cartilage elements found in the pectoral fins of batoid rays. The arrange-ment and connectivity are similar to portions found in the Atlantic stingray.


AcknowledgmentsThe authors would like to acknowl-

edge funding from the Office of NavalResearch through the MURI programon Biologically Inspired AutonomousSea Vehicles (Program Manager: Dr.R. Brizzolara, Contract No. N00014-08-1-0642) and the David and LucillePackard Foundation.

ReferencesAbbott, I.H., & von Doenhoff, A.E. 1949.

Theory of Wing Sections: Including a

Summary of Airfoil Data. New York:

McGraw-Hill.

Ayers, J., Wilbur, C., & Olcott, C. 2000.

Lamprey robots. In: Proceedings of 1st Inter-

national Symposium on Aqua Bio-mechanisms,

eds. Wu, T., & Kato, N. Honolulu, Hawaii:

Tokai University Pacific Center.

Anderson, J.M., & Chhabra, N.K. 2002.

Maneuvering and stability performance of

a robotic tuna. Integr Comp Biol. 42(1):

118-26. doi: 10.1093/42.1.118.

Bandyopadhyay, P.R. 2005. Trends in bio-

robotic autonomous undersea vehicles.


doi: 10.1109/JOE.2005.843748.

Barrett, D., Grosenbaugh,M., &Triantafyllou,

M. 1996. The optimal control of a flexible

hull robotic undersea vehicle propelled by

an oscillating foil. In: Proceedings of the Sym-

posium on Autonomous Underwater Vehicle

Technology, AUV’96. 1-9. Monterey, CA:

IEEE. doi: 10.1109/AUV.1996.532833.

Blidberg, D.R. 2001. The development of

autonomous underwater vehicles (AUVs):

A brief summary. In: IEEE ICRA Conference,

Seoul Korea, May 21-25, 2001.

Borgen, M.G., Washington, G.N., & Kinzel,

G.L. 2003. Design and evolution of a piezo-

electrically actuated miniature swimming

vehicle. IEEE-ASME T Mech. 8(1):66-76.

doi: 10.1109/TMECH.2003.809131.

Breder, C.M., Jr. 1926. The locomotion of

fishes. Zoologica. 4:159-297.

Brower, T.P.L. 2006. Design of a manta

ray inspired underwater propulsive mechanism

for long range, low power operation. Master’s

thesis, Tufts University.

Cai, Y., Bi, S., & Zheng, L. 2010. Design

and experiments of a robotic fish imitating

cow-nosed ray. J Bionic Eng. 7(2):120-6.

doi: 10.1016/S1672-6529(09)60204-3.

Chen, Z., Um, T.I., & Bart-Smith, H. 2011.

A novel fabrication of ionic polymer-metal

composite membrane actuator capable of

3-dimensional kinematic motions. Sensor

Actuat A-Phys. 168(1):131-9. doi: 10.1016/

j.sna.2011.02.034.

Clark, R.P., & Smits, A.J. 2006. Thrust

production and wake structure of a batoid-

inspired oscillating fin. J Fluid Mech. 562:

415-29. doi: 10.1017/S0022112006001297.

Colgate, J.E., & Lynch, K.M. 2004.

Mechanics and control of swimming: A

review. IEEE J Oceanic Eng. 29(3):660-73.

Combes, S.A., & Daniel, T.L. 2001. Shape,

apping and exion: wing and fin design for

forward flight. J Exp Biol. 204(12):2073-85.

Crespi, A., Badertscher, A., Guignard, A., &

Ijspeert, A.J. 2004. An amphibious robot

capable of snake and lamprey-like locomotion.

In: Proceedings of the 35th International

Symposium on Robotics ISR 2004. Paris,

France: ISR publisher.


& MacIver, M.A. 2010. Aquatic manoeuver-

ing with counter-propagating waves: a novel


8(60):1041-50.

Dennler, G., Scharber, M., Ameri, T.,

Denk, P., Forberich, K., Waldauf, C., &

Brabec, C. 2008. Design rules for donors

in bulk-heterojunction tandem solar cells?

Towards 15% Energy-Conversion Eficiency.

Adv Mater. 20(3):579-83. doi: 10.1002/

adma.200702337.

Dobbins, P. 2007. Dolphin sonar: Modelling

a new receiver concept. Bioinspir Biomim. 2:19.

doi: 10.1088/1748-3182/2/1/003.

Drucker, E.G., & Jensen, J.S. 1997. Kine-

matic and electromyographic analysis of steady

pectoral fin swimming in the surfperches.

J Exp Biol. 200:1709-23.

Festo, M. 2008. Fluidic Muscle Brochure.

http://www/festo.com/cms/en-us_us/5030.htm.

Fish, F.E., & Lauder, G.V. 2006. Passive and

active ow control by swimming fishes and

mammals. Annu Rev FluidMech. 38:193-224.

doi: 10.1146/annurev.fluid.38.050304.092201.

Fish, F.E., Lauder, G.V., Mittal, R., Techet,

A.H., Triantafyllou, M.S., Walker, J.A., &

Webb, P.W. 2003. Conceptual design for the

construction of a biorobotic AUV based on

biological hydrodynamics. In: Proceedings of

the 13th International Symposium on Un-

manned Untethered Submersible Technology.

Durham, NH: AUSI (Autonomous Undersea

Systems Institute).


Howle, L.E. 2011. The humpback whale’s

ipper: Application of bio-inspired tubercle

technology. Integr Comp Biol. 211:1857-67.

Gao, J., Bi, S., Xu, Y., & Liu, C. 2007.

Development and design of a robotic manta

ray featuring flexible pectoral fins. In: IEEE

International Conference on Robotics and

Biomimetics, December 15-18, 2007. ROBIO

2007, 519-23. Yalong Bay of Sanya, China:

IEEE.

Gibb, A., Jayne, B., & Lauder, G. 1994.

Kinematics of pectoral fin locomotion in the

bluegill sunfish Lepomis machrochirus.

J Exp Biol. 189:133-61.

Heine, C. 1992. Mechanics of apping fin

locomotion in the cownose ray, Rhinoptera

bonasus (Elasmobranchii, Myliobatidae).

Ph.D. thesis, Duke University.

Jaffe, J., Moore, K., McLean, J., & Strand,

M. 2001. Underwater optical imaging: status

and prospects. Oceanography. 14(3):66-76.

Jalbert, J., Baker, J., Duchesney, J., Pietryka,

P., Dalton, W., Blidberg, D., … Holappa, K.

2003. Solarpowered autonomous underwater

vehicle development. In: Proceedings of the

Thirteenth International Symposium on Un-

manned Untethered Submersible Technology.

Durham, NH: Autonomous Undersea Sys-

tems Institute.


Kato, N. 1998. Locomotion by mechanical

pectoral fins. J Mar Sci Technol. 3(3):113-21.

doi: 10.1007/BF02492918.

Kato, N., Ando, Y., Tomokazu, A., Suzuki,

H., Suzumori, K., Kanda, T., & Endo, S.

2008. Elastic pectoral fin actuators for biomi-

metic underwater vehicles. In: Bio-mechanisms

of Swimming and Flying, eds. Kato, N., &

Kamimura, S., Part 3, 271-82. Springer.

doi: 10.1007/978-4-431-73380-5_22.

Kato, N., & Furushima, M. 2002. Pectoral

fin model for maneuver of underwater

vehicles. In: Autonomous Underwater

Vehicle Technology, 1996. AUV’96,

Proceedings of the 1996 Symposium on

IEEE. pp. 49-56. Durham, NH: Autonomous

Undersea Systems Institute.

Klausewitz, W. 1964. Der lokomotions-

modus der Flugelrochen (Myliobatoidei).

Zool Anz. 173:111-20.

Kocak, D., & Caimi, F. 2005. The current

art of underwater imaging with a glimpse

of the past and vision of the future. Mar

Technol Soc J. 39(3):5-26. doi: 10.4031/

002533205787442576.

Lauder, G.V., & Jayne, B.C. 1996. Pectoral

fin locomotion in fishes: Testing drag-based

models using three-dimensional kinematics.

Am Zool. 36(6):567-81.

Lauder, G.V., Nauen, J.C., & Drucker, E.G.

2002. Experimental hydrodynamics and evo-

lution: function of median fins in ray-finned

fishes 1. Integr Comp Biol. 42(5):1009-17.

doi: 10.1093/icb/42.5.1009.

Licht, S., Polidoro, V., Flores, M., Hover, F.S.,

& Triantafyllou, M.S. 2004. Design and

projected performance of a apping foil AUV.


doi: 10.1109/JOE.2004.833126.

Listak, M., Martin, G., Pugal, D., Aabloo, A.,

& Kruusmaa, M. 2005. Design of a semi-

autonomous biomimetic underwater vehicle for

environmental monitoring. In: Computational

Intelligence in Robotics and Automation.

Espoo, Finland: IEEE.

Low, K.H., & Willy, A. 2006. Biomimetic

motion planning of an undulating robotic

fish fin. J Vib Control. 12:1337-59.

doi: 10.1177/1077546306070597.

Mojarrad, M. 2000. AUV biomimetic

propulsion. In: OCEANS 2000 MTS/IEEE

Conference and Exhibition. pp. 2141-6.

Providence, RI: IEEE.

Moored, K.W. August 2010. The design of a

novel tensegrity-based synthetic pectoral fin

for bio-inspired propulsion. Ph.D. thesis,

University of Virginia.

Moored, K.W., & Bart-Smith, H. 2007. The

analysis of tensegrity structures for the design

of a morphing wing. J Appl Mech. 74:668-76.

doi: 10.1115/1.2424718.

Moored, K.W., & Bart-Smith, H. 2009. In-

vestigation of clustered actuation in tensegrity

structures. Int J Solids Struct. 46:3272-81.

doi: 10.1016/j.ijsolstr.2009.04.026.

Moored, K.W., Dewey, P.A., Leftwich, M.C.,

Bart-Smith, H., & Smits, A.J. 2011a.

Bio-inspired propulsion mechanisms based

on manta ray locomotion. Mar Technol Soc J.

45(4):110-8.

Moored, K.W., Kemp, T.H., Houle, N.E., &

Bart-Smith, H. 2011b. Analytical predictions,

optimization and design of a tensegrity-based

artificial pectoral fin. Int J Solids Struct.

(Accepted for Publication).

Palmisano, J., Geder, J., Ramamurti, R.,

Liu, K.J., Cohen, J., Mengesha, T., … Ratna,

B. 2008. Design, development, and testing

of flapping fins with actively controlled cur-

vature for an unmanned underwater vehicle.

In: Bio-mechanisms of Swimming and Flying:

Fluid Dynamics, Biomimetic Robots, and

Sports Science, eds. Kato, N., & Kamimura,

S., 283-94. Berlin: Springer.

Palmisano, J., Ramamurti, R., Lu, K.J.,

Cohen, J., Sandberg, W., & Ratna, B. 2007.

Design of a biomimetic controlled-curvature

robotic pectoral fin. In: IEEE International

Conference on Robotics and Automation

(ICRA). 966-73. IEEE.

Pederzani, J.-N., Moored, K.W., Fish, F.E.,

& Haj-Hariri, H. 2011. A numerical inves-

tigation of the hydrodynamic signature of

batoid swimming. In: Society for Integrative

and Comparative Biology, Annual meeting.

Salt Lake City, UT: Oxford Univ Press Inc.



oscillation. J Exp Biol. 204(2):379-94.

Russo, R.S., Blemker, S., Fish, F.E., Moored,

K.W., & Bart-Smith, H. 2011. Form function

relationship between ray skeletal architecture

and ray locomotion. In: Annual meeting of

the Society for Integrative and Comparative

Biology. Salt Lake City, UT: Oxford Univ

Press Inc.

Sasko, D.E., Dean, M.N., Motta, P.J., &

Hueter, R.E. 2006. Prey capture behavior

and kinematics of the Atlantic cownose ray,

Rhinoptera bonasus. Zoology. 109(3):171-81.

doi: 10.1016/j.zool.2005.12.005.

Schaefer, J.T., & Summers, A.P. 2005.

Batoid wing skeletal structure: Novel

morphologies, mechanical implications, and

phylogenetic patterns. J Morphol. 264(3):

298-313. doi: 10.1002/jmor.10331.

Sfakiotakis, M., Laue, D., & Davies, B.

2005. An experimental undulating-fin device

using the parallel bellows actuator. In:

Robotics and Automation, Proceedings 2001

ICRA. IEEE International Conference on.

pp. 2356-62. Seoul, Korea: IEEE.

Sfakiotakis, M., Lane, D.M., & Davies,

J.B.C. 1999. Review of fish swimming modes

for aquatic locomotion. IEEE J Oceanic Eng.

24(2):237-52. doi: 10.1109/48.757275.

Stocking, J.B., Eberhardt, W.C., Shakhsheer,

J.R., Paulus, M.A., & Calhoun, B.H. 2010.

A capacitance-based whisker-like artificial

sensor for fluid motion sensing. In: IEEE

Sensors. Waikoloa, Big Island, Hawaii.

Summers, A. 2000. Stifiening the stingray

skeleton: An investigation of durophagy in

myliobatid stingrays (Chondrichthyes, Batoidea,

Myliobatidae). J Morphol. 243(2):113-26.

doi: 10.1002/(SICI)1097-4687(200002)

243:2<113::AID-JMOR1>3.0.CO;2-A.

Suzumori, K., Endo, S., Kanda, T., Kato, N.,

& Suzuki, H. 2007. A bending pneumatic

rubber actuator realizing soft-bodied manta

swimming robot. In: 2007 IEEE International


Conference on Robotics and Automation.

pp. 4975-980. IEEE.

Takagi, K., Yamamura, M., Luo, Z., Onishi,

M., Hirano, S., Asaka, K., & Hayakawa, Y.

2007. Development of a rajiform swimming

robot using ionic polymer artificial muscles.

In: Intelligent Robots and Systems, 2006

IEEE/RSJ International Conference on.

pp. 1861-6. Beijing, China: IEEE.

Tangorra, J., Anquetil, P., Fofonoff, T., Chen,

A., Zio, M., & Hunter, I. 2007. The applica-

tion of conducting polymers to a biorobotic

fin propulsor. Bioinspir Biomim. 2:S6.

doi: 10.1088/1748-3182/2/2/S02.

Tangorra, J., Davidson, S., Hunter, I.,

Madden, P., Lauder, G., Haibo, D., …

Mittal, R. 2008a. The development of a

biologically inspired propulsor for unmanned

underwater vehicles. IEEE J Oceanic Eng.

32(3):533-50.

Tangorra, J., Lauder, G.,Madden, P.,Mittal, R.,

Bozkurttas, M., & Hunter, I. 2008b. A

biorobotic apping fin for propulsion and

maneuvering. In: Robotics and Automation,

2008. ICRA 2008. IEEE International

Conference on. pp. 700-5. Pasadena, CA:

IEEE.

V. y Alvarado, P., Chin, S., Larson, W.,

Mazumdar, A., & Youcef-Toumi, K. 2010. A

soft body under-actuated approach to multi

degree of freedom biomimetic robots: A

stingray example. In: Biomedical Robotics and

Biomechatronics (BioRob), 2010 3rd IEEE

RAS and EMBS International Conference

on. pp. 473-8. Tokyo, Japan: IEEE.


Labriform propulsion in fishes: kinematics

of flapping aquatic flight in the bird wrasse

Gomphosus varius (Labridae). J Exp Biol.

200:1549-69.

Wang, Z., Hang, G., Wang, Y., Li, J., &

Du, W. 2008. Embedded SMA wire actuated

biomimetic fin: a module for biomimetic

underwater propulsion. Smart Mater

Struct. 17:025039. doi: 10.1088/0964-

1726/17/2/025039.

Webb, D., & Simonetti, P. 1999. The

Slocum AUV: An environmentally propelled

underwater glider. In: International Sympo-

sium on Unmanned Untethered Submersible

Technology. 75-85. Durham, NH: Autono-

mous Undersea Systems Institute.

Webb, P.W. 1994. Mechanics and Physiology

of Animal Swimming: The Biology of Fish

Swimming. Cambridge, UK: Cambridge

University Press. 45-62 pp.

Yan, Q., Zhang, S., & Yang, J. 2010. Initial

implementation of basic actuated unit of a

flexible pectoral fin driven by SMA. In:

Mechatronics and Automation (ICMA),

2010 International Conference on.

pp. 899-904. Xi’an, China: IEEE.

Yang, S., Qiu, J., & Han, X. 2009. Kine-

matics modeling and experiments of pectoral

oscillation propulsion robotic fish. J Bionic

Eng. 6(2):174-9. doi: 10.1016/S1672-6529

(08)60114-6.

Yang, Y., Chen, J., Engel, J., Pandya, S.,

Chen, N., Tucker, C., … Liu, C. 2006.

Distant touch hydrodynamic imaging with

an artificial lateral line. P Natl Acad Sci.

103(50):18891.

Yong-hua, Z., Shi-wu, Z., Jie, Y., & Low, K.

2007. Morphologic optimal design of bionic

undulating fin based on computational fluid

dynamics. In: Mechatronics and Automation,

2007. ICMA 2007. International Conference

on. pp. 491-6. Harbin, Heilongjian, China:

IEEE.

Yu, J., Hu, Y., Fan, R., Wang, L., & Huo, J.

2007. Mechanical design and motion

control of a biomimetic robotic dolphin.

Adv Robotics. 21(3):499-513.

doi: 10.1163/156855307780131974.

Yu, J., Tan, M., Wang, S., & Chen, E. 2004.

Development of a biomimetic robotic fish

and its control algorithm. IEEE Sys Man

Cybern B. 34(4), 1798-810.

Zhou, C., & Low, K. 2010. Better endurance

and load capacity: An improved design of

manta ray robot. J Bionic Eng. 7:S137-44.

doi: 10.1016/S1672-6529(09)60227-4.


P A P E R

Bioinspired Propulsion MechanismsBased on Manta Ray LocomotionA U T H O R SKeith W. MooredPeter A. DeweyMechanical and AerospaceEngineering, Princeton University

Megan C. LeftwichPhysics Division, Los AlamosNational Laboratory

Hilary Bart-SmithMechanical and AerospaceEngineering, University of Virginia

Alexander J. SmitsMechanical and AerospaceEngineering, Princeton University

A B S T R A C TMobuliform swimmers are inspiring novel approaches to the design of underwater

vehicles. These swimmers, exemplified bymanta rays, present amodel for new classesof efficient, highly maneuverable, autonomous undersea vehicles. To improve ourunderstanding of the unsteady propulsion mechanisms used by these swimmers,we report detailed studies of the performance of robotic swimmers that mimic aspectsof the animal propulsive mechanisms. We highlight the importance of the undulatoryaspect of producing efficientmanta ray propulsion and show that there is a strong inter-action between the propulsive performance and the flexibility of the actuating surfaces.Keywords: mobuliform, manta ray, unsteady, swimming, flexible actuators

Introduction

I n recent years, there has beenconsiderable interest in developingnovel underwater vehicles that usepropulsion systems inspired by bi-ology (Colgate & Lynch, 2004;Bandyopadhyay, 2005). Such vehicleshave the potential to open up new mis-sion capabilities and improve maneu-verability, efficiency, and speed (Fishet al., 2003, 2011). Here we explorehow various aspects of biological loco-motion relate to performance in theparticular case of ray-like swimming,with the aim of informing the designof new vehicles.

The kinematic motion of batoid fish(rays) is based on the chordwise travelingwave that is a hallmark of their motion(Rosenberger, 2001). Species are classi-fied as being oscillatory if the travelingwave wavelength is longer than thechord of their fin and undulatory if thewavelength is less than the chord of

their fin. The manta ray is an exampleof an oscillatory swimmer. Previously,Clark and Smits (2006) explored thethrust production and efficiencies of anartificial pectoral fin that captured thetraveling wave motion and observedefficiencies upwards of 50% for an oscil-latory motion at a fixed flow velocity.

To study the swimming of mantas,we use artificial or robotic devices thatgenerate a simple baseline motion thatapproximates biological kinematics.The complexity of the motion is thenprogressively increased by addingmorekinematic features until the motionresembles the biology very closely. Ateach level of complexity, various perfor-mancemetrics aremeasured.We explorethe role of spanwise curvature, the effectsof a spanwise traveling wave and tipspeed modulation, which have not beenpreviously investigated, and, the roleof a chordwise traveling wave motionis investigated in the performance ofan actively and passively flexible fin.

ExperimentsTwo biorobotic devices were devel-

oped and tested. First, an artificial pec-

toral fin able to produce root-fixedpure heaving motions was developedand will be referred to as the heavingfin. The fin was cast from a flexibleplastic and actuated with variable de-grees of spanwise curvature. The mea-surements on the heaving fin wereconducted in a tow tank (Figure 1).This facility tows a fin through stillwater at a fixed velocity, U, in a tankmeasuring 5-m long, 1-m deep, and1.5-m wide, and we directly measurethe net force produced, T, and thepower imparted into the fluid, Pf ,by a flapping fin structure. Thepropulsive efficiency of the motion,ηp ¼ ‐TU=

‐Pf , can then be calculatedfrom the thrust and power mea-surements averaged over a cycle,‐T and

‐Pf , respectively. If the fin wereunconstrained and free-swimming,then the net force would cause thefin to accelerate or decelerate to anew velocity where there is no averagenet force. Constraining the fin allowsfor the measurement of force produc-tion and is a commonly used experi-mental approach (Anderson et al.,1998).


Second, an artificial pectoral fin ca-pable of generating a chordwise travel-ing wave motion was developed andwill be referred to as the travelingwave fin. This fin was used to measurethe free-swimming performance in atank measuring 6.7-m long, 1-mdeep, and 1-m wide. This fin wasalso cast from a flexible plastic, but itwas activated using a number of rigidspars in the spanwise direction. By re-ducing the number of actuating spars,the degree of passive flexibility of thefin could be varied. The travelingwave wavelength was controlled bychanging the phase differences be-tween adjacent spars. The steadyswimming speed U and mean powerinput over a cycle

‐Pf were measured,and hence, the energy economy ζcould be found, where ζ ¼ U=

‐Pf .Energy economy is the inverse of

cost of transport, both of which areused extensively in the biological liter-ature (Schmidt-Nielsen, 1972; Fishet al., 1991; Liao et al., 2003; Liao,2004). Energy economy, however, isa more appropriate engineering metricas the dimensions are distance/per unitenergy (the units could be miles pergallon for instance). Efficiency is alsoan appropriate performance measure,but from an experimental point-of-

view it can only be measured whenthere is net thrust production or thefin is not in a free-swimming mode.Thus, for the experiments in the towtank, efficiency was a measurableperformance metric, but in the free-swimming cases, economy is used in-stead because the efficiency could notbe directly measured.

The performance is measured as afunction of the Strouhal number,St = fA/U, where f is the frequency ofmotion, A is the peak-to-peak trailing-edge amplitude of motion at the mid-span, andU is the free-stream velocity.The Strouhal number is a measureof the lateral to streamwise spacing ofthe shed vortices in the wake and to alarge extent governs the structure ofthe wake. It has been shown to be acritical parameter in describing the ef-ficient propulsion of oscillating foilsand plates (Anderson et al., 1998;Buchholz & Smits, 2008) and forswimming (Clark & Smits, 2006;Borazjani & Sotiropoulos, 2008,2009) and flying animals (Tayloret al., 2003).

Fixed Velocity Experiments:Heaving Motion

The skeletal structure of the heav-ing fin is composed of three connected

hinged plates. The angular position ofeach hinge is individually controlled bya linear actuator. The fin allows forout-of-plane motion with no pitchingor undulation in the chordwise direc-tion. The skeletal structure is embed-ded into a compliant PVC polymer.The PVC is molded around the struc-ture into a fin with a trapezoidal plan-form shape. This shape was chosen tobe a simple representative shape of themanta ray as well as to maximize spac-ing for the internal structure. The finhas a span length of b = 28 cm andan average chord length of ―c = 19cm, with an aspect ratio, AR = b2/S,of 1.47, wherein S is the planformarea. The cross-sectional shape is aNACA 0020 airfoil. The trailing edgeis stiffened by a thin metal sheet at-tached to the main structure.

Three flapping mode shapes wereexplored: flat root-fixed heave (Fig-ure 2(a)), curved root-fixed heave (Fig-ure 2(b)), and curved root-fixed heavewith a span-wise traveling wave (Fig-ure 2(c)). For each of these modeshapes the tip speed of the fin can bemodulated. Previous kinematics stud-ies of ray locomotion (Heine, 1992;Rosenberger, 2001) found that inorder to swim faster oscillatory raysdo not vary their beat frequency, butinstead they vary the tip speed oftheir fin while holding frequency andamplitude constant. This can beviewed as modifying the actuationwaveform to suit a particular mode ofswimming. To implement this modein our experiments, the time-varyingwaveform was varied from a pure sinu-soid towards an almost square waveform (Figure 3). This allows the fre-quency and amplitude to be heldfixed while the maximum fin tipspeed is increased. The thrust and pro-pulsive efficiency were measured foreach prescribed motion.

FIGURE 1

Tow tank facility consisting of an artificial pectoral fin, tow tank, motor, carriage, and a controland measurement system.


Free-Swimming Experiments:Traveling Wave Motion

The traveling wave fin was testedunder free-swimming conditions in astationary tank to explore the roleof flexibility in ray-like propulsion.Low-friction carts were attached tothe fin actuation mechanism to forma carriage that was mounted on tracksabove the tank (see Figure 4). Thechord of the fin was aligned parallelto the tracks, which allowed for a sin-gle degree of freedom and made theswimming direction of the carriageunidimensional, and no transversemotion of the carriage was permitted.The root of the fin abutted against anacrylic sheet that was in contact withthe free surface of the water to mini-mize surface waves. Upon actuatingthe fin, the cart propelled itself downthe length of the tank.

An elliptical planform fin with anaspect ratio of 1.6 and a NACA 0020cross-section was cast using a flexiblePVC plastic. Four aluminum sparswere embedded into the fin to provideactuation. A push-rod connected eachspar to a gear in a gear-train, driven bya DC motor, that produced a sinusoi-dal rotation of the spar about a pivotpoint located at the root chord of thefin. This results in a linearly increasingamplitude of motion along the span ofthe fin (Figure 5). The wavelength ofthe traveling wave λ could be variedby changing the phase difference be-tween the actuating spars.

In addition, the number of actuat-ing spars could be varied to allow for acertain degree of passive flexibility inthe fin. When four actuating sparsare used, the locomotion of the finwas prescribed for the entire chord of

the fin, with a traveling wave wave-length λa, and this fin will be referredto as the active fin.When the two trail-ing edge spars are removed, the trailinghalf of the chord of the fin will pas-sively respond to the leading edge actu-ation and external fluid forces. This finwill be referred to as the passive fin(Figure 5). For the passive fin, thetraveling wave along the chord is gen-erated by the first two gears in the geartrain and the passive response of thetrailing edge. Due to these compound-ing factors, the precise wavelengthalong the chord for the passive fin isunknown, and we define an analo-gous wavelength, λp such that λa =λp when the offset between the firsttwo gears is identical for both the ac-tive and passive fin. We define the di-mensionless wavelengths λ

a ¼ λa=Cand λ

p ¼ λp=C , where C is the rootchord of the fin. This study focuses oncases with λ

a ;λp > 1, representative

of oscillatory swimmers (Rosenberger,2001).

Results and DiscussionFlat and Curved Modes

The first two modes of swim-ming, the flat mode (Figure 2(a)) andthe curved mode (Figure 2(b)), weres tud ied us ing the heav ing fin.The thrust coefficient is defined byCT ¼ T = 1

2 ρU2S, wherein T is the

thrust, ρ is the water density, and Sis the planform area of the fin. Simi-larly, the power coefficient is definedby Cp ¼ P= 1

2 ρU3S, wherein P is the

power input to the water (as definedby Clark & Smits, 2006). Figure 6(a)shows that the thrust production andpower input increases as the Strouhalnumber increases, as found in previousstudies by Anderson et al. (1998) andDong et al. (2006). The flat mode ofswimming produces more thrust and

FIGURE 3

Holding frequency and amplitude constant while modulating tip speed can be achieved by vary-ing the actuation waveform from a sine wave to a square wave.

FIGURE 2

Swimming modes ranging from an artificial/simple motion that approximates manta ray locomo-tion to a more complex biologically inspired motion: (a) flat root-fixed heaving motion, (b) curvedroot-fixed heaving motion, and (c) curved root-fixed heaving motion with a span-wise travelingwave (tip lag effect).


uses more power than the curvedmodethroughout the Strouhal range. Thisresults is not unexpected as the flatmode of swimming sweeps out a largervolume of fluid than the curvedmode of swimming, causing the over-all increase in the thrust and powercoefficients.

Because the two modes of swim-ming are dissimilar, comparing thethrust or efficiency as a function ofStrouhal number may be misleading.For a fixed Strouhal number, differentamounts of thrust and power are pro-duced by each swimming mode. Abetter comparison is the thrust or ef-ficiency as a function of the powercoefficient because each swimmingmode can be compared at a fixedpower input.

Figure 7a shows this comparison.For both modes of swimming thepower input increases as the thrust in-creases, although at higher powerinput the thrust increases at a slowerrate. For all power inputs, more thrustis produced for the flat mode of swim-ming than for the curved mode ofswimming, while the thrust tends tothe drag of the motionless fin as thepower input tends to zero. Interest-ingly, observations of swimmingmanta rays do not indicate that theyuse this higher-performance flatmode and instead use the curvedmode for swimming. This suggeststhat the curved mode, coupled withanother mechanism (perhaps a chord-wise traveling wave), more fully char-acterizes the swimming mechanics ofmanta rays.

Figure 7(b) shows the propulsiveefficiency as a function of Strouhalnumber. The efficiency of the curvedmode first rises quickly rise withincreasing Strouhal number, and thena peak in efficiency is attained, fol-lowed by a slow decline in efficiency

FIGURE 4

Tank facility for the chordwise traveling wave experiments: (a) perspective view and (b) frontview. Drawings not to scale.

FIGURE 5

Traveling wave actuation system for (left) active fin and (right) passive fin.

FIGURE 6

Flat mode compared to curved mode: (a) thrust performance as a function of St and (b) powercoefficient as a function of St.


at the higher Strouhal numbers. Thistrend is characteristic of efficiencycurves for oscillating foils and plates(Clark & Smits, 2006; Andersonet al., 1998; Heathcote et al., 2006a,2006b; Heathcote & Gursul, 2007;Buchholz & Smits, 2008). In general,the efficiency crosses from negative(net drag) to positive (net thrust) andcontinues to increase while at highvalues of St there is a decline in effi-ciency that follows potential flow the-ory ( Jones et al., 1998). Thus, a peakin efficiency is expected at an interme-diate Strouhal number. The peak effi-ciency for the curved mode is about20% occurring at a St = 0.2, which isin the range of 0.2 < St < 0.4 wheremost swimming and flying speciescruise (Taylor et al., 2003). Further-more, unpublished work on mantarays swimming at different speedsshow that a ray swimming at approxi-mately 2 m/s has a Strouhal number ofabout 0.21 (Fish, 2010), which is ingood agreement with our observedvalue of 0.2 for the peak efficiency ofthe curved mode. The peak in effi-ciency for the flat mode is not capturedin our experiments, but the highestvalue seen is about 22%.

Tip Lag ModeWe now study the effects of tip lag

for the heaving fin undergoing thecurved mode of swimming. Tip laghas been observed as an impor-tant feature of manta ray kinematics(Klausewitz, 1964), and here we com-pare three tip lags: 0, 3.2%, and11.3%, where the degree of tip lag isthe tip deflection normalized by thespan as the fin root section crossesthe neutral plane (Figure 8). With notip lag, the entire span crosses the neu-tral plane at the same time.

The results shown in Figure 9(a)indicate that tip lag increases the thrustand the power coefficients, and thepeak in efficiency broadens somewhatfor 11.3% tip lag. However, the peak

efficiency appears to decrease slightlywith increasing tip lag (Figure 9(b)), al-though the trend is barely outside theuncertainty limits. Perhaps manta raysutilize this kinematic mode to gain aslight thrust increase with no loss ofefficiency. However, the artificial finshape did not capture the sweep thatis exhibit by manta rays, as it wasthought to be unimportant for thrustproduction. The sweep could directlyaffect the streamwise velocity fieldwhen tip lag is present and thus moredirectly affect the thrust and efficiency.In fact, cownose and bullnose rays donot have significant sweep and do notdisplay any significant tip lag (Heine,1992; Rosenberger, 2001). Further-more, if a chordwise undulation werepresent than tip lag would have agreater impact on the streamwisevelocity field even when the planformhas no sweep.

Tip Speed ModulationAs indicated earlier, myliobatoid

rays have been shown to regulate thetip speed of their fins (while holdingfrequency constant) to increase theirswimming speed. To explore this ex-perimentally, the actuation waveformwas modified from a sine wave of agiven frequency and amplitude towardsa square wave with the same frequencyand amplitude, but with an increasedmaximum tip speed (Figure 3). A quartic

FIGURE 7

Flat mode compared to curved mode: (a) thrust coefficient as a function of the power coefficientand (b) efficiency as a function of Strouhal number.

FIGURE 8

Diagram showing tip lag effect. The magnitude of tip lag is measured as the tip deflection nor-malized by the span as the fin root section crosses the neutral plane.


function was used to modulate the tip speed while fixing the frequency andamplitude, as given by

x tð Þ ¼ a t ϕð Þ4þb t ϕð Þ2þA; 0 ≤ t ≤ T =2

a ¼ Usqrmax

2ϕ3þ Aϕ4

; b ¼ 2Aϕ2

þ Usqrmax

2ϕ

ϕ ¼ 1=4f ;U sqrmax ¼ 2παU∞St

The amplitude, A, the frequency, f, and the maximum tip speed, U sqrmax , all de-

termine the shape of the quartic function. The period is T. By holding the fre-quency and amplitude constant the tip speed can be modulated by varying αbetween 1 and 2.667. When α is 1, the quartic function matches a sine wave,but when α is 2.667, the maximum tip speed is 2.667 times faster than the max-imum tip speed of a sine wave without varying the frequency or amplitude. Itshould be noted, however, that many animals increase their swimming speedby modulating their frequency of motion while fixing the amplitude. Frequencymodulation will increase the Strouhal number by increasing frequency.

Tests were conducted at two Strouhal numbers, 0.2 and 0.25, and the ampli-tude was fixed at A/b = 0.44 for α =1-2.4. Figure 10a shows the thrust coefficientdependence on the tip speed, α, for a Strouhal number of 0.2. As the tip speed isincreased the thrust coefficient increases linearly, indicating that tip speed modu-lation can be used to increase thrust production, as exhibited by rays (Heine,1992; Rosenberger, 2001).

Figure 10(b) shows the thrust coefficient plotted against the power coefficientfor both frequency modulation and tip speed modulation. Tip speed modulationproduces less thrust than frequency modulation for the same input power, sug-gesting that tip speed modulation is not an efficiency strategy. However, incorpo-rating a chordwise traveling wave may change this outcome. Additionally, as thevalue of α is increased, the fin has an increasing period of effectively no motion. Ina free-swimming test, the increasing resting time for the fin would result in anunpowered gliding period over part of the flapping cycle. This would result in aburst-and-coast behavior that could improve the economy since forward motionwould occur without any input power for part of the flapping cycle. Alternatively,

tip speed modulation could an effec-tive flight/sprinting mode, where effi-ciency is not important.

Chordwise Traveling WaveWe now investigate the effects of a

chordwise traveling wave, using thetraveling wave fin (the experimentalarrangement was described in Free-Swimming Experiments: TravelingWave Motion).

Clark and Smits (2006) investi-gated the thrust and efficiency of an ar-tificial pectoral fin using a chordwisetraveling wave motion and found effi-ciencies peaking near 50% for optimalconditions (St ≈ 0.25 and λ

a ≈ 4‐6).The efficiencies were measured at pre-determined Strouhal numbers. For thecurrent work, this constraint is not im-posed, and the fin was instead actuatedat a given frequency and wavelengthand allowed to freely swim down thelength of a tow tank. In doing so, thefin attains its self-propelled swimmingspeed and Strouhal number for thatfrequency and wavelength.

Figure 11 shows the steady velocityachieved as a function of input flap-ping frequency for different wave-lengths of actuation. The velocity isgiven in root-chord lengths (CL) persecond, where Croot = 0.254 m. Forthe active fin, an almost linear increasein velocity is observed with increasingfrequency, a result in agreement withprevious studies that found thrustcoefficients increasing with frequency(Anderson et al., 1998). Peak velocitieswere found to occur for λ

a ¼ 6 at thehighest flapping frequencies. In theseinstances the velocity was upwards of2 CL/s, corresponding to a dimension-al velocity of 0.51 m/s, highlightingthat this form of propulsion mayprove fruitful for future underwater ve-hicle designs that demand relativelyhigh speeds.

FIGURE 9

Tip lag mode: (a) efficiency as a function of St and (b) peak efficiency dependence on tip lag.

(1)


Figure 11 also reveals the role ofpassive flexibility. Here, the fin was ac-tuated using only the two anteriorspars, leaving the posterior half of thefin to respond passively to the forcingby the actuators and the fluid forces.The velocity of this fin still increaseswith increasing frequency, but thetrend is no longer linear. Tests in stillwater indicate that the passive fin has aresonant frequency of about 2.4 Hz,defined as the frequency at which thetrailing edge amplitude is maximized.

As resonance is approached, the ampli-tude of the trailing edge motion in-creases, resulting in the fin velocityincreasing at an enhanced rate (in com-parison to a linear trend). It should benoted that the swimming velocities forthe passive fin were approximately80% of those of the active fin.

The free-swimming Strouhal num-ber for the active fin, along with mantaray field data (Fish, 2010), are dis-played in Figure 12 (the Strouhalnumber for the passive fin is notshown since the trailing edge excursionof the passive fin is unknown). The ex-

perimental data collapse onto a singlecurve, indicating that the Strouhalnumber for the freely swimming activefin does not depend on the wave-length. As the swimming velocity in-creases, the Strouhal number beginsto enter the regime presumed to be ef-ficient (St = 0.2-0.4; Taylor et al.,2003). Hence, the optimal swimmingspeed for the traveling wave fin, fromthe perspective of efficiency, is likelyto be ≥2 CL/s. The biological andexperimental data display the sametrend, whereby an increase in swim-ming velocity yields a decrease inStrouhal number. Borazjani andSotiropoulos (2008, 2009) were ableto show, for carangiform and anguili-form swimming, that the Strouhalnumber for self-propulsion approachesthe efficient regime observed in natureonly with increasing swimming veloc-ity. The current study supports thisconclusion for mobuliform swimmingas well.

The energy economy, ζ =Vc/Pf , forthe active and passive fins is shown inFigure 13. In the case of the shortestwavelength(λ* = 3), the active fin hasa higher energy economy than the pas-sive fin, but for the longer wavelengthsthe energy economy for the passive fin

FIGURE 10

(a) Thrust coefficient increasing with tip speed increase and (b) tip speed modulation comparedto frequency modulation. The parameter α is the ratio of maximum tip speed of the square waveactuation compared to the maximum tip speed of a sine wave of the same frequency and am-plitude, α ¼ U sqr

max=Usinmax .

FIGURE 11

Fin velocity in chord lengths per second as afunction of flapping frequency. The wavelengthλa refers to the active fin while λp refers to thepassive fin. (Color versions of figures availableonline at: http://www.ingentaconnect.com/content/mts/mtsj/2011/00000045/00000004.)

FIGURE 12

Strouhal number as a function of steady-stateswimming velocity for the active fin com-pared with biological data of the manta ray(Fish, 2010).

FIGURE 13

Energy economy as a function of flapping fre-quency. The wavelength λa refers to the ac-tive fin while λp refers to the passive fin.


exceeds that of the active fin despite adecrease in the overall swimming speed(Figure 11). Clearly, by removing thetwo trailing edge spars in creating thepassive fin, the power consumptiondecreases significantly compared tothe active fin. The highest energyeconomy recorded was 0.18 CL/Jfor the passive fin with λ

p ¼ 12 andf = 2 Hz, but its economy is still trend-ing upward with increasing frequency,which may reflect the fact that themaximum test frequency was stillbelow the resonant frequency of thefin (2.4 Hz). Leftwich and Smits(2010) found thrust production of apassively flexible artificial lamprey tailto increase as resonance is approached.The current work suggests that the en-ergy economy also benefits by a systemexploiting the resonant modes of a fin.

The increase in energy economywith frequency may also be a Strouhalnumber effect. Figure 12 indicates thatthe fin begins to enter the “efficient”regime (St = 0.2-0.4) with increasingswimming velocity (which occurs athigher flapping frequencies; see Fig-ure 11). While the energy economy isnot a direct measure of efficiency, thetwo parameters are inherently linked,so it is not altogether surprising thatthe energy economy increases as theStrouhal number approaches the sup-posedly optimal range. The active finwith λ

a ¼ 12 displays a maximum atf = 1.6 Hz. It is believed that thispeak is related to changes in the wakestructure that result from the three-dimensionality of the wake associatedwith this fin. Dewey et al. (2011)found that an increasing wavelengthcauses the wake to bifurcate into a dou-ble wake structure that is less efficient,and it is believed to be responsible forthe maximum observed in Figure 13.It may be that the other cases willeventually reach a maximum due to a

similar mechanism, but further testingis required to support this suggestion.

ConclusionsThis study has explored various ki-

nematic modes associated with biolog-ical propulsion based on themanta ray.The results highlight the interdepen-dence of the kinematic motions andfluid–structure interactions on the per-formance characteristics of the animal.

A purely heaving fin was used to ex-amine three kinematic modes of swim-ming (flat, curved, and tip lag) as wellas variation in the actuation waveform(sine waveform to square waveform). Itwas found that the flat mode of swim-ming produces higher efficiency andthrust compared to the curved modeof swimming. Furthermore, there wasno performance benefit found by in-corporating tip lag into the motion orby modulating tip speed instead of fre-quency. These results are counter tothe hypothesis that as the motion be-comes more biologically similar thethrust or efficiency performance willincrease. The maximum thrust co-efficient was found to be about 0.7 atSt = 0.45, and the maximum pro-pulsive efficiency was about 22% atSt = 0.15. These performance metricswere low, as expected, due to the ab-sence of a chordwise traveling wavemotion. What was not expected wasthat the performance would be in-sensitive to curvature, tip lag, and tipspeed variations. Without flexibility,the motion of the fin may be too con-strained and not “natural” enough toachieve the performance benefits ofthe kinematic variations explored.

The presence of a chordwise travel-ing wave to generate an undulatorymotion appears to be of prime impor-tance for efficient propulsion. For ex-ample, Clark and Smits (2006) found

efficiencies upwards of 50%. In free-swimming experiments of artificialfins similar to that studied by Clarkand Smits, we found that they wereable to generate speeds similar tothose observed in nature, of the order2 CL/s. When the fin was actively ac-tuated, that is to say that the travelingwave motion was defined for all pointson the chord of the fin, the steady-stateStrouhal number obtained by the os-cillating fin was found to be indepen-dent of the wavelength and exhibitedthe same trend as the manta ray innature. That is, at low swimmingvelocities, both the manta ray and theartificial fin display high steady-stateStrouhal numbers, but the Strouhalnumbers decrease with increasingswimming speed and they approachthe regime where efficient propulsionis hypothesized to exist (St = 0.2-0.4). Introducing passive flexibilityinto the fin, by restricting the actua-tion to the leading edge and lettingthe rest of the fin respond passively tothe actuation and the external fluidforces, improved the energy economy.It was found that the passively actuatedfin achieved steady state swimmingspeeds that were approximately 80%of that of the actively actuated fin,but because there was a significant de-crease in the power required to propelthe passively actuated fin the energyeconomy increased.

AcknowledgmentsThe authors would like to thank

Daphne Rein-Weston, Dan Quinn,and Dr. Melissa Green for their aidin developing the low-friction carriageexperiment. We would also like tothank Professor Frank Fish for cor-respondence regarding manta rays innature. The authors would like to ac-knowledge funding from the Office


of Naval Research through theMURI program on Biologically-Inspired Autonomous Sea Vehicles(grant N0001408-1-0642), theDavid and Lucille Packard Founda-tion, the National Science Foundation(grant CMS-0384884), and the Virgi-nia Space Grant Consortium.

Corresponding Author:Alexander J. SmitsMechanical and AerospaceEngineering, Princeton UniversityPrinceton, NJ 08544Email: [email protected]

ReferencesAnderson, J.M., Streitlien, K., Barrett, D.S.,

& Triantafyllou, M.S. 1998. Oscillating

foils of high propulsive efficiency.

J Fluid Mech. 360:41-72. doi: 10.1017/

S0022112097008392.

Bandyopadhyay, P.R. 2005. Trends in

biorobotic autonomous undersea vehicles.


doi: 10.1109/JOE.2005.843748.

Borazjani, I., & Sotiropoulos, F. 2008.

Numerical investigation of the hydrodynamics

of carangiform swimming in the transitional

and inertial flow regimes. J Exp Biol.

211:1541-58. doi: 10.1242/jeb.015644.

Borazjani, I., & Sotiropoulos, F. 2009.

Numerical investigation of the hydrodynamics

of anguilliform swimming in the transitional

and inertial flow regimes. J Exp Biol.

212:576-92. doi: 10.1242/jeb.025007.

Buchholz, J.H.J., & Smits, A.J. 2008.

The wake structure and thrust performance

of a rigid low-aspect-ratio pitching panel.

J Fluid Mech. 603:331-65. doi: 10.1017/

S0022112008000906.

Clark, R.P., & Smits, A.J. 2006. Thrust

production and wake structure of a

batoid-inspired oscillating fin. J Fluid

Mech. 562:415-29. doi: 10.1017/

S0022112006001297.

Colgate, J.E., & Lynch, K.M. 2004. Me-

chanics and control of swimming: A review.


doi: 10.1109/JOE.2004.833208.

Dewey, P.A., Carriou, A., & Smits, A.J.

2011. On the relationship between efficiency

and wake structures of a batoid-inspired

oscillating fin. J Fluid Mech. Under review.

Dong, H., Mittal, R., & Najjar, F.M. 2006.

Wake topology and hydrodynamic perfor-

mance of low-aspect-raio flapping foils.

J Fluid Mech. 566:309-43. doi: 10.1017/

S002211200600190X.

Fish, F.E. 2010. Private Communication.

Fish, F.E., Fegely, J.F., & Xanthopoulos,

C.J. 1991. Burst-and-coast swimming in

schooling fish (Notemigonus crysoleucas) with

implications for energy economy. Comp

Biochem Phys A Phys. 100:633-7.

doi: 10.1016/0300-9629(91)90382-M.

Fish, F.E., Haj-Hariri, H., Smits, A.J.,

Bart-Smith, H., & Iwasaki, T. 2011.

Biomimetic swimmer inspired by the

manta ray. In: Biomimetics: Nature-Based

Innovation, ed. Bar-Cohen, Y., 495-523.

Boca Raton, FL: CRC Press.

Fish, F.E., Lauder, G.V., Mittal, R., Techet,

A.H., Triantafyllou, M.S., Walker, J.A., &

Webb, P.W. 2003. Conceptual design for the

construction of a biorobotic AUV based on

biological hydrodynamics. In: Proceedings

of the 13th International Symposium on

Unmanned Untethered Submersible

Technology. Durham, NH: Autonomous

Systems Undersea Institute.

Heathcote, S., & Gursul, I. 2007. Jet

switching phenomenon for a periodically

plunging airfoil. Phys Fluids. 19(2):027104.

doi: 10.1063/1.2565347.

Heathcote, S., Wang, Z., & Gursul, I. 2006a.

Effect of spanwise flexibility on flapping wing

propulsion. In: 36th AIAA Fluid Dynamics

Conference and Exhibit. Reston, VA: American

Institute of Aeronautics and Astronautics Inc.

Heathcote, S., Wang, Z., & Gursul, I.

2006b. Flexible flapping airfoil propulsion

at low Reynolds numbers. AIAA J. 45(5):

1066-79. doi: 10.2514/1.25431.

Heine, C. 1992. Mechanics of flapping fin

locomotion in the cownose ray, Rhinoptera

bonasus (Elasmobranchii, Myliobatidae).

Ph.D. Thesis Duke University.

Jones, K.D., Dohring, C.M., & Platzer,

M.F. 1998. Experimental and computational

investigation of the KnollerBetz effect.

Zool Anz. 173:1240-6.

Klausewitz, W. 1964. Der lokomotionsmodus

der Flugelrochen (Myliobatoidei). AIAA J.

37:1240-6.

Leftwich, M.C., & Smits, A.J. 2010. Thrust

production by a mechanical swimming

lamprey. Exp Fluids. 50(5):1349-55.

doi 10.1007/s00348-010-0994-x.

Liao, J.C. 2004. Neuromuscular control

of trout swimming in a vortex street: Impli-

cations for energy economy during the

Karman gait. J Exp Biol. 207:3495-506.

doi: 10.1242/jeb.01125.

Liao, J.C., Beal, D.N., Lauder, G.V., &

Triantafyllou, M.S. 2003. Fish exploiting

vortices decrease muscle activity. Science.

302:1566-9. doi: 10.1126/science.1088295.



oscillation. J Exp Biol. 204(2):379-94.

Schmidt-Nielsen, K. 1972. Locomotion:

Energy cost of swimming, flying, and running.

Science. 177:222-8. doi: 10.1126/science.

177.4045.222.

Taylor, G.K., Nudds, R.L., & Thomas, A.L.R.

2003. Flying and swimming animals cruise

at a Strouhal number tuned for high power

efficiency. Nature. 425(6959):707-11.

doi: 10.1038/nature02000.


P A P E R

Inspired by Sharks: A Biomimetic Skeletonfor the Flapping, Propulsive Tail of anAquatic RobotA U T H O R SJohn H. Long, Jr.Department of Biology,Vassar College

Tom KoobMiMedx Group, Inc.

Justin SchaeferDavid Geffen School of Medicine,University of California

Adam SummersFriday Harbor Labs,University of Washington

Kurt BantilanDepartment of Biology,Vassar College

Sindre GrotmolDepartment of Biology,University of Bergen

Marianne PorterDepartment of Biology,Vassar College

A B S T R A C TThe vertebral column is the primary stiffening element of the body of fish. This

serially jointed axial support system offers mechanical control of body bendingthrough kinematic constraint and viscoelastic behavior. Because of the functionalimportance of the vertebral column in the body undulations that power swimming,we targeted the vertebral column of cartilaginous fishes—sharks, skates, and rays—for biomimetic replication. We examined the anatomy and mechanical properties ofshark vertebral columns. Based on the vertebral anatomy, we built two classes ofbiomimetic vertebral column (BVC): (1) one in which the shape of the vertebraevaried and all else was held constant and (2) one in which the axial length of theinvertebral joint varied and all else was held constant. Viscoelastic properties ofthe BVCs were compared to those of sharks at physiological bending frequencies.The BVCs with variable joint lengths were then used to build a propulsive tail, con-sisting of the BVC, a vertical septum, and a rigid caudal fin. The tail, in turn, wasused as the propeller in a surface-swimming robot that was itself modeled after abiological system. As the BVC becomes stiffer, swimming speed of the robot in-creases, all else being equal. In addition, stiffer BVCs give the robot a longer stridelength, the distance traveled in one cycle of the flapping tail.Keywords: biomimetics, robot, vertebral column, propulsion

Propulsive Functions ofVertebral Columns

I n sharks and other fish, thebody’s primary skeleton is the ver-tebral column, which runs from thehead to the caudal fin (Summers &Long, 2006). The vertebral column isa jointed framework to which musclesattach and on which the muscles pullto create the traveling waves of flexurethat transfer momentum from thebody to the surrounding fluid. Thevertebral column is composed of rigid

elements, called vertebrae, connectedby flexible intervertebral joints(Grotmol et al., 2003; Koob &Long, 2000). The joints and theiradjoining vertebrae limit the body’s ki-nematic degrees of freedom, constrain-ing bending primarily to the lateraldirection in response to loads imposedby muscle, inertia, and external fluidforces (Grotmol et al., 2006; Porteret al., 2009; Symmons, 1979; Schmitz,1995). In this way, the vertebralcolumn functions to control dynamicreconfigurations of the self-propellingbody.

In roll-stable sharks and fish, lateralbody bending is characterized by theoverlay of harmonic and transient mo-

tions that range from small travelingflexures to large-amplitude standingwaves (Long et al., 2010). These bend-ing motions produce the propulsiveforces that create forward swimming,turning maneuvers, and rapid accelera-tions. Across many different kinds offish, the stiffness of the bending joints,measured as the apparent Young’s mod-ulus, E, ranges from 0.1 to 8 MPa(Long et al., 2002). The E is a me-chanical property, sometimes calledthe “material stiffness,” that measuresthe contribution of the material, inde-pendent of its geometric arrangementin the structure, to the structure’s resis-tance to changing shape when an exter-nal load is applied to it.


Joints with any appreciable stiffnessat all may at first seem to be a paradox:why not have low-stiffness joints thatcost little, in terms of mechanicalwork, to bend? The answer seems tobe two-fold: (1) Stiff joints increasetheir resistance, in terms of the abso-lute bending moment, M (in units ofNm), in proportion to the magnitudeof bending curvature, κ (m−1). This re-sistance, which can grow nonlinearlywith κ, serves as a brake, limitinglateral bending (Long et al., 2002).(2) Stiff joints store and release moremechanical work, so-called “elasticenergy,” than flexible joints (Long,1992). The amount of work releasedin elastic recoil is also in proportionto E and to the square of κ. Hence,the vertebral column functions bestas a spring when muscles have reachedtheir functional limits, at the end of a

large-κ bend when connective tissuesare stiffest.

To explore the mechanical designspace of vertebral columns, we createdbiomimetic vertebral columns (BVCs).The BVCs are modeled after the ver-tebral column of sharks. We chosesharks’ vertebral columns since theyare structurally simple, compared tothose of bony fish, consisting of cylin-drical centra, small neural and hemalarches, and thin intervertebral joints(Figure 1). Together, a centrum andits arches are called a vertebra, and inthe cartilaginous sharks, skates, andrays, the vertebrae (plural form of‘vertebra’) are composed of mineral-ized cartilage. The compressive stiff-ness of these vertebrae, measured byE, ranges from 25 to 500 MPa, over-lapping the lower range of E for bone(Porter et al., 2006). Compared to the

bone of mammals, for a given value of Ethe vertebrae of sharks are stronger, wherestrength is measured in terms of break-ing stress (Porter & Long, 2010). Thus,in some ways, vertebral columnsmade ofmineralized cartilage perform better thanvertebral columns made of bone.

In summary, vertebral columns serveat least three important propulsive func-tions during swimming: (1) they controldynamic reconfigurations of the bodyby limiting the kinematic degrees offreedom, (2) they brake high-amplitudebends by virtue of their stiffness, and(3) they integrate muscle work overtime by recoiling elastically.

To build a BVC that can function aspart of an aquatic propulsion system,we (1) characterized the morphology(size and shape) of the vertebral col-umns of sharks, (2) measured the me-chanical properties of those vertebralcolumns as they underwent sinusoidalbending, (3) used that informationabout morphology and mechanicalproperties to design BVCs, (4) testedBVCs as they underwent the dynamicbending characteristic of swimmingand propulsion, and (5) tested theBVCs as the primary skeleton in theflapping tail of an aquatic robot.

Morphology of Sharks’Vertebral Columns

In three individuals of the black-tip shark, Carcharinus limbatus, andthe bonnethead shark, Sphrynatiburo, we measured, from radio-graphs, the following features fromthe head to the beginning of the caudalfin (Figure 2): (1) the length of centrum,c, (2), the diameter of the centrum,d, (3) the length of the intervertebraljoint, j, and (4) the cone angle, Ξ,of the capsule of the joint. Black-tip sharks, members of the familyCarcharhinidae, were chosen because

FIGURE 1

Vertebral columns of two species of sharks. For scale, each centrum shown here is between 4- and6-mm long in the head-to-tail direction. Head is to the left; tail is to the right.


they are known to be fast swim-ming predators of fish. In contrast,bonnethead sharks, members of thehammerhead family Sphyrnidae, areknown less for their speed and morefor their maneuverability and abilityto find and eat crustaceans. Of sim-ilar adult body size, the two speciesrepresent contrasting swimming stylesand ecologies. Three individuals ofeach species were used for this study.

In the bonnethead shark, all themorphological features, except d,increased in size from the head to theend of the abdomen and then decreasedtowards the caudal fin (Figure 3). Inblacktip shark, only Ξ and c varied

from head to caudal fin. The sig-nificance (α = 0.05) of the morpholog-ical variation was determined with amultivariate analysis of covariance(MANCOVA), with species and posi-tion as main effects and individual asthe covariate (JMP 8.0.2., SAS Insti-tute, Cary, NC). Following an iden-tity model MANCOVA, univariateANCOVAs were also run.

The variation in the morphology ofthese vertebral columns was used toguide the construction of the BVCs.For each of the sharks’ morphologicalfeatures, we indicated which dimen-sions were used (see black arrows onthe ordinates, Figure 3).

Mechanical Properties ofSharks’ Vertebral Columns

Using the same freshly dissectedvertebral columns from which mor-phology was measured, we conducted3-point dynamic bending tests usingan MTS model Mini Bionix 858(Eden Prairie, MN) with a 500 Nload cell. For blacktip sharks, eachvertebral column was cut into fivesegments of 19 vertebrae each. Forbonnethead sharks, each vertebralcolumn was cut into five segments of14 vertebrae each. The number of seg-ments in each species was varied tokeep the absolute length of each testsegment approximately equal. Eachsegment was subjected to sinusoidalbending at a frequency, f (Hz), andmaximum curvature, κ (m−1), variedto hold constant the time rate ofchange of κ, which is equivalent tothe strain rate (actuator displacementamplitude of 2 mm s−1).

To characterize the viscoelasticproperties of the vertebral column dur-ing bending, the apparent storage andloss moduli, E′ and E″ (MPa), respec-tively, were measured at each combi-nation of two species, five segmentpositions, and three κ. The E ′ mea-sures the purely elastic component ofthe stiffness; it is the force proportionalto the magnitude of the bending of thevertebral column. The E″measures thepurely viscous component of the stiff-ness; it is the force proportional to thevelocity of the bending of the vertebralcolumn. These properties were calcu-lated from the following formulae:E ′ ¼ E*cos δ a n d E ″ ¼ E*sin δ ,wherein E* ¼ FmaxL3

48Iymaxand δ i s the

phase lag (radians) between the dis-placement and load signals. Moreover,Fmax is the force (N) measured at theload cell, L is the gauge length of thespecimen (m), I is the specimen’s

FIGURE 2

Measuring vertebral morphology of blacktip and bonnethead sharks. Representative X-raysshow the heavily mineralized vertebral centra, which possess an “X” shape in this two-dimensionalview that is from cone-shaped joint capsules. The dark space between vertebrae is the intervertebraljoint. The morphology of each vertebra and intervertebral joint was measured from digitized land-marks (blue dots). (Color versions of figures available online at: http://www.ingentaconnect.com/content/mts/mtsj/2011/00000045/00000004.)


second moment of area (m4), and ymax(m) is the distance from the presumedneutral plane of bending (transversecenter of specimen) and the lateral-most fibers of the specimen.

In blacktip sharks, the E ′ and E″values were of greater magnitude

( p < 0.05) than those of bonnetheadsharks (Figure 4). In both species,E ′ increased towards the tail, an effectthat is amplified at higher values of κ,as indicated by a significant (p < 0.05)interaction term. The significance ofthe variation in E ′ and E″ was deter-

mined using ANOVA, with species,position, and κ as main effects (JMP8.0.2., SAS Institute, Cary, NC).

Since the data blacktip and bonnet-head sharks were taken at a single am-plitude of strain rate (2.0 mm s−1), wesought additional information abouthow E′ and E″ vary with changes instrain rate. We also wanted to test thehypothesis that the intervertebral cap-sule, which contains liquid underabove-ambient pressure, uses its inter-nal fluid pressure to alter the apparentE′ and E″ of the vertebral column. Be-cause blacktip and bonnethead sharkswere not available for these tests,spiny dogfish, Squalus acanthias, wereused. Fresh 10-vertebrae segmentswere removed from the region of thefirst dorsal fin in three dogfish. Eachsegment was pressure-clamped at theterminal vertebrae and end-loadedwith bending moments, M (for exper-imental configuration, see Long et al.,2011). The bending motion was deliv-ered via moment arms attached to asingle-axis linear actuator using anMTSmodel Tytron 250 (Eden Prairie,MN) and a 50-N load cell. To test theeffects of both f and κ on E ′ and E″,each segment was bent sinusoidally ateach combination of five f values andthree κ values. In addition, to test theeffects of the integrity of the fluid-filledintervertebral joint capsule on E′ andE″, we repeated this suite of testsafter (a) puncturing a single joint cap-sule located in the middle of the seg-ment and (b) puncturing three jointcapsules, including the first one punc-tured and two adjoining capsules.

Increases in f increased only E′( p < 0.05) while increases in κincreased both E ′ and E″ (Figure 5).The only significant effect of punc-turing the intervertebral capsule waswhen three capsules were punctured,and even then only E″ increased. The

FIGURE 3

Vertebral morphology of sharks. Four dimensions were used to characterize the size and shape ofthe vertebral centra and the intervertebral joints. The means of three individuals for each speciesare shown; individuals ranged from 0.59 to 0.91 m in overall body length. The error bars indicatethe standard error of the mean. Black arrows show the specific dimensions represented in ourBVCs. MANCOVA, using the identity method, calculated a significant Wilkes λ (p < 0.0001),with a significant interaction of species and position and significant main effect of species; thecovariate, individual, was also significant. Partial correlations among the response variablesranged from a low of 0.38 between j and d to a high of 0.79 between Ξ and c. Significancelevel is indicated (*p < 0.05, **p < 0.01, ***p < 0.001).


significance of the variation in E′ andE″ was determined using ANCOVA,with puncture, f, and κ as main effectsand individual as the covariate (JMP8.0.2., SAS Institute, Cary, NC).

In summary, the vertebral columnsof sharks have mechanical propertiesthat are highly variable. As speciesand anatomical position change, so,too, do E ′ and E″. Within a given ver-tebral segment, the apparent storagemodulus, E ′, and the apparent lossmodulus, E″, can be altered by thebending that they undergo. Increasingthe segment’s curvature, κ, increasesbothE′ andE″; increasing the segment’sfrequency of bending, f, increases the

E′. Knowing the mechanical behaviorof shark vertebral columns under real-istic bending conditions creates specifi-cations for BVCs.

Designing BVCsTo begin to understand how to

control the mechanical behavior ofBVCs, we built two classes of shark-inspired BVC: (1) BVC with variablecone angle, Ξ (BVCΞ): vertebraewere created with variable Ξ and theBVC had constant joint length, j,and (2) BVCwith variable joint length,j (BVC j): vertebrae were created witha constant Ξ and the BVC had vari-

able j. In addition to exploring the ef-fects of the structures Ξ and j, wealso varied the amount of cross-linkingof the hydrogel material forming thejoint. Thus, we explored the BVC“morphospace,” the variety of designsdescribed by three dimensions: Ξ, j,and cross-linking. Part of this explora-tion involved the challenge of makingcomposite structures that concatenateflexible and rigid elements. After fabri-cation and mechanical testing of bothclasses of BVC, we selected a singleclass, the BVCj, for performance test-ing in a tail-flapping aquatic robot.

In the BVCΞ , vertebrae were de-signed in software (SolidWorks,Dassault Systèmes SolidWorks Corp.,Concord, MA) to have the followingvalues of Ξ: 15°, 30°, and 45° (Fig-ure 6). These values correspond tolow, medium, and high values of Ξmeasured in sharks (see Figure 3).The diameter, d, and axial length, c,of the vertebrae were fixed at 1 cmfor both. The j of the column wasfixed at 0.25 cm.

Vertebrae were fabricated with arapid prototyper (Z-Corp, model310), which produced a porous, plas-ter part that was subsequently infil-trated with cyanoacrylate (EZ bond5cps, K&R International, DiamondBar, CA). This process yielded verte-brae with mean compressive moduli,E (MPa) of 43, 50, and 61 for verte-brae with values of Ξ at 15°, 30°,and 45°, respectively. These values ofE are within the range measured forshark vertebrae (Porter et al., 2006).

Vertebrae of a givenΞ were assem-bled into a BVCΞ in two stages. First,seven vertebrae were linked together,spaced at the fixed j, with eight horsehairs (E in tension of 900 MPa)arrayed axially and affixed to theouter circumference of the vertebrae.These horse hairs served as first

FIGURE 4

Mechanical properties of the vertebral columns of sharks in sinusoidal bending. Points are meansfrom three individuals. Error bars are the standard error of the mean. Size of the symbol indicatesthe relative magnitude of the curvature, κ. Significance level is indicated (n.s. = not significant,*p < 0.05, **p < 0.01, ***p < 0.001).


approximations of the intervertebralligaments found in the vertebral col-umns of sharks. Second, a 10% por-cine gelatin solution was injected inbetween the vertebrae; the gelatin wassolidified at 4° C. Once solidified, eachBVCΞ was then subjected to oneof three fixation treatments: 0, 1%,or 5 % glutaraldehyde, a chemicalagent that cross-links the collagen inthe hydrogel. A total of nine differenttypes of BVCΞ were produced, witheach possible pairwise combination ofΞ and glutaraldehyde concentration.

In the BVCj, vertebrae were de-signed to have a Ξ of 90°, which cre-ated ring-shaped vertebrae (Figure 7).The d and c of the vertebrae werefixed at 0.5 and 1.0 cm, respectively.

The overall length of the BVCj wasfixed at 8.4 cm. As the number ofnonterminal vertebrae were variedfrom 0 to 11, j varied from 720 to0.5 mm. These values of j created arange that extended below and abovethe range of j measured in sharks (seeFigure 3).

Ver tebrae were mi l l ed f romDelrin™, a polyoxymethylene thermo-plastic. Delrin has a compressive Eof 3.1 GPa (Delrin Design Guide,Module III, from DuPont), whichlies in the middle of the range ofE values reported for shark vertebrae(Porter et al., 2006).

The ring vertebrae had an inner di-ameter of 0.8 cm, which matched theouter diameter of hydrogels made from

10% porcine gelatin fixed in 2.5%glutaraldehyde (Long et al., 2006).Vertebrae were slid onto the hydrogel,spaced evenly at the desired j, and af-fixed to the hydrogel with cyanoacry-late adhesive. A total of 12 differenttypes of BVCj were produced, onetype for each of the 12 different valuesof j. Three replicates of each type wereproduced and tested. Please note thatin the BVCj horse hairs were omittedbecause at all but the smallest valuesof j, the hairs cut into the hydrogel dur-ing bending.

Mechanical Propertiesof BVCs

The E ′ and E″ of the BVCs weremeasured in two different kinds ofsinusoidal bending test, which cor-responded to the tests performed onsharks’ vertebral columns. In theBVCΞ, 3-point bending tests wereconducted in a manner identical withthose on the blacktip and bonnetheadsharks. In the BVCj , end-loaded bend-ing tests were conducted in a manneridentical with those on the spiny dog-fish sharks. The E′ and E″ data for theBVCj have been analyzed previously(Long et al., 2011). In the analysishere, the data have been reanalyzedto calculate the mechanical work re-quired to bend the BVCj and the me-chanical work recovered as recoil.

In the BVCΞ , both E′ and E″ in-creased as the glutaraldehyde concen-tration increased, E ′ and E″ decreasedas the Ξ increased, and E ′ increasedand E″ decreased as κ increased (Fig-ure 8). The significance of the varia-tion in E ′ and E ″ was determinedusing ANOVA, with glutaraldehydeconcentration, Ξ, and κ. as maineffects ( JMP 8.0.2., SAS Institute,Cary, NC).

FIGURE 5

Mechanical properties of the vertebral column vary as a function of cycle frequency, f, and the in-tegrity of the intervertebral joint in the spiny dogfish, Squalus acanthias. Points are means fromthree individuals. Error bars are the standard error of the mean. Size of the symbol indicatesthe relative magnitude of the curvature, κ. Significance level is indicated (n.s. = not significant,*p < 0.05, **p < 0.01, ***p < 0.001).


Compared to the mechanical prop-erties of shark vertebral columns, theBVCΞ have values of E ′ that have awider range, overlapping the lowervalues and exceeding the sharks’ highervalues by an order of magnitude (com-

pare Figures 8 and 4). In contrast, theE″ values of the BVCΞ overlap onlywith those of the bonnethead shark;the BVCΞ has much lower values ofE″ than either the blacktip or spinydogfish shark. Moreover, the E ′ for

BVCΞ decreases as κ increases; wemeasured the opposite trend in sharks(see Figures 4 and 5). Hence, theBVCΞ is not a good biological modelin this sense. Our hypothesis as to thesource of this strain softening is thatthe horse hairs force the column tobend primarily by compression, ratherthan by a combination of tension andcompression.

In the BVCj, both E ′ and E″ de-creased nonlinearly as j increased(Figure 9). Compared to the mechan-ical properties of dogfish vertebralcolumns, the BVCj span a nearly iden-tical range of E ′ and E″ values. Thegreatest sensitivity to changes in j oc-curred at the smallest values of j (Fig-ure 9) in the region that correspondsto the j measured in the vertebral col-umns of sharks (Figure 3). In datashown elsewhere (Long et al., 2011),E ′ and E ″ of the BVC j increasedwith increasing κ, just as in sharks(see Figure 5 herein). Moreover the E′increased with increasing f, as like-wise seen in sharks (Figure 5).

The mechanical work to bend theBVCj increased with increasing κ andincreasing E′ (Figure 9). The mechan-ical work recovered as elastic recoil,Wrecoil , is a function of the resil-ience, R, which, over all the testingconditions and sizes of joints, averaged76%.

BVCs in Aquatic VehiclesThe flexible skeletons of sharks and

fish are inspiring the design of novelpropulsive systems and aquatic vehi-cles (for review, see Fish, 2006; Long,2007, 2011). Fins with life-like flexi-bility have been built to propel a 0.7 mlong robotic turtle (Long et al., 2006)and a 0.4 m long robotic knife-fish (Curet et al., 2011). Bodies withlife-like flexibility have been built to

FIGURE 6

BVCs (BVCΞ) with variable cone angles, Ξ, and constant joint length, j.

FIGURE 7

BVCs (BVCj) with variable joint lengths, j, and constant cone angle, Ξ.


propel a 0.7-m long robotic electricray (Krishnamurthy et al., 2010),a 0.5-m long robotic trout (Kruusmaaet al., 2011), a 0.12-m long mechan-ical sunfish (McHenry et al., 1995),and a 0.5-m long mechanical pick-erel (Conte et al., 2010). Of theseself-propelled aquatic vehicles, onlythe mechanical pickerel has anythingresembling a vertebral column:a piece of spring steel designed torelease mechanical work to poweraccelerations.

The BVCj presented here was in-vented to propel a surface swimming,0.3-m long tadpole robot (Longet al., 2006; Doorly et al., 2009),known as Tadro4 (Figure 10). BVCj

were attached to a servo motor thatcreated a sinusoidally varying pitchingmotion of a tail. That pitch bent theBVCj, creating a bending momentthat propagated down the length ofthe BVCj in a traveling wave that, inturn, oscillated the terminal caudalfin. In this configuration, withoutdistributed muscles, the BVCj actsas both a transmission system, transfer-ring momentum from the servo motorto the caudal fin, and as a propeller,directly transferring momentum tothe surrounding fluid.

Since Tadro4 was built to behavereactively, with sensorimotor feedbacksystems creating foraging and predatoravoidance, we needed a version thatcould be programmed to swim straightusing a constant flapping frequency ofthe tail, f, and lateral amplitude of thecaudal fin. That modified version ofTadro4 was called MARMT (MobileAutonomous Robot for MechanicalTesting), and it had a hull length of17 cm and a tail length of 10 cm(Long et al., 2011).

Outfitted with a given BVC j ,MARMT’s steady swimming perfor-mance was measured as swimming

FIGURE 8

The mechanical properties of the BVCs (BVCΞ) with variable cone angles, Ξ, and constant jointlength, j. Horizontal bars indicate the median, the lower and upper limits of the box indicate the25th and 75th percentiles, respectively, and the whiskers indicate the range.

FIGURE 9

The mechanical properties of the BVCs (BVCj) with variable joint length, j, and constant coneangle, Ξ. Top row: points represent the means of E ′ and E ″ pooled across f and κ; error barsare one standard error of the mean. Bottom row: points are not pooled.


speed, U, and stride length, the slopeof the line of U regressed onto f,which measures the distance therobot travels over one period of theflapping tail (Figure 11). As f increased

for any BVCj, so, too, did the U. TheBVCj with greater values of E ′ pro-duced a more rapid increase in U,over the same range of change in f,compared to BVCj, with smaller values

of E′ (Figure 11, top panel). The stridelength of MARMT increased initially,doubling as E ′ doubled, before taper-ing off.

When the BVCj operates in theflapping tail, MARMT’s swimmingperformance is clearly linked to themechanical properties of the BVCj , E′in the case shown here. Those me-chanical properties are, in turn, underthe control of the structure of theBVCj . Thus these experiments, takentogether, demonstrate the functionalrelationship between the structure ofthe BVCj and the performance of aself-propelled aquatic robot.

SummaryUsing the morphology and me-

chanical properties of the vertebralcolumns of sharks as our biologicaltarget, we built and tested a series ofBVC. The mechanical behavior ofthe BVCs, measured by the storageand loss moduli over a range of bend-ing frequencies and curvatures, can bealtered by changing (1) the materialproperties of the hydrogel that makesup the intervertebral joint, (2) thelength of the intervertebral joint, or(3) the shape of the vertebrae. BVCsare sufficient to function as propulsiveelements in swimming aquatic ro-bots: in Tadro4 and MARMT theBVC converts a simple pitch oscilla-tion from a servo motor into a waveof bending that drives the caudal finlaterally.

Having identified variables that in-fluence the mechanical behavior ofBVCs, we offer a few observations forthose wishing to build jointed, flexiblebiomimetic skeletons for use in flexi-ble, flapping propulsive systems:

(1) Design of biomimetic systems:Engineered systems that are much

FIGURE 10

The aquatic robot, Tadro4, is propelled by a BVC (BVCj). Tadro4 is a fully-autonomous surface-swimmer with a flattened circular body and propulsive undulatory tail. It is modeled after fish likethe extinct Drepanaspis and the living electric ray, Narcine. Using sensory input from photoresis-tors and IR proximity detectors, Tadro4 searches for and swims up light gradients while avoidingcollisions. Tadro4 is propelled by its submerged BVCj, which is wrapped in a thin membrane, at-tached to a caudal fin, and actuated by an oscillating servomotor. Tadro4 was developed by Doorlyet al. (2009). Photo of Drepanaspis specimen 8462, American Museum of Natural History. Photoof adult Narcine is courtesy of Dr. Steve Kajiura.


simpler than the targeted biolog-ical system can match and extendthe targeted range of mechanicalbehaviors.(2) Control of mechanical proper-ties: The spacing of rigid elementsin a flexible matrix is more impor-tant than the shape of the rigid ele-ments or the material properties ofthe flexible material.(3) Control of reconfiguration: Be-cause of the strain- and strain-rate-dependence of viscoelastic materials,

no passive, flexible propulsive sys-tem, if its E ′ and E″ matches thatof the vertebral column of sharks,will produce constant motions overa wide range of motor inputs.

This work is a straight-forward exam-ple of one method of biomimeticdesign (Fish, 2006; Long, 2007):describe, test, build, and test. Startby identifying a specific operationalcontext—aquatic undulatory pro-pulsion in this case. Then describe,quantitatively, the biological system’s

functional morphology. Next, testthe morphology’s mechanical behav-ior under physiologically relevant test-ing conditions. Finally, build and testsimple biomimetic models of the sys-tem that change just a single structuralvariable over a wide range. Repeat thisprocess with different variables, ceterisparibus, until the designer knowswhich variables permit the natural sys-tem and its operational range to bemimicked or extended in biomimeticform.

AcknowledgmentsWe thank Carl Bertsche, Nicole

Doorly, Carina Frias, AndresGutierrez,Jonathan Hirokawa, Kira Irving, DougPringle, Foster Ranney, HannahRosenblum, Hassan Sahktah, SoniaRoberts, Elise Stickles, Josh Sturm,and Janese Trimaldi for their help indesigning, building, and testing ver-tebral columns, BVCs, and the aquaticrobots. This work was supported bythe National Science Foundationof the USA (DBI-0442269 andIOS-0922605).

Lead Author:John H. Long, Jr.Department of Biology,Vassar College124 Raymond Avenue,Poughkeepsie, NY 12604-0513Email: [email protected]

ReferencesConte, J., Modarres-Sadeghi, Y., Watts, M.N.,

Hover, F.S., & Triantafyllou, M.S. 2010. A

fast-starting mechanical fish that accelerates

at 40 m s-2. Bioinspir Biomim. 5(3):1-9.

doi: 10.1088/1748-3182/5/3/035004.


& MacIver, M.A. 2011. Aquatic manoeuvering

with counter-propagating waves: A novel

FIGURE 11

Swimming performance of a surface-swimming robot, MARMT, propelled by a tail with the BVCj asthe primary skeleton. MARMT is a version of Tadro4 (Figure 10) modified for mechanical testingover a range of flapping frequencies of the tail, f (Hz). For all types of BVCj tested, swimming speedof MARMT, U, increased linearly with increases in f (all R2 values > 0.92). The rate of change ofU with respect to f is the stride length (distance traveled per period of the flapping cycle); it wasgreatest with BVCj having larger storage moduli, E ′. Three replicates of each kind of BVCj weretested (N = 36). Means (N = 12) are shown here. Data reanalyzed from Long et al., 2011.



8(60):1041-50. (10.1098/rsif.2010.0493).

Doorly, N., Irving, K., McArthur, G.,

Combie, K., Engel, V., Sakhtah, H., …

Long, J.H., Jr. 2009. Biomimetic evolutionary

analysis: Robotically-simulated vertebrates in

a predator-prey ecology. In: Proc. 2009

IEEE Symp. Artificial Life. 147-54. Nashville,

TN: IEEE (Institute of Electrical and

Electronic Engineers). doi: 10.1109/

ALIFE.2009.4937706.

Fish, F.E. 2006. Limits of nature and

advances of technology: What does biomi-

metics have to offer to aquatic robots? Appl.

Bionics Biomech. 3(1):49-60. doi: 10.1533/

abbi.2004.0028.

Grotmol, S., Kryvi, H., Keynes, R., Krossoy,

C., Nordvik, K., & Totland, G.K. 2006.

Stepwise reinforcement of the notochord

and its intersection with the myoseptum:

An evolutionary path leading to the develop-

ment of the vertebra? J Anat. 209:339-57.

doi: 10.1111/j.1469-7580.2006.00618.x.

Grotmol, S., Kryvi, H., Nordvik, K., &

Totland, G.K. 2003. Notochord segmen-

tation may lay down the pathway for the

development of the vertebral bodies in the

Atlantic salmon. Anat Embryol. 207:263-72.

doi: 10.1007/s00429-003-0349-y.

Koob, T.J., & Long, J.H., Jr. 2000. The

vertebrate body axis: Evolution and me-

chanical function. Am Zool. 40(1):1-18.

doi: 10.1668/0003-1569(2000)040[0001:

TVBAEA]2.0.CO;2.

Krishnamurthy, P., Khorrami, F., de Leeuw,

J., Porter, M., Livingston, K., & Long,

J.H., Jr. 2010. An electric ray inspired

biomimetic autonomous underwater vehicle.

Am Control Conf. 2010:5224-9.

Kruusmaa, M., Salumäe, T., Toming, G.,

Ernits, A., & Ježov, J. 2011. Swimming speed

control and on-board flow sensing of an

artificial trout. In: Proceedings of IEEE

International Conference of Robotics and

Automation (IEEE ICRA 2011). Shanghai,

China: IEEE (Institute of Electrical and

Electronic Engineers).

Long, J.H., Jr. 1992. Stiffness and damping

forces in the intervertebral joints of bluemarlin,

(Makaira nigricans). J Exp Biol. 162:131-55.

Long, J.H., Jr. 2007. Biomimetic robotics:

Building autonomous, physical models to test

biological hypotheses. Proc. Inst. Mech. Eng,

C, J. Mech. Eng. Sci. 221:1193-200.

Long, J.H., Jr. 2011. Biomimetics: Robotics

based on fish swimming. In: Encyclopedia of

Fish Physiology: From Genome to Environ-

ment, ed. Farrell, A.P., 1, 603-12. San Diego:

Academic Press.

Long, J.H., Jr., Koob-Emunds, M., Sinwell,

B., & Koob, T.J. 2002. The notochord of

hagfish,Myxine glutinosa: Viscoelastic properties

and mechanical functions during steady

swimming. J Exp Biol. 205:3819-31.

Long, J.H., Jr., Koob, T.J., Irving, K.,

Combie, K., Engel, V., Livingston, N., …

Schumacher, J. 2006. Biomimetic evolutionary

analysis: Testing the adaptive value of verte-

brate tail stiffness in autonomous swimming

robots. J Exp Biol. 209(23):4732-46.

doi: 10.1242/jeb.02559.

Long, J.H., Jr., Krenitsky, N., Roberts, S.,

Hirokawa, J., de Leeuw, J., & Porter, M.E.

2011. Testing biomimetic structures in

bioinspired robots: How vertebrae control

the stiffness of the body and the behavior

of fish-like swimmers. Integr Comp Biol.

51(1):158-75. doi: 10.1093/icb/icr020.

Long, J.H., Jr., Porter, M.E., Liew, C.W., &

Root, R.G. 2010. Go reconfigure: How fish

change shape as they swim and evolve. Integr

Comp Biol. 50(6):1120-39. doi: 10.1093/icb/

icq066.

McHenry, M.J., Pell, C.A., & Long, J.H., Jr.

1995. Mechanical control of swimming speed:

Stiffness and axial wave form in an undulatory

fish model. J Exp Biol. 198:2293-305.

Porter, M.E., Beltran, J.L., Koob, T.J., &

Summers, A.P. 2006. Material properties

and biochemical composition of mineralized

vertebral cartilage in seven elasmobranch

species (Chondrichthyes). J Exp Biol.

209:2920-8. doi: 10.1242/jeb.02325.

Porter, M.E., & Long, J.H., Jr. 2010.

Vertebrae in compression: Mechanical

behavior of arches and centra in the gray

smooth-hound shark (Mustelus californicus).

J Morph. 271:366-75.

Porter, M.E., Roque, C.M., & Long, J.H., Jr.

2009. Turning maneuvers in sharks: Predict-

ing body curvature from body and vertebral

morphology. J Morph. 270:954-65.

doi: 10.1002/jmor.10732.

Schmitz, R.J. 1995. Ultrastructure and

function of cellular components of the

intercentral joint in the percoid vertebral

column. J Morph. 226(1):1-24.

doi: 10.1002/jmor.1052260102.

Summers, A.P, & Long, J.H., Jr. 2006. Skin

and bones, sinew and gristle: The mechanical

behavior of fish skeletal tissues. In: Fish

Biomechanics, eds. Shadwick, R.E., & Lauder,

G.V., 141-77. San Diego: Academic Press.

Symmons, S. 1979. Notochordal and elastic

components of the axial skeleton of fishes

and their functions in locomotion. J Zool.

189(2):157-206. doi: 10.1111/j.1469-

7998.1979.tb03958.x.


P A P E R

Lateral-Line-Inspired Sensor Arrays forNavigation and Object IdentificationA U T H O R SVicente I. FernandezAudrey MaertensDepartment of MechanicalEngineering, MassachusettsInstitute of Technology (MIT)

Frank M. YaulDepartment of ElectricalEngineering and ComputerScience, MassachusettsInstitute of Technology

Jason DahlDepartment of MechanicalEngineering, MassachusettsInstitute of Technology

Jeffrey H. LangDepartment of ElectricalEngineering and ComputerScience, MassachusettsInstitute of Technology

Michael S. TriantafyllouDepartment of MechanicalEngineering, MassachusettsInstitute of Technology

A B S T R A C TThe lateral line is a critical component of fish sensory systems, found to affect

numerous aspects of behavior, including maneuvering in complex fluid environ-ments with poor visibility. This sensory organ has no analog in modern ocean ve-hicles, despite its utility and ubiquity in nature, and could fill the gap left by sonarand vision systems in turbid, cluttered environments.

To emulate the lateral line and characterize its object-tracking and shape recog-nition capabilities, a linear array of pressure sensors is used along with analyticmodels of the fluid in order to determine position, shape, and size of various objectsin both passive and active sensing schemes. We find that based on pressure infor-mation, tracking a moving cylinder can be effectively achieved via a particle filter.Using principal component analysis, we are also able to reliably distinguish betweencylinders of different cross section and identify the critical flow signature informationthat leads to the shape identification. In a second application, we employ pressuremeasurements on an artificial fish and an unscented Kalman filter to successfullyidentify the shape of an arbitrary static cylinder.

Based on the experiments, we conclude that a linear pressure sensor array foridentifying small objects should have a sensor-to-sensor spacing of less than 0.03(relative to the length of the sensing body) and resolve pressure differences of atleast 10 Pa. These criteria are used in the development of an artificial lateral lineadaptable to the curved hull of an underwater vehicle, employing conductive poly-mer technologies to form a flexible array of small pressure sensors.Keywords: underwater sensing, artificial lateral line, pressure sensor arrays

Introduction

The lateral line organ is a uniquesensory mechanism in fish, enablingcomplex behaviors based on the inter-pretation of local fluid mechanics. Forexample, the blind Mexican cavefish(Astyanax fasciatus) is able to navigatenew environments at high speedwithout collision and to identify andremember features of the environ-ment (Montgomery et al., 2001; vonCampenhausen et al., 1981). This sur-prising feat is accomplished relyingprimarily on its lateral line organ for

sensory feedback. All fish have thisorgan, although not all use it to theextent of the blind cavefish (seeMontgomery et al., 2001). Many fun-damental behaviors in fish have beenidentified by biologists to be lateral-line-mediated, including trackingprey by their wake (Pohlmann et al.,2004) and recognizing nearby physicalobjects (von Campenhausen et al.,1981). Although the lateral line is a bi-ological organ, its functionality wouldtranslate well to needs in underwatervehicle design, such as with object de-tection, navigation, and flow sensing.

The lateral line organ consists oftwo subsystems responding separately

to velocity and to pressure gradientson the surface of the fish, both usingthe same underlying sensory element,the neuromast, which responds directlyto flow velocity. In the case of thepressure gradient measurements, thesevelocity sensors are embedded under-neath the skin in canals periodicallyopening via pores to the external flow(van Netten, 2006). Biological studieshave demonstrated the ability of thefish to use their lateral line to interro-gate their environment through bothactive and passive sensing. In activesensing, a fish uses the flow generatedby repeatedly gliding near new objectsat a short distance (von Campenhausen


et al., 1981) to interrogate their shape.In particular, blind cave fishwere foundto detect and discriminate betweenstationary objects or openings of dif-ferent geometries in still water (vonCampenhausen et al., 1981; Weissert& von Campenhausen, 1981; Burt dePerera, 2004). During passive sensing, amoving object generates a flow field thatis detected by the stationary lateral line ofa still fish. Vogel and Bleckmann (2000)demonstrated that goldfish use theirlateral line to passively detect and dis-criminate the size, velocity and shape ofpassing rods in still water. One key ele-ment emerging from these studies is thatin both active and passive settings, theobject identification behavior is tied tothe pressure gradient measurements ofthe lateral line and appears independentof the velocity measurements.

While the lateral line organ’s role inmany fish behaviors is becoming pro-gressively better understood, therestill remain many questions about thelevel of information detail available viathe lateral line. The recent advances inunderstanding the central processingof the lateral line have all been with re-spect to the oscillating dipole stimulus(Curcic-Blake & van Netten, 2006,Goulet et al., 2007), which is inappli-cable to both passive and active objectsensing. In the passive case, a dipolemodel neglects the wake that formsabout a moving object. In the activesensing situation, the interaction be-tween the two bodies in still waterleads to very different pressure distri-butions. Due to the difficulty in study-ing the neurological aspects of how afish utilizes the stimulus of the lateralline, an artificial representation of thelateral line can give insight on the useof pressure sensing for biologically in-spired, engineered applications. Thispaper investigates the ability of pres-sure sensor arrays, emulating the lateral

line organ, to distinguish the shapes ofphysical objects through both activeand passive sensing, while also identi-fying physical requirements for a con-structed artificial lateral line to be usedfor underwater vehicle navigation ap-plications. The results demonstratethe promise of a lateral-line-like sensorfor autonomous underwater vehicles(AUVs) in severe environments.

Previous approaches with artificiallateral lines have taken a biomimeticapproach of reproducing the lateralline at the structural level, by recreat-ing a sophisticated neuromast-likecantilevered sensors and using a canalsystem similar to that of the fish inorder to obtain pressure gradientmeasurements (Chen et al., 2006;Yang et al., 2008). Yet, at its core,the portion of the biological lateralline associated with the behaviors ofinterest can be viewed as a more ele-mentary processing unit, taking dis-tributed pressure measurements asinputs and, via some processing thatremains to be understood, extractinginformation about the environment.Thus, the current work instead takesa bioinspired approach in which thelateral line is abstracted by a linearpressure sensor array and the resultinginformation is processed relying onmodern inference algorithms. In addi-tion to providing potential clues as tothe fundamental processes takingplace in the biological sensory system,the present abstracted approach, by di-rectly focusing on the desired func-tionality of the artificial lateral line,bears significant engineering benefits.As a sparse artificial lateral line can bereadily implemented, making use ofdiaphragm-based pressure sensors, theinitial focus is able to shift from sensorfabrication to the inference and process-ing. In turn, insights from the infer-ence results are then used to develop

specifications for optimal sensor de-sign. As another benefit of using pres-sure sensors to approximate the lateralline, note that the available pressureinformation includes, but is not limitedto, discrete pressure gradients (as in thelateral line canal system).

First, we present two experimentsdemonstrating passive object detectionthrough the use of an artificial lateralline. In this study, an artificial lateralline is used to track the size and locationof a moving cylinder and separately toidentify the shape of a moving cylinderbetween known possibilities. Second,we present an experiment demonstrat-ing active object detection, where amoving artificial lateral line is used toidentify the completely unknownshape of an object. In this study, thesensor arrangement is studied withrelation to the applied problem inorder to identify constraints of sensorspacing. Finally, we discuss the con-struction and testing of a prototype ar-tificial lateral line using conductivepolymer materials for use in object de-tection and navigation applications.

Passive CylinderIdentification

In passive object identification, aninteraction between the object and theenvironment generates a pressure fieldthat stimulates the sensor array. In theexamples considered here, this inter-action is between a moving cylinderand still water. The sensor array is alsostationary in the water. The problem ofidentifying a moving cylinder via thepressure along a lateral-line-like sensoris addressed in two stages. First, we dis-cuss an approach for determining theposition and size of a cylinder, focusingon a single shape, and subsequently, weconsider the question of distinguishingbetween different shapes.


Experimental SetupTwo experiments were used to an-

alyze passive object detection techni-ques. In the first experimental setup(see Figure 1A), the position trackingof a circular cylinder was experimen-tally tested using off-the-shelf pressuresensors (Honeywell 19C015PG4K)arranged in a linear array. Seven sen-sors were embedded along a 46-cmsquare flat plate spaced 1.9 cm(0.75 inch) apart, as shown in Fig-ure 1A. Using a mechanical stagewith mill imeter accuracy in thex-y plane, a 32-mm diameter circularcylinder was passed in front of the sen-sor array at constant velocity andknown position. These experimentswere performed in a 3.6 m × 1.2 m ×1.2 m water tank at the Singapore-MIT Alliance for Research and Tech-nology (SMART) Centre. Pressuresignals were amplified by a factor of1000 immediately adjacent to the sen-sors using AD620 instrumentationamplifiers and were next sampledusing an NI USB-6210 analog-to-digital converter. The sensors werecalibrated using static pressure.

In the second passive detection ex-perimental setup, as with the first, avertically mounted cylinder passes infront of a stationary linear pressuresensor array at constant velocity (Fig-ure 1B). In this experiment, fourHoneywell 242PC15M pressure sen-sors were enclosed in a streamlinedcylindrical enclosure. The sensorswere mounted rigidly with a small di-ameter outlet in order to minimizenoise. Due to onboard amplification,the sensors were only amplified by afactor of 10 before data acquisition(NI USB 6210). These experimentswere performed in the MIT tow-ing tank facility (36.6 m × 2.4 m ×1.2 m). As the goal of these experi-ments and subsequent analysis is to

distinguish between two cylindercross-section shapes known a priori,both square and round cross-sectioncylinders were towed past the sensorarray. In order to provide a reliablecommon denominator for shape clas-sification, the data was gathered froma variety of speeds, cylinder sizes, anddistances from the sensor array asshown in Table 1. Calibration of thepressure sensors was completed in situby comparing the amplitude of pres-

sure oscillations due to regularly gener-ated waves, using a wavemaker, to thetheoretical amplitude based on linear-ized surface wave theory.

Passive Cylinder TrackingTracking the position and size of a

circular cylinder can be interpreted asthe first step or lowest level of shapeidentification. This task is complicatedby the wake that forms when a cylindermoves. In order to track the position

FIGURE 1

Schematics of two experimental apparatuses for passive object identification and sample pressuredata from one apparatus. (A) Amoving circular cylinder that is towed past an array of seven stationarypressure sensors on a flat wall. (B) An array of four pressure sensorsmounted on a stationary stream-lined body with a square cylinder towed past the body. (C) A representative data set corresponding tothe layout in part B, filtered with a cut-off frequency of 100 Hz. (Color versions of figures availableonline at: http://www.ingentaconnect.com/content/mts/mtsj/2011/00000045/00000004.)


and size, an accurate and preferablysimple model for relating the cylinderstate to the pressure measurements isneeded as well as a technique for esti-mating the state in real time. Potentialflow solutions are an immediate candi-date for modeling an object in a fluidsince they provide an analytic solutionwith relatively few parameters to definethe complete flow field.

The base model chosen for trackinga circular cylinder in the present ex-periments is a Rankine half-body inpotential flow. This well-known struc-ture is described by the superpositionof a source and a free stream. A mirrorimage of the half-body provides thenecessary boundary conditions todefine the wall in the model. Thismodel was found to approximate thepressure field of a cylinder with awake when compared with pressuremeasurements obtained through a vis-cous numerical simulation. The radiusof the cylinder was taken as the dis-tance from the location of the sourceto the nearby forward stagnationpoint in the flow. It is important tonote that the dipole model frequentlyused in a vibratory setting in studieswith fish (e.g., Coombs & Janssen,1990) does not model the flow wellabout a steadily translating body dueto the absence of the wake in themodel. It is not surprising that a poten-tial flow model like the Rankine half-body approximates the experimentalsituation, since the flow outside theseparated region in the wake can bereasonably considered as potentialflow. Although the real situation isnot steady and the pressure far fromthe cylinder was found to convergemore slowly than in numerical simula-tions, a Rankine model captures theamplitude and shape of the pressure re-sponse well in the immediate vicinityof the cylinder. Using the velocity po-

tential from this model, the pressure atthe sensor locations is calculated usingBernoulli’s equation.

Although the pressure sensors usedin these experiments measure pressurewith respect to a fix reference value,one key element in the analysis fortracking a cylinder is to consider thedifference in pressure between adja-cent pressure sensors instead of the ab-solute pressure. Long-wavelength,small-amplitude disturbances at thefree surface of the tank, caused by themotion of test cylinder, were found tocause pressure fluctuations on theorder of 20 Pa, significant comparedto the 100 Pa pressure signals fromthe cylinder. Due to the long wave-length, taking the difference in pressuresignals effectively removes this contam-ination. As mentioned in the Introduc-tion, the portion of the lateral line thathas been associated with object identifi-cation responds to pressure differences(Coombs, 2001), so it is interestingthat the use of pressure differences isnecessary even when absolute pressuremeasurements are available.

For the purpose of tracking theposition and size of a cylinder, the for-ward model relating the hidden pa-rameters of interest to the availablemeasurements has been defined bythe Rankine half-body representationof the cylinder and wake, in conjunc-tion with the Bernoulli equation toobtain the pressure at the sensors. Totackle the inverse problem and esti-mate the cylinder state based on thepressure measurements, several addi-tional issues require consideration.First, the variables of interest are theposition (x and y as labeled in Fig-ure 1A) and radius (R) of the cylinder.In order to cast the problem in theform of a hidden Markov model (see,for example, Cappe et al., 2005), inwhich the state at each timestep de-

pends solely on the previous iteration,the velocity (u) must to be included inthe state vector. Using this hiddenMarkov model formulation allows forthe use of well-known estimation algo-rithms for solving the inverse problem,such as the Kalman filter extensions fornonlinear models (Gelb, 1974; Julier& Uhlmann, 2004). Second, the for-ward model requires a strict constrainton two of the state variables,R and y, inorder to correspond to a physical real-ization: 0 < R < y. Unfortunately, thereis no natural boundary in the measure-ment model that corresponds to thesephysical constraints, as the potentialflow model works uniformly well andthere is no sudden shift in the predictedpressure. As a consequence of this, an es-timation technique such as the extendedKalman filter fails when it is applied tonaturally noisy experimental data.

The successful tracking of the posi-tion and size of a cylinder was accom-plished instead using a particle filter(see Branko et al., 2004). This generalinverse problem technique operates byusing a number of samples, which rep-resent possible states. To each sample,there is a corresponding weight thatroughly tracks the quality of that sam-ple. At each time step, each sample isupdated to a new value randomly cho-sen from a distribution based on theprevious sample, and the weight ofthat sample is updated based on howlikely it is to have generated the mea-surements. The key advantage of thisapproach is that, in specifying thenoise distributions that govern the up-date of the samples, it is possible tolimit the range of state vectors tothose that correspond to a physical sys-tem. In the present case, the x positionand the velocity u are assumed to haveGaussian noise distributions, which isthe typical assumption for an uncon-strained variable where the noise is


generated by the accumulation ofmany small random influences. Thenoise distribution associated with they position is assumed to be a log-normal distribution, and for the radiusit is assumed to be a Gaussian distribu-tion that is truncated at 0 and y and re-normalized. These distributions satisfythe constraints between the variablesand have intuitive limiting behaviors.When the mean is far from zero withrespect to the standard deviation, thelognormal is approximately symmetricabout the mean. In the case of thetruncated Gaussian distribution, ifthe previous value of the radius isgreater than the new y value, then thedistribution is weighted heavily tolarger values and has a maximumlikelihood of y. The main cost tousing a particle filter is that it fre-quently requires a large number ofsamples to generate repeatable accurateresults. For this application, it wasfound that approximately 150 particlesare sufficient.

A typical result of the particle filterimplementation on experimental datais shown in Figure 2. In the figure,the black line corresponds to the esti-mate of the variable being outputfrom the filter, and the red line cor-responds to the true value. Note thatthe filter is initialized with a uniformdistribution and does not begin toconverge until there are significantpressure differences. As seen from theresults, the particle filter estimate gen-erally tracks the x position and radiuswell but underestimates the y position.The velocity is also underestimated inthis example, although both over-estimates and underestimates of the ve-locity were recorded in general. Theseresults demonstrate that using a steadypotential flow model, the radius andaspects of the position can be accuratelytracked. With further refinements of

the model, the underestimation ofthe y position could likely be over-come. The use of a simple analyticalmodel here makes it more likely thatequivalent information is also avail-able to a fish via its lateral line, and sim-ilarly likely that it could be adapted forquickly tracking objects near a hull.

Passive CylinderShape Classification

The second set of experiments isconcerned with the more complicatedproblem of shape classification. Fun-damentally, we consider whether it ispossible to distinguish between twosimilar shapes after a single pass of anobject past the sensor array, using arobust test for the decision. The testin question must identify whethera passing cylinder has a square orround cross section. Since the decisionis based on a single pass, the stochasticnature of the flow in the wake of thecylinder must be overcome.

The large data set of pressure ob-tained from the experimental setup inFigure 1B, corresponding to cylindersof different sizes, velocities, distances,and shapes provides a broad range ofparameters for verifying the ability ofany decision rule. In addition, thelarge number of runs for each pointin the test matrix allows for an accuratecomparison of the mean pressure re-

sponses. As shown in Figure 3, thereare distinct differences between theaverage pressure measured due to thecircular and square cross sections; anal-ogous differences exist for comparisonsin all the tested sizes and velocities. It isnotable that the most easily distin-guished differences in the pressure sig-nals occur after the zero crossing point(circled).

The analysis presented here devel-ops a test to determine, after the fact,whether a square or circular cylinderhas passed the sensors. Based on theobservations of the mean pressuretraces, it is tempting to choose featuressuch as the location of the minimumpressure to classify the shape of thecylinder. This approach is limited inthat the features would be localizedin the pressure trace and, therefore,highly susceptible to the type of noiseobserved in Figure 1C. Instead, we usea principal component analysis (PCA)to identify a limited number of featuresin the form of weighted sums over thefull pressure data from a single run.This type of feature depends on all thedata and is less susceptible to localizedperturbations such as those in thewake. By carefully applying the PCAto a training data subset, the resultscan be directed towards a decision rulein classifying the shape of the cylinders.

In implementation, PCAworks as asingular value decomposition of the

TABLE 1

Matrix of experiments for passive cylinder shape classification.

7.62 cm Diameter 5.08 cm Diameter

0.5 m/s 0.75 m/s 0.5 m/s 0.75 m/s

Square 100 runs 100 runs 100 runs 100 runs

Round 100 runs 100 runs 100 runs 100 runs

Each of the experiments listed corresponds to the cylinder passing at a distance of 5.1 mm from thesensors. Additional data, not listed in the table, were collected with the 5.08-cm-diameter cylinders passing1.27 cm from the sensors at 0.75 m/s.


sample covariance matrix of the data(Jolliffe, 2002; Jackson, 2003). One im-portant detail in the implementation isthat the mean of each initial feature (asample point in the pressure trace)must be removed. This mean is takenacross all training data, not over eachclass. With the mean removed, thesample covariance is straightforwardto calculate. Each resulting principalcomponent is a vector that accountsfor the maximum possible variance,subject to having unit area and beinguncorrelated with all the previous prin-cipal components.

Since only four sensors were avail-able for this experiment (as opposedto seven in the previous cylinder track-ing experiments) and the cylinderstimuli are translating at constant ve-locity, an equivalence between thepressure time history and the spatialpressure field was used in the analysis.This approximation is very accurate inregions in front of the cylinder, but lessso in the unsteady wake areas. In orderto compare data sets with different cyl-inder velocities on the same spatialscale, the 0.5 m/s velocity data isdown-sampled appropriately. In addi-tion, the data is aligned by the pointat which the pressure crosses zero(marked in Figure 3), which forms areliable internal placement marker.

Finally, the PCA approach does notnaturally produce features which cap-ture the cylinder shape. The decom-position of the covariance matrixgenerates principal components thatcapture the majority of the data vari-ance in the first few components.Therefore, the resulting principal com-ponents will be most useful if the pri-mary cause of differences in the data isdue to the shape of the cylinders.While the variation in the data due tothe flow separation and the randomphase of the wake is impossible to

FIGURE 2

Results of a cylinder tracking experiment using a particle filter. In the first four parts, each verticalslice in the surface corresponds to the probability density function for that variable at the time ofthe slice, given the previous pressure measurements. In these sections, the red line represents thetrue value of the parameter at each time instant, and the black line corresponds to the expectedvalue of the distribution. The final part shows the corresponding pressure difference measure-ments, filtered at 60 Hz for anti-aliasing.


remove, the velocity of the cylinderstrongly correlates with the pressuresignal amplitude and masks the re-sponse to shape. Normalizing eachpressure trace by the maximum pres-sure effectively removes this maskingeffect since the pressure amplitude isconsistently proportional to the veloc-ity squared.

Using the first three principal com-ponents based on a training set of datafrom a single sensor and covering all ofthe experimental parameters, a highlysuccessful classification test is ob-tained. These three principal compo-nents form a space in which the datapoints generated from experimen-tal runs are grouped into two roughlyellipsoidal clouds, which can be opti-mally separated by a decision plane,minimizing the sum of squared erroron the training data. Since the projec-tions of the data points on each axis arebased on a linear combination of thepressure data with the corresponding

principal component, a single vectorof weights can be found that corre-sponds to an axis perpendicular tothe decision plane (Figure 4B). Thisresults in a single linear combina-tion using these weights with thenormalized pressure measurementsto determine the score of an experi-mental run, with a positive valueimplying a square cross section. Whenapplied to the test data (excluding thetraining data), this test produceda very small misclassification rateof 1.2%.

While the decision test encapsu-lated in the first two parts of Figure 4results in a surprisingly high accuracy,the difficulty with using principalcomponents is that there is generallylittle corresponding intuition for whyit works so well. The key element ofthe decision test is given by the deci-sion weights, illustrated in Figure 4B.There are three distinct main sections

FIGURE 3

A comparison of the mean pressure measured at one experimental configuration reveals cleardifferences between the two cylinder shapes. The standard deviation is also plotted, offset by−2.0 kPa.

FIGURE 4

The elements of the PCA-based classification test (A and B), and the accuracy of the test usingdifferent subsections of the full decision coefficients (C).


to the decision weights based on thePCA results, labeled II, III, and IV inFigure 4B. Region II is the smallest inmagnitude and corresponds to the re-gion between the maximum pressureand the zero crossing. Region III corre-sponds to the area of largest differencebetween the two shapes shown in Fig-ure 3. Region IV weighs an area of thepressure response that is directly influ-enced by the wake of the cylinder. Thisregion has considerable variation in thedata, but it is unclear whether it is re-lated to the cylinder shape. Regions Iand IV on either side are very lightlyweighted. By zeroing out regions, it ispossible to examine the importance ofthe remaining portion of the decisionweights to the accuracy in classifyingthe shape. In this zeroing process (Fig-ure 4C), we find that data on the rearside of the zero-crossing point is themost important in classifying cylindershape. In fact, the data before the zero-crossing point appears to add almostno new information for classifyingthe shape.

The emphasis on region IV impliesthat the wake of the cylinder contains asubstantial portion of the informationbeing used to classify the shape. Ini-tially, this may appear to be throughvortex spacing, governed by the Strou-hal number. Although this could helpin classifying the shape if particularsizes were being compared, in these ex-periments the large round cylinder andthe small square cylinder had nearlyequal vortex spacing, meaning thathalf of the data would be difficult toseparate by this trait. This observation isinconsistent with the 1% error rate.

The results based on principal com-ponents demonstrate that it is possibleto distinguish between two relativelysimilar cross sections of moving cylin-ders based on an artificial lateral linesensor. This has been hinted at based

on experiments with fish, but theclear identification of shape withoutregard to changes in velocity or sizehas not been demonstrated. Given theimportance of the wake in identifyingthe shape in the experiment, it is pos-sible that the ability of the lateral line,and any artificial analogs, to identifyshapes does not extend to arbitraryshapes. In particular, the bluntnessof the leading face of the cylinderhas a strong impact on the locationand flow direction at the point offlow separation. If two shapes withsimilar sharp flow separation are em-ployed, it may be much harder to dis-tinguish them.

Active CylinderIdentification

In active object identification, weuse pressure measurements from an ar-tificial lateral line sensor array to locateand identify stationary objects in stillwater without prior information ontheir shape. The problem is consid-ered from a two-dimensional per-spective in which a moving fish-likebody glides at constant speed past acolumn-like stationary object ofunknown location, size, and cross-section geometry. The moving bodyis equipped with a finite number ofpressure sensors distributed over itssurface. The viscous effects areignored in the model used to calculatethe pressure and the limits of this as-sumption are discussed.

Hassan (1985, 1992) showed thatin the case of a fish gliding towards acylinder or a wall, the pressure in-creases noticeably in the front regionwhen the fish gets close to the obstacle.He also showed (Hassan, 1985) thatfor a fish gliding past a cylinder, thereis a univocal relationship betweenspacio-temporal hydrodynamic signa-

ture and size and distance to the cylin-der, an indication that an artificiallateral line may be able to distinguishthese features. More recent studies(Sichert et al., 2009; Bouffanais et al.,2011) have considered parameteriza-tions of the shape of an object relevantto potential flow models, giving ananalytic representation of the flowfield that can be easily related to pres-sure and velocity measurements.

Experimental SetupExperiments were conducted in the

SMART Centre water tank. The com-plete setup is shown in Figure 5A. The

FIGURE 5

Active sensing experimental layout. (A) A pictureof the experimental set-up used for active ob-ject identification (the foil is moving towardsthe photographer). (B) A schematic of thecross-section of the same experiment showingthe location of the pressure ports.


moving body used to both generate the flow and sense pressure was a NACA0018 foil (chord c = 15 cm and span s = 60 cm) cast with internal 0.318 cm PVCtubing to transmit pressure from taps at the foil’s midspan to the top. Honeywell19C015PG4K pressure sensors were mounted on top of the foil, and measure-ments were collected at a sampling rate of 500 Hz via a NI USB-6289 DAQ.The location of the sensor ports is shown in Figure 5B. The foil was dragged atvelocity v = 0.5 m/s past a static cylinder of elliptical cross section oriented at var-ious angles. At its closest point, the foil was 5-10 mm away from the cylinder.

The Forward and Inverse ProblemIt is believed that blind cave fish can encode separately the distance, size and

shape of objects (Burt de Perera, 2004). Therefore, a convenient and potentiallybiologically relevant way to parameterize the problem is to characterize an objectby two parameters accounting for its position, one size parameter and several shapeparameters. Another desirable feature of this parameterization is that the num-ber of shape parameters needed to account for the pressure decreases with thedistance to the object. Bouffanais et al. (2011) proposed such a characterization:

S θð Þ ¼ a þ R eiθ þ ∑k¼1

∞μke

ikθ

¼ x þ iy þ R eiθ þ ∑k¼1

∞ jμkjeik θαkð Þ

; θ ∈ 0; 2π½

wherein a (a = x + iy) refers to the location of the object, R to its size and each μk(μk = |μk|e

ikαk ) term is associated with a (k + 1)-gonal type of perturbation of theshape from that of a circle. As k increases, the impact of the μk term on the pres-sure field decays very quickly with the distance from the cylinder. For an ellipse,as in these experiments, only the first shape parameter μ1 is non-zero and there-fore the subscripts will be dropped.

Given the moving object, its trajectory, and the location, size and shape (a,R, μ) of the stationary object, the velocity potential anywhere in the flow fieldcan be expressed in terms of a singularity distribution over the surface of the ob-jects. The problem is solved numerically using a source panel method: the surfaceof each object is broken up into line segments of constant source strength (364 linesegments were used for the moving body and 150 for the stationary object). Thepressure at the sensor locations on the surface of the moving body can then beexpressed in terms of the potential at these points using the unsteady Bernoulliequation. The combination of the parameterization of the object shape andthe numerical potential flow model defines the forward methodology, relatingthe pressure measurements to the variables of interest (location, size, andshape). In experiments, the pressure measured by the sensors is corruptedby noise. The two main sources of noise are laboratory noise (electrical andmechanical) and the background noise of the fluid flow (which includes viscouseffects).

For proper inversion, the technique must be (1) robust to noise, (2) capable ofhandling non-linearity since the pressure does not depend linearly on the charac-terizing parameters of the stationary object, and (3) dynamic, in order to be used

for navigation of underwater vehicles.A particularly suitable algorithm forsuch inversion is the unscented Kalmanfilter (UKF) ( Julier & Uhlmann,2004), which is a robust dynamicprobabilistic signal filtering techniquefor highly nonlinear systems. TheUKF is more accurate than the moretraditional extended Kalman filter forhighly nonlinear problems and doesnot require the computation of deriva-tives for which no analytic expressionsare available. It also propagates the sta-tistics with fewer samples than themore powerful particle filter, whichmakes it better suited for real timeapplications.

Subtle modifications to the UKFare necessary before applying it to ourproblem. Non-physical configura-tions, such as bodies intersecting eachother cannot be identified by an UKFassuming Gaussian distributions. Thiscan skew statistics, producing unusableresults. To avoid this problem, if afterupdating the location of the movingobject, the estimated mean configura-tion is not valid, the estimated positionis shifted ‘out of the way.’ If the meanconfiguration is valid, the parameter αthat determines the spread of the sam-ples around the mean in the unscentedtransform (Wan & van der Merwe,2001) is chosen small enough (foreach time step) that all the samplesare valid.

The UKF and the forward modelare combined to solve the inverse prob-lem: locating and identifying a cylinderusing pressure measurements. In theresults and analysis discussed here,the steady pressure due to the constantvelocity of the foil has been subtractedfrom the pressure signal. The measure-ment covariance matrix used in theUKF was calculated for each runbased on the pressure measured 0.5-0.3 s before the characteristic drop


of pressure at the second sensor (seeFigure 7C).

Active Sensing Resultsand Analysis

Due to the amount of noise in theexperiments and the fact that it washighly correlated, all attempts to si-multaneously fully identify the loca-tion and geometry of the cylinderwere unsuccessful. However, fishhave been observed to pass severaltimes in front of new objects (vonCampenhausen et al., 1981), and itseems reasonable to assume that theyfirst locate the objects and estimatetheir size before refining their estimateof the shape. A similar approach is usedhere: the first pass is used to get a firstestimate of the position (x and y) andsize (R) of the cylinder. A second passusing the same data refines the first es-timates and estimates the shape para-meters (|μ| and α). This approach isconsistent with the dynamic hierarchi-cal access associated with the shapecharacterization parameters used byBouffanais et al. (2011). An exampleof such a process of cylinder geometryreconstruction is shown in Figure 6 fora cylinder with shape parameters R =3.81 cm and μ = 0.2i and the foil pass-ing 6.5 mm away from it.

Both behavioral studies (Weissert& von Campenhausen, 1981) andmathematical modeling (Hassan,1985) have shown that the presenceof the static object in still water cannotbe detected until the fish (or here thefoil) is very close to it (on the orderof the width of the moving body). Ascan be seen in Figure 6A, as soon asthe foil is close enough to the cylinder,the position (x and y) and size (R) esti-mates converge steadily towards thetrue value of the parameters. The sec-ond run allows for very good orienta-

tion (α) estimation (Figure 6B). Theaspect ratio of the ellipse (|μ|) is themore difficult feature to reconstruct.The estimated parameter |μ| first wan-ders around the actual value of theparameter before decreasing towardsa lower value. As can be seen in Fig-ure 6F even though the shape of theellipse has not been exactly recon-structed, the estimate of the half ofthe ellipse that is closest to the foil isreasonably accurate. This observationconfirms that object identificationbased on pressure sensing becomesless reliable for features further fromthe sensors.

The results demonstrate that objectlocalization and object recognition arepossible with experimental pressuremeasurements; however, there are lim-itations to the proposed method. Themain limitation is the use of potentialflow models to establish the governingequations for the filter and the simula-tions (the importance of viscosity onthe stimulus to the lateral line system

of fish is also discussed in Windsor &McHenry, 2009). A fourth pressuresensor placed near the tail of the foilmeasured an oscillating pressure signalcharacteristic of the wake, unlike thefirst three sensors. Particle imagevelocimetry was used to visualize theflow and compare it to the flow pre-dicted by the inviscid model. As canbe seen in Figures 7A and 7B, thetwo flows are nearly identical almosteverywhere, but as the foil passes thecylinder, separation occurs over asmall region on the foil (in the orangeellipse). Comparing the pressure mea-surements and simulations (Fig-ure 7C), we can see that the potentialflow model gives very good approxi-mations of the pressure measurementsuntil the second sensor passes the cyl-inder (black dotted line). After thatpoint at which separation occurs, theinviscid assumption is violated andthe model is no longer valid. There-fore, only the measurements beforethe black dotted line on the figure

FIGURE 6

Results of a cylinder detection and identification in two passes (A and B) using an UKF. The coloreddashed lines show the true value of the parameters. (a–f) The corresponding shape estimate (redcircle or ellipse), the actual cylinder (green dotted ellipse), and the position of the foil at varioustimes. (Color versions of figures available online at: http://www.ingentaconnect.com/content/mts/mtsj/2011/00000045/00000004.)


have been used to locate and identifythe cylinder.

Sensor Array ConstraintsThe use of off-the-shelf sensors se-

verely limits the density which can beachieved in the sensor array whilemaintaining the sensitivity to devia-tions from the static pressure neededfor engineering applications. Fromthe perspective of sensor sensitivity,the weakest signals are those associatedwith tracking the position and size of amoving cylinder. The maximum abso-lute pressure was as small as 30 Pa in

some of the experiments that still pro-vided successful tracking results. Thepressure data for the active sensing ex-periments was of a similar scale. Anysensors used for an artificial lateralline application must provide the sen-sitivity and correspondingly low noisefloor to distinguish small changes inpressure of at least 10 Pa.

The experiments in active sensingof objects provide a guide to the op-timal density of sensors needed forsimilar applications, since the curvedsurface of the foil and the self-generatedflow increase in the importanceof the instantaneous measurementsof the spatial pressure distribution.Simulations were performed to exam-ine the affect of sensor density on theability to identify a stationary object.The simulations consisted of thesame foil described earlier gliding at0.5 m/s and passing 7 mm away froma cylinder with the geometry para-meters R0 = 1.5 cm and μ = 0.2i.White Gaussian noise of standarddeviation 2 Pa was added to thesimulated pressure measurements. Be-tween 10 and 70 pressure measure-ment points were evenly distributedalong the front three quarters of thefoil (see Figure 8C), and the sampling

FIGURE 7

Comparison between simulated (A) and experimental (B) flow field. The green and orange ellipsesshow where the flows differ. (C) The pressure data (plain line) is filtered with a cut-off frequency of100 Hz. (Color versions of figures available online at: http://www.ingentaconnect.com/content/mts/mtsj/2011/00000045/00000004.)

FIGURE 8

Convergence time (A) and final error (B) of the object identification as a function of sensorspacing. (C) The location of the sensors for a spacing of 6 mm.


rate was chosen such that f = (number of sensors) / 20,000. The error was cal-culated as E ¼ 1

5∇RR0

þ ∇xR0

þ ∇yR0

þ∇μx þ∇μy

. For each simulation, the con-

vergence time (time elapsed before E < 0.2) and the final error were calculated.Each case was simulated 12 times, and the mean and standard deviations of thetime of convergence and final error were computed (Figures 8A and 8B). Thegoal of this procedure was to identify a sensor density below which the lack ofspatial density cannot be compensated by a higher sampling frequency.

Both the convergence time and final error plots suggest that, at least for theconfiguration considered, the performance of the object identification decreasesas the sensor spacing exceeds 5 mm, which corresponds to a spacing of 0.03 rel-ative to the length of the foil. The optimal spacing that these observations suggestscales favorably with the actual spacing of lateral line canal neuromasts in thetrunk canal of the blind Mexican cave fish, which is roughly 0.02 body lengths(measured from image in Windsor & McHenry, 2009). This optimal spacingderived from simulations is specific to identifying the shape of cylindrical objectsin which the cross section size is on the same order or slightly smaller than thebody length. In general, the optimal spacing for measuring the pressure distri-bution may scale with the size of the object being detected as well as the bodylength. In addition, based on the observed separation that occurs on the surfaceof the sensing body, the pressure sensors should be distributed only over roughlythe forward half of the body in order to maximize the effectiveness of the sensorarray without waste.

A New Approach for a Flexible Pressure Sensor ArrayIn order to achieve the optimal spatial distributions and sensitivity of pres-

sure sensors in an effective artificial lateral line, it is necessary to develop sensorarrays on smaller scales than those possible using off-the-shelf technologies. Inaddition to the spatial and sensitivity constraints, it would be necessary that thesensor be applied directly on a surface in order for a lateral line sensor to bepractical for ocean vehicles, instead of fabricating the sensing body aroundthe sensor as was necessary in the experiments reported here. To addressthese problems, we have begun to develop a thin, flexible, one-dimensionalarray of pressure sensors to meet the pressure and spatial resolution require-ments. The array is designed to conform to an AUV’s hull without protrudingsignificantly.

While MEMS pressure sensors are typically made with a silicon substrate(Senturia, 2001), the flexible sensor array described here is made entirely of asilicone elastomer material, allowing it to be rugged, waterproof, and flexible.A conductive polymer is used as the pressure sensitive element in order to func-tion while maintaining mechanical compatibility with the rest of the flexiblestructure.

The conductive polymer is composed of the silicone Polydimethylsiloxane(PDMS) doped with conductive carbon black particles (Ding et al., 2007). Thismaterial has been used for chemical sensors (Andreadis et al., 2007) but not for ahigh-resolution pressure sensor. Prior work involving large pressure sensing arrayshas concentrated on tactile pressure sensing, which requires reduced sensitivityand greater dynamic range (Someya et al., 2004; Harsanyi, 2000). In con-trast, the carbon black pressure sensors are designed specifically for the lateral line

application, where small pressure var-iations in the tens of Pascals are ofinterest.

Pressure Sensor Designand Fabrication

An individual pressure sensing cellis depicted and diagrammed in Fig-ure 9. Its active components consistof a resistive strain gauge patternedon the surface of a 10-mm-wide,1-mm-thick elastomer membrane,as shown in Figures 1A and 1B.PDMS was chosen as the membranematerial due to its low tensile mod-ulus (Schneider et al., 2008), whichimproves sensitivity. A differentialpressure across the surfaces of themembrane causes the membrane todeflect. For small pressures, the deflec-tion is linear (Senturia, 2001). The de-flection then induces strain in theresistive strain gauge. The resulting re-sistance change is measured using thefour terminals of the strain gauge,as depicted in Figures 9B and 9C. Aconstant current is applied throughthe outer terminals, and the voltage ismeasured across the inner voltage-tapterminals. This four-point probe mea-surement desensitizes the device to var-iations in contact resistance betweenthe wires and the strain gauge. Thevoltage taps are positioned to capturethe greatest resistance change at thecentral edge of the membrane wherethe greatest strain occurs (Senturia,2001).

For a lateral line application on anunderwater vehicle, the differentialpressure between sensors in the arrayis of interest, but the depth of theaquatic vehicle, which corresponds tothe absolute pressure, is not. Theslow-varying absolute pressure maybe cancelled out by low frequencypressure equilibration using the air


channel diagrammed in Figure 9A.This equilibration prevents damageto the sensors caused by the waterdepth at which the vehicle operates.It also relaxes the design requirementson the sensor, allowing it to have athin membrane that is sensitive whilenot requiring it to withstand largepressures.

In order to verify the performanceof the carbon black sensing elementin the context of a flexible pressure sen-sor, a single pressure sensing cell wasfabricated (Figure 9C). PDMS wascast in a mold to form the membranestructure overhanging a cavity asshown in Figure 9A. To form thestrain gauge, conductive carbon blackparticles approximately 1 μm in diam-eter were uniformly mixed intoPDMS. Patterning of the strain gaugewas accomplished with a stencil mask,which yielded a 100-μm-thick layer.The carbon black strain gauge term-inals were bonded to aluminum wirewith conductive epoxy. For testing

purposes, a glass slide was used to sim-ulate the rigid hull.

Sensor PerformanceThe dynamic response of the single

test pressure sensor was characterizedusing a manual pressure source inde-pendently measured using a Honey-well pressure sensor. The resistance ofthe test sensor’s carbon black straingauge was recorded as the pressurevaried. Figures 10A and 10B showthe resistance of the strain gauge as afunction of time, compared to thepressure measured by the off-the-shelfsensor.

From the dynamic response data inFigure 10, the sensitivity can be esti-mated as 0.55% change in resistanceover 100 Pa. An amplification circuitwas used to cancel the DC offset volt-age and amplify the small resistancechanges by 100, resulting in a roughly55% change in voltage over 100 Pa.This is more than sufficient for detect-

ing pressure variations on the order of10 Pa, so it is adequate for the lateralline application.

Figure 10C depicts the appliedpressure plotted against the measuredresistance for the dynamic tests in Fig-ures 10A and 10B, as well as a statictest. The static test was performed byapplying a series of pressure stepswith a 1-min hold time and recordingthe resistance at the end. The pressurewas stepped up, down, and up again tocharacterize sensor hysteresis. Theslope of the curves represents the sen-sitivity of the sensor, and the asym-metry between the up and down stepsis indicative of hysteresis. This is ex-pected because the PDMS membraneis a viscoelastic material that experi-ences both creep and stress relaxation(Schneider et al., 2008). Both behaviorsare byproducts of its low tensile modu-lus but are well understood and can bemodeled and accounted for using signalprocessing techniques. The creepcauses increased strain for a given pres-sure, resulting in the static responsehaving a greater slope than the dynamicresponse. The dynamic responses aremore linear and consistent from cycleto cycle than the static response, be-cause the creep is not significant onthe timescale of these tests.

In addition to the polymer creep,there is the time-dependent relaxationbehavior inherent in the conductivepolymer. From the data in Fig-ures 10A and 10B, there is a veryshort time constant in the carbonblack strain gauge resistance datawhen the pressure is being steppedup, but a much longer time constantwhen the pressure is being steppeddown. This is thought to be a productof the way the carbon black conductivepolymer responds to strain (Ding et al.,2007). However, as shown in dynamicresponse transfer curves of Figure 10C,

FIGURE 9

Diagrams and photo of a single pressure sensing cell with a 10-mm square membrane. The deviceis 3-mm thick.


this behavior is repeatable and consis-tent during cyclic loading, so futurework will consist of using signal pro-cessing to compensate for this effect.

Creating a Sensor ArrayExtending the single carbon black

sensor cell into an array involves bothminiaturization of each sensor celland routing of the carbon black straingauges. A prototype array of four pres-sure sensing cells is shown in Fig-ure 11. The four sensors share twocommon current terminals, and each

individual sensor has two voltagetaps, reducing the necessary wiring.Each cell has a 5-mm-wide, 0.5-mm-thick square membrane. The dimen-sions of the cells have been scaleddown by a factor of 2 from the singletest pressure sensor. The spatial resolu-tion of the array is determined by thesensor cell spacing, which is 7-mmcenter-to-center for this array.

Further miniaturization is possibleby reducing the dimensions of themembrane and strain gauge patternin order to achieve the sub-5 mm spac-ing found optimal in the active sensing

simulations. Ultimately, the mem-brane thickness and the width of thestrain gauge patterns will be the limit-ing factors. The carbon black–basedflexible sensor array is a promisingtechnology for the lateral line applica-tion because it meets the pressure andspatial resolution criteria, it has repeat-able and predictable performance, andit can conform to the hull of an aquaticvehicle without protruding.

ConclusionThis paper has considered several

aspects of object identification basedentirely on a lateral-line-like pressuresensor array, with the aim of ex-amining its potential for applicationin underwater vehicles. Object identi-fication was divided into passive objectidentification, where an external flowinteracts with the object to stimulatethe sensor array, and active objectidentification, where self-generatedflow is used to interrogate an objectin still water. In each case, experimentswere used to examine the problem ofestimating the position, size, andshape of a cylinder based on a lineararray of pressure sensors in a realisticnoise environment. Constraints withoff-the-shelf sensors in the experimen-tal implementation of an artificial lat-eral line, have led to the developmentof a flexible artificial lateral line thatmakes use of a new polymer sensingtechnology.

The passive sensing of a movingcylinder separated the identificationof the position and size from theshape into different experiments. Thewake generated by the motion ofthe cylinder was a critical componentin the success of each part. For the cyl-inder tracking, a simple steady poten-tial model of the wake was sufficientto allow a particle filter to track the

FIGURE 10

Dynamic and static responses of the single cell pressure sensor, normalized by the strain gaugeresistance at atmospheric pressure Rinitial ∼ 20 kΩ. Parts A and B compare the output of an off-the-shelf Honeywell sensor with the output of the carbon black pressure sensor when the samepressure is applied to both. Part C presents dynamic transfer curves using the dynamic datafrom parts A and B as well as static data obtained by applying a constant pressure to the sensorand holding it for 1 min at each data point.


position and size in real time. In orderto distinguish between a square andcircular cylinder, a decision rule de-rived from PCA was highly successful.This rule demonstrated robustness indealing with cylinders of differentsizes, distances, and speeds. In analyz-ing the decision rule, it was found thatpressure data in the highly variableregion near and aft of the flow separa-tion is vital in determining the cor-rect shape. This implies limitationsin the types of shapes that can be dis-tinguished passively, as dissimilarshapes with similar wakes may causedifficulty.

By recreating an active sensing en-counter with a foil standing in for thevehicle or fish, we demonstrated thepossibility of identifying the locationand shape of a cylinder without priorknowledge of its shape. A potential

flow model was chosen for simplicityand to avoid heavy calculations thatwouldmake it impossible to run the al-gorithm in real time. Using this modelwith an UKF allowed us to obtain avery good object location estimateand reasonable ellipse identificationusing three pressure sensors. The re-sults stress the importance of thehead lateral line of the fish (that hadbeen emphasized in the case of thefish moving towards an obstacle byHassan, 1986) when it passes an ob-ject, due to the timing and locationof flow separation. To make more useof a biomimetic trunk lateral line, thenext step is to develop a model thattakes the separation into account tobe able to extract information fromthe second half of the data.

For both active and passive sensing,we have found that information about

the location and size of a cylinder isavailable via an artificial lateral line.Beyond this point, however, there arestrong differences in the two scenarios.In the case of passive sensing, the flowseparation and wake are integral com-ponents to the pressure distributionand must be accounted for from thebeginning. In fact, it appears theshape informationmay largely be avail-able through these components. Incontrast, for active sensing flow sep-aration does not occur immediately,allowing it to be ignored in the ini-tial shape estimation as done in thispaper. Using pre-separation data lim-its the number of measurementsavailable but allows the use of a gen-eral model for identifying arbitraryshapes.

Simulations based on the activesensing experiments have shown thatincreasing the sensor spacing ofapproximately 0.03 body lengthsachieves the fastest and most accurateobject identification for cylindrical ob-jects with a diameter on the same scaleas the sensing body. Coupled with thesensitivity (less than 10 Pa) required tosuitably measure the pressure signals inthe experiments, this defines the speci-fications for an artificial lateral linethat could be used to identify or locatesimilar objects for an underwatervehicle. Of significant additional con-cern is the need for a sensor that canbe easily mounted to a hull withoutsignificantly disturbing the flow.Based on the design and test resultsshown, the carbon black-based flexiblesensor array is a promising technologyfor the lateral line application. Used asa sensing element, it can meet the pres-sure and spatial resolution criteria, ithas repeatable and predictable perfor-mance, and it is able to conform tothe smooth curves of a hull while pro-truding only 3 mm.

FIGURE 11

Diagram and photo of an array of four pressure sensors, each with a 5-mm transparent squaremembrane. The sensors share common current terminals. Each individual sensor has two voltagetap terminals to measure the resistance of the part of the carbon black strain gauge located on thecorresponding membrane.


AcknowledgmentsThe authors gratefully acknowledge

the support of the Singapore-MITAlliance for Research and Technology(SMART) program’s Center for Envi-ronmental Sensing and Modelingand that of the National Oceanic andAtmospheric Administration’s SeaGrant program under project numberR/RT-2/RCM-17.

Lead Author:Vicente I. FernandezDepartment ofMechanical Engineering,Massachusetts Institute of Technology5-424, 77 Massachusetts AvenueCambridge, MA 02139Email: [email protected]

ReferencesAndreadis, N., Chatzandroulis, S., Goustouridis,

D., Kosma, V., Beltsios, K., & Raptis, I. 2007.

Fabrication of conductometric chemical sensors

by photolithography of conductive polymer

composites. J Microelec Eng. 84(5-8):

1211-4. doi: 10.1016/j.mee.2007.01.050.

Bouffanais, R., Weymouth, G.D., & Yue,

D.K.P. 2011.Hydrodynamic object recognition

using pressure sensing. P R Soc A. 467(2125):

19-38. doi: 10.1098/rspa.2010.0095.

Branko, R., Arulampalam, S., & Gordon, N.

2004. Beyond the Kalman Filter: Particle

Filters for Tracking Applications. Boston:

Artech House. 318 pp.

Burt de Perera, T. 2004. Spatial parameters

encoded in the spatial map of the blind

Mexican cave fish, Astyanax fasciatus.

Anim Behav. 68(2):291-5. doi: 10.1016/

j.anbehav.2003.11.009.

Cappe, O., Moulines, E., & Ryden, T. 2005.

Inference inHiddenMarkovModels.NewYork:

Springer Science+Business Media, Inc. 652 pp.

Chen, J., Engel, J., Chen, N., Pandya, S.,

Coombs, S., & Liu, C. 2006. Artificial lateral

line and hydrodynamic object tracking. In:

Micro Electro Mechanical Systems 2006,

MEMS 2006 Istanbul. 19th IEEE Interna-

tional Conference on Micro Electro Mechan-

ical Systems. pp. 694-7. Istanbul: IEEE.

doi: 10.1109/MEMSYS.2006.1627894.

Coombs, S. 2001. Smart skins: Information

processing by lateral line flow sensors. Auton

Rob. 11(3):255-61.

Coombs, S., & Janssen, J. 1990. Behavioral

and neurophysiological assessment of the lat-

eral line sensitivity in the mottled sculpin,

Cottus bairdi. J Comp Physiol A. 167(4):

557-67. doi: 10.1007/BF00190827.

Curcic-Blake, B., & van Netten, S. 2006.

Source location encoding in the fish lateral

line canal. J Exp Biol. 209(8):1548-59.

doi: 10.1242/jeb.02140.

Ding, T., Wang, L., & Wang, P. 2007.

Changes in electrical resistance of carbon-

black-filled silicone rubber composite during

compression. J Polym Sci Part B Polym

Phys. 45(19):2700-6. doi: 10.1002/

polb.21272.

Gelb, A. 1974. Applied Optimal Estimation.

Massachusetts: MIT Press. 382 pp.

Goulet, J., Engelmann, J., Chagnaud, B.,

Franosch, J., Suttner, M., & Hemmen, J.

2007. Object localization through the lateral

line system of fish: Theory and experiment.

J Comp Phys A 194(1):1-17. doi: 10.1007/

s00359-007-0275-1.

Harsanyi, G. 2000. Polymer films in sensor

applications: A review of present uses and

future possibilities. Sens Rev. 20(2):98-105.

doi: 10.1108/02602280010319169.

Hassan, E.S. 1985. Mathematical analysis

of the stimulus for the lateral line organ.

Biol Cybern. 52(1):23-36. doi: 10.1007/

BF00336932.

Hassan, E.S. 1986. On the discrimination

of spatial intervals by the blind cave fish

(Anoptichthys jordani). J Comp Physiol A.

159(5):701-10. doi: 10.1007/BF00612042.

Hassan, E.S. 1992. Mathematical description

of the stimuli to the lateral line system derived

from a three dimensional flow field analysis:

I. The case of moving in open water and of

gliding toward a plane surface. Biol Cybern.

66:443-52. doi: 10.1007/BF00197725.

Jackson, J.E. 2003. A User’s Guide to

Principal Components. New Jersey: Wiley.

592 pp.

Jolliffe, I.T. 2002. Principal Component

Analysis. New York: Springer. 502 pp.

Julier, S.J., & Uhlmann, J.K. 2004. Un-

scented filtering and nonlinear estimation.

Proc IEEE. 92(3):401-22. doi: 10.1109/

JPROC.2003.823141.

Montgomery, J.C., Coombs, S., & Baker,

C.F. 2001. The mechanosensory lateral line

system of the hypogean form of Astyanax

fasciatus. Eviron Biol Fishes. 62(1-3):87-96.

doi: 10.1023/A:1011873111454.

Pohlmann, K., Atema, J., & Breithaupt, T.

2004. The importance of the lateral line

in nocturnal predation of piscivorous

catfish. J Exp Biol. 207(17):9271-8.

doi: 10.1242/jeb.01129.

Schneider, F., Fellner, T., Wilde, J., &

Wallrabe, U. 2008. Mechanical properties

of silicones for MEMS. J Micromech

Microeng. 18(6):065008. doi: 10.1088/

0960-1317/18/6/065008.

Senturia, S.D. 2001. Microsystem Design.

Boston: Kluwer Academic Publishers.

720 pp.

Sichert, A., Bamler, R., & van Hemmen, J.

2009. Hydrodynamic object recognition:

When multipoles count. Phys Rev Lett.

102(5):058104. doi: 10.1103/

PhysRevLett.102.058104.

Someya, T., Sekitani, T., Iba, S., Kato, Y.,

Kawaguchi, H., & Sakurai, T.A. 2004.

Large-area, flexible pressure sensor matrix

with organic field-effect transistors for artifi-

cial skin applications. P Natl Acad Sci.

101(27):9966-70. doi: 10.1073/pnas.

0401918101.

van Netten, S. 2006. Hydrodynamic detec-

tion by cupulae in a lateral line canal:

functional relations between physics and


physiology. Biol Cybern. 94(1):67-85.

doi: 10.1007/s00422-005-0032-x.

Vogel, D., & Bleckmann, H. 2000. Behav-

ioral discrimination of water motions caused

bymoving objects. J Comp Physiol A. 186(12):

1107-17. doi: 10.1007/s003590000158.

von Campenhausen, C., Riess, I., &

Weissert, R. 1981. Detection of stationary

objects by the blind cave fish Anoptichthys

jordani (Characidae). J Comp Physiol A Sens

Neural Behav Physiol. 143(3):369-74.

doi: 10.1007/BF00611175.

Wan, E., & van der Merwe, R. 2001.

The unscented Kalman filter. In: Kalman

filtering and neural networks, ed. Haykin, S.,

221-80. New York: Wiley. doi: 10.1002/

0471221546.ch7.

Weissert, R., & von Campenhausen, C.

1981. Discrimination between stationary

objects by the blind cave fish Anoptichthys

jordani (characidae). J Comp Physiol A

Sens Neural Behav Physiol. 143(3):375-81.

doi: 10.1007/BF00611176.

Windsor, S.P., & McHenry, M.J. 2009. The

influence of viscous hydrodynamics on the

fish lateral-line system. Integr Comp Biol.

49(6):691-701. doi: 10.1093/icb/icp084.

Yang, Y., Chen, N., Tucker, C., Pandya, S.,

Jones, D., & Liu, C. 2008. Biomimetic flow

sensing using artificial lateral lines. ASME

Conf Proc. 2007(43025):1331-8.


P A P E R

A Conserved Neural Circuit-BasedArchitecture for Ambulatory andUndulatory Biomimetic RobotsA U T H O R SJoseph AyersAnthony WestphalDaniel BlusteinDepartment of Biology andMarine Science Center,Northeastern University

A B S T R A C TThe adaptive capabilities of underwater organisms result from layered extero-

ceptive reflexes responding to gravity, impediment, and hydrodynamic and opticalflow. In combination with taxic responses to point sources of sound or chemicals,these reflexes allow reactive autonomy in the most challenging of environments.We are developing a new generation of lobster and lamprey-based robots that op-erate under control by synaptic networks rather than algorithms. The networks,based on the command neuron, coordinating neuron, and central pattern generatorarchitecture, code sensor input as labeled lines and activate shape memory alloy-based artificial muscles through a simple interface that couples excitation to con-traction. We have completed the lamprey-based robot and are adapting this sensor,board, and actuator architecture to a new generation of the lobster-based robot. Thenetworks are constructed from discrete timemap-based neurons and synapses andare instantiated on the digital signal processing chip. A sensor board integrates in-puts from a short baseline sonar array (for beacon tracking and supervisory con-trol), accelerometer, a compass, antennae, and optionally chemosensors. Actuatorcontrol is mediated by pulse-width duty cycle coding generated by the electronicmotor neurons and a comparator and power field-effect transistor (FET) systemhoused on low- and high-current driver boards. These circular boards are stackedin a tubular hull with the processor and batteries. This system can readily mimic thebiomechanics of the model organisms by the addition of hydrodynamic control sur-faces. The behavioral set results from chaining sequences of exteroceptive reflexesreleased by sensory feedback from the environment.Keywords: biomimetic, robot, UUV, lobster, lamprey

Introduction

The innate behavior of underwateranimals provides an effective model forthe adaptive behavior of unmannedunderwater vehicles (Ayers, 2004).Underwater animals must respond toa broad variety of environmental chal-lenges including turbidity, hydro-dynamic flow, heterogeneous andhighly structured bottom types andimpediment. Their relatively neutralbuoyancy renders them especially sus-ceptible to hydrodynamic perturba-tion. As a result, they have evolved abehavioral set that includes a broadvariety of compensatory responses toperturbation. This behavioral set re-sults from layered exteroceptive re-flexes responding to exteroceptivesensor input resulting from changesin orientation relative to gravity, im-pediment, chemical cues, and hydro-dynamic and optical flow (Ayers,2004; Blustein & Ayers, 2010).These layered exteroceptive reflexes canform taxic responses to point sourcesof sound or chemicals (Westphalet al., 2011). As the point sourcesformmotivational cues for goal achiev-ing behavioral sequences, they can

guide reactive autonomy in the mostchallenging of environments. Thetask is to capture these performanceadvantages in engineered devices.

The Biological ModelWe are developing a new genera-

tion of lobster and lamprey-basedrobots that operate under control bysynaptic networks rather than algo-rithms. Previous generations of thesevehicles were controlled by finitestate machines that were organized

around the elements of the correspond-ing neurobiological models (Ayerset al., 2000; Ayers & Witting, 2007).

The neuronal circuits that controlour current generation of vehicles arebased on the command neuron, coordi-nating neuron, central pattern genera-tor (CCCPG) architecture (Figures 1aand 1b) of innate animal behavior(Kennedy & Davis, 1977; Stein, 1978;Pearson, 1993). The networks are or-ganized into segmental central patterngenerators (CPGs) that control append-ages or axial body musculature in the


animal models and the robots (Ayerset al., 2010). The CPGs are coordi-nated among themselves by a categoryof neurons called coordinating neu-rons that pass status informationfrom a governing CPG to a governedCPG that alters its period to remaincoordinated at a particular phase,depending on the ratio of intrinsic fre-quencies of the governing and gov-erned CPGs (Selverston & Ayers,2006). This temporal resetting occurson a cycle-by-cycle basis to entrain theCPGs in a particular gait in the case ofwalking or to ensure propagation of awave of flexion down the body duringundulation (Figure 1).

The CPGs are brought into oper-ation and modulated by a category ofneurons called command neurons

(Kupfermann & Weiss, 1978). Com-mand neurons generally constitute thelocus at which the decision to evoke abehavioral act is made and projectfrom the brain through the central ner-vous system to bring the segmentalCPGs into operation. They typi-cally perform this process through themechanism of neuromodulationthrough second messengers that alterboth the cellular properties and syn-aptic connectivity within the CPG(Dickinson, 2006). By this mech-anism or through direct synapticmodulation, the same CPG can oftenproduce variations on a behavioral actin response to different commands(Selverston & Ayers, 2006).

The sensors we employ are con-figured to encode sensory input as a la-

beled line code (Bullock, 1968). In thisform of coding, each sensory neuron isa unique source of information. Theinformation consists of (1) the natureof the sensory stimulus (light, opticalflow, chemicals, bumps, etc.), (2) thereceptive field or position of the stim-ulus on the body and (3) the mag-nitude of the stimulus coded as anaction potential train where the fre-quency of the action potentials isproportional to the logarithm of thestimulus intensity. We configurethese sensory elements in networksthat filter out features of the envi-ronment through lateral inhibition,range fractionation and motion detec-tion. These filtered outputs provideinput to the command neurons torelease behavior.

We have completed the lamprey-based robot and are adapting this sen-sor, board, and actuator architecture toa new generation of the lobster-basedrobot (Figure 2). The lamprey robotfeatures an electronic nervous sys-tem that we are adapting to the newlobster-based robot. The key featureof this architecture is that it is general-izable between all animal models.

Electronic Nervous SystemsThe discrete time map-based

(DTM) neuron and synapse equations(Rulkov, 2002) phenomenologicallymodel neuronal activity. The modelhas two state variables x and y, two

FIGURE 1

CCCPG architecture with exteroceptive reflex. Labeled circles represent neurons; synapses areshown as connecting lines with triangular (excitatory) or circular (inhibitory) endpoints. (a) Neu-ronal circuit-based controller for a walking lobster robot. The effector organs of each body seg-ment are controlled by CPGs that contain a neuronal oscillator, a pattern generator and sets ofmotor neuron pools. The CPGs are coordinated among themselves by a set of coordinating neu-rons (CoN) that provide information about the activity status of a governing oscillator to a gov-erned oscillator. The CPGs are brought into operation by a set of command neurons (CN) thatinitiates their operation and controls their average frequency and amplitude. (b) Neuronalcircuit-based controller for a swimming lamprey robot. Slight modification of the CCCPG archi-tecture and effectors transforms the system’s motor output from walking to swimming. (c) TheCNs are organized into exteroceptive reflexes that are released by neuronally coded sensor infor-mation (rounded rectangles: heading from a compass, target orientation from SBA) through sen-sory interneurons, which mediate in place rotation and yaw during locomotion.

FIGURE 2

Biomimetic robots. (a) Fourth generation lobster-based robot. (b) Second generation lamprey-based robot.


control parameters α and σ, and aparameter β for integrated synapticinput. Variations in α and σ can con-figure neurons into a silent type, aspiking type, a bursting type and achaotic type (Ayers & Rulkov, 2007).Similar control parameters for the syn-apse instruments determine the syn-aptic strength, relaxation rate, releasethreshold, and reversal potential thatdetermine whether the synapse is excit-atory or inhibitory. The electronic ner-vous systems are first prototyped intheNational Instruments LabVIEW™software. Neuron and synapse in-struments are configured with dif-ferent properties, and the modeledneurons and synapses are wired to-gether in LabVIEW™.

Figure 3 demonstrates a simpleCPG circuit configured to illustrateoperation of the four types of neuronsin our CPGs. Here, a command neu-ron (1) initiates an oscillation betweena bursting neuron (2) and a spikingfollower (3) using a slow modulatorysynapse. A fourth coordinating neuron(4) can be activated in bursts to entrainthe bursting pattern evoked by thecommand. In contrast to coordinatingneurons that reset the timing of the os-cillation on a cycle-by-cycle basis byperturbation, command neurons initiateoperation of the circuits and modulatetheir average frequency and amplitudeas parameters (Figure 3b–c).

Board ArchitectureThe networks that control the ro-

bots are constructed from DTM neu-rons and synapses in procedural C andare instantiated on a Texas Instru-ments digital signal processing (DSP)chip. A common board architectureis used to control both the swim-ming and walking robots (Figure 4a).The board set consists of four types:

FIGURE 3

DTM network integration. (a) The modeled neuronal circuit: (1) command neuron, (2) burstingneuron, (3) spiking neuron, (4) coordinating neuron. (b) Parametric modulation of 2 and 3 gen-erates an antagonistic bursting pattern. Voltage vs. simulation iteration traces are shown for eachcomponent of the network. The trace below neuron 1 shows the current injected to initiate activity.(c) Perturbation of the bursting pattern in 2 by a coordinating neuron (3) to entrain the evokedrhythm. The trace below 4 shows the injected current into that neuron. Adapted from Ayers andRulkov (2007).

FIGURE 4

Configuration of robots. (a) Current implementation of board set of the lamprey-based robot. Thesonar stack processes the analog hydrophone signals from the short baseline array (SBA). Theelectronic nervous system is instantiated on a Texas Instruments TMS320C6727 chip on the CNSDSP board. Sensors and the SBA processor are housed on the sensor array board. Low- and high-current drivers provide the current pulse trains that activate the nitinol actuators. The robot oper-ates on a 12-V, 4.5-Ah NiMH battery pack. (b) Configuration of the lamprey robot. A hinge in thepitch module allows the undulator to alter its pitch relative to the hull to dive or climb. (c) Con-figuration of the lobster robot. A tubular hull similar to the lamprey robot houses the electronicsand batteries. Anterior claw-like and posterior abdomen-like hydrodynamic control surfaces pro-vide a thrust vector into the substrate to increase traction. Externally mounted sensors includeoptical flow detectors, hydrodynamic flow sensors on the antennae, and sonar transducers forbeacon tracking and supervisory input.


(1) The DSP board houses the DSPchip and interconnects to sensor andactuator boards. (2) A sensor arrayhouses a compass, inclinometers, ac-celerometers and a processor board toderive azimuth and inclination devia-tion signals from the short baselinearray stack. (3) A low-current driverboard receives logic signals from themotor neuron output from the DSPchip and in turn controls (4) a high-current driver that applies current toheat the individual nitinol actuators.

Artificial muscles constructed fromthe shape memory alloy nitinol moveboth the walking legs and the undula-tory body axis. The nitinol is operatedon a thermal cycle. When cooled byseawater, it can be deformed into mar-tensite state that is associated withabout a 5% length increase. Whenheated by electrical current, it trans-forms into the austenite state and con-tracts rapidly. Increases in the lengthof one muscle are produced by thecontraction of its antagonist. Boththe amplitude and velocity of the con-tractions can be graded by pulse-widthduty cycle modulation of the drivepulses. Excitation-contraction cou-pling with the motor neurons is medi-ated by a comparator circuit on the lowcurrent boards that thresholds theaction potentials to generate a squarewave pulse that controls a power onthe high-current boards to activatethe actuators. Changes in the firing fre-quency of the motor neurons providethe duty cycle modulation.

A common DSP board (Figure 4a)interfaces the sensor array board tothe current driver boards. A separateregulator board provides the load tothese boards via a 12 V NiMh batterypack. Feed-through connectors in theend caps lead the current conduc-tors to the actuators. The boards arestacked in a tubular hull whose length

can be varied to accommodate a varietyof mission packages.

Behavioral SetThis system can readily mimic the

biomechanics of the model organismsby the addition of hydrodynamic con-trol surfaces. Turns in the undulatoryrobot are mediated by modulationof the amplitude of the flexions tothe two sides as in the animal model(Ayers, 1989). The direction of propa-gation of the flexion waves along thebody axis can be reversed to mediatebackward swimming. Dives andclimbs can be mediated by alterationof the pitch of the hull relative to theundulator (Figure 4b). Dorsal flexionof the hull generates a low-pressurearea above the hull to mediate a climbwhile ventral flexion generates a low-pressure area below the hull to me-diate diving.

The primary response to hydro-dynamic flow in the lobster is to orientinto the flow, lower the anterior con-trol surfaces and elevate the posteriorcontrol surfaces. As the lobster is onlyslightly negatively buoyant, this createsa thrust vector into the substrate andincreases the traction of the legs onthe bottom. The three degree of free-dom walking legs of the lobster robotallow the vehicle to walk in all direc-tions (Ayers &Witting, 2007). Altera-tions in the degree of depression canregulate the height above the substrate,while variations along the long bodyaxis regulates pitch. Biasing the depres-sion on the two sides can correspond-ingly regulate roll to maintain primaryorientation on tilted substrates.

The behavioral set of both robotsis organized around exteroceptive re-flexes (Kennedy & Davis, 1977). Aninnate releasing mechanism composedof sensory neurons and interneurons

filters incoming information to extractrelevant features of the environmentsuch as bumps, tilt, hydrodynamicand optical flow (Figures 1c and 5a).These sensory releasers are coded in in-terneurons that, in turn, activate com-mand systems. The interneurons uselateral inhibition from low-thresholdto high-threshold elements to producerange fractionation so that differentranges of a scalar input are coded bydifferent sensory neurons, providingthe capability for detailed circuit logic.

An example of such exteroceptivereflexes are those involved in the me-diation of the yaw plane responses tohydrodynamic flow that occur duringrheotaxis. Lobsters typically walkwith their antenna projected to thefront (Figure 5a, I). If wave surgeoccurs from the side, it bends theupstream antenna medially and thedownstream antenna laterally (Fig-ure 5a, II). Our hypothesis is thatthis perturbation activates a rheotaxicinterneuron that activates the back-ward walking command on the up-stream side and the forward walkingcommand on the downstream side.This would cause the animal/vehicleto rotate in place into the flow. Whileorienting into flow, the animals projecttheir antenna laterally, which wouldswitch control to another bilateralpair of surge interneurons (Figure 5a,III). As the most upstream antennaewould be bent more than the moredownstream antenna, and these in-terneurons project to the contralateralforward walking commands, theanimal/vehicle would continue toyaw into the flow until current to thetwo antenna is balanced ensuringproper orientation into the flow formaximal hydrodynamic stability. Thehydrodynamic control surfaces canthen ensure proper traction to over-come the perturbation.


This overall control scheme appliesto a variety of environmental circum-stances and perturbations in the yaw,pitch, and roll planes. Many extero-ceptive reflexes form taxic systems.For example, the three hydrophoneshort baseline sonar array (SBA) onthe lamprey robot reports the devia-tion of the sonar beacon relative tothe hull orientation in terms of incli-nation and azimuth (Westphal et al.,2011). The azimuthal signal mod-ulates swim command systems tocause the vehicle to yaw toward thebeacon (Figure 1c) while the incli-nation signal modulates the pitch sys-tem to cause the vehicle to climb/divetoward the beacon. Taken togetherthese layered reflexes will cause thevehicle to home on a sonar beacon. Asimilar 2-D SBA is planned for thelobster robot to control yaw taxis.

The sonar transducers also providea capability for supervisory control.The vehicles will be sent supervisorycommands that specify a heading andodometry information for distance.The command will include a propen-sity to negotiate or investigate obsta-cles depending on the mission. At theend of the search vector, the vehiclewould annunciate its location to along baseline sonar array and be senta new search vector. By this mecha-nism an operator could supervise sev-eral robots simultaneously.

ConclusionThe neuronal mechanisms of

innate behavior can be applied to abroad variety of biomimetic under-water robots. Minimal modificationof neuronal components can easily

alter locomotory outputs of a network,as we have shown to produce bothswimming and walking. Even hybridsystems can be achieved, such as thegait of an alligator resulting from acombination of the lamprey and lob-ster CPG networks. Sensor packagescan be adapted with little restruc-turing. Layered exteroceptive reflexnetworks provide capabilities for navi-gation, investigation and obstacle ne-gotiation in unpredictable near-shoremarine environments with a mini-mum of supervisory control.

Lead Author:Joseph AyersDepartment of Biology andMarine Science CenterNortheastern University,East Point, Nahant MA 01908Email: [email protected]

ReferencesAyers, J. 1989. Recovery of oscillator function

following spinal regeneration in the Sea

Lamprey. In: Neuronal and Cellular Oscillators,

ed. Jacklett, J., 371-406. New York: Marcel

Dekker Inc.

Ayers, J. 2004. Underwater walking. Arthro-

pod Struct Dev. 33(3):347-60. doi: 10.1016/

j.asd.2004.06.001.

Ayers, J., & Rulkov, N. 2007. Controlling

biomimetic underwater robots with electronic

nervous systems. In: Bio-mechanisms of

Animals in Swimming and Flying, eds.

Kato, N., & Kamimura, S., 295-306.

Tokyo: Springer-Verlag.

Ayers, J., Rulkov, N., Knudsen, D., Kim,

Y-B., Volkovskii, A., & Selverston, A. 2010.

Controlling underwater robots with electronic

nervous systems. Appl Bionics Biomech.

7:57-67. doi: 10.1080/11762320903244843.

Ayers, J., Wilbur, C., & Olcott, C. 2000.

Lamprey robots. In: Proceedings of the

International Symposium on Aqua

FIGURE 5

Neural simulation of rheotaxis. (a) Epochs of a lobster’s rheotaxic behavioral response to watersurge shown with the predominant active neural reflex circuit. In the network diagrams, top circlesrepresent sensory neurons corresponding to high (H),medium (M), or low (L) antennal bending inthe lateral or medial direction. Black ovals represent interneurons that project to bilateral com-mands for forward (F) or backward (B) walking. In I, as a lobster walks forward, bilaterally balancedlow lateral bending of the antennae is observed, which serves to sustain forward locomotion. In II,left-to-right water surge (blue arrows) causes a high medial bend of the ipsilateral antenna and ahigh lateral bend of the contralateral antenna eliciting rheotaxis. In III, bilaterally asymmetricallateral antennal bending mediates yawing upstream during forward walking. (b) Voltage vs.time traces for the neurons of the rheotaxis circuit. Dashed lines distinguish the behavioral epochsshown in a.


Biomechanisms, eds. Wu, T., & Kato, N.,

Tokai University.

Ayers, J., & Witting, J. 2007. Biomimetic

approaches to the control of underwater

walking machines. Philos T R Soc Lond A.

365:273-95. doi: 10.1098/rsta.2006.1910.

Blustein, D., & Ayers, J. 2010. A conserved

network for control of arthropod exteroceptive

optical flow reflexes during locomotion. Lect

Notes Artif Int. 6226:72-81.

Bullock, T.H. 1968. Representation of infor-

mation in neurons and sites for molecular

participation. P Natl Acad Sci. 60:1058-68.

doi: 10.1073/pnas.60.4.1058.

Dickinson, P.S. 2006. Neuromodulation of

central pattern generators in invertebrates and

vertebrates. Curr Opin Neurobiol. 16(6):

604-14. doi: 10.1016/j.conb.2006.10.007.

Kennedy, D., & Davis, W.J. 1977. Organi-

zation of invertebrate motor systems. In:

Handbook of Physiology, eds. Geiger, S.R.,

Kandel, E.R., Brookhart, J.M., &Mountcastle,

V.B., Section 1, Vol. 1, Pt. 2:1023-87.

Bethesda, MD: American Physiological

Society. Section 1, Part 2.

Kupfermann, I., & Weiss, K. 1978. The

command neuron concept. Behav Brain

Sciences. 1:3-39.

Pearson, K.G. 1993. Common principles of

motor control in vertebrates and invertebrates.

Annu Rev Neurosci. 16:265-97. doi: 10.1146/

annurev.ne.16.030193.001405.

Rulkov, N.F. 2002. Modeling of spiking-

bursting neural behavior using two-dimensional

map. Phys Rev E. 65:041922. doi: 10.1103/

PhysRevE.65.041922.

Selverston, A., & Ayers, J. 2006. Oscillations

and oscillatory behavior in small neural circuits.

Biol Cybern. 95:537-54. doi: 10.1007/

s00422-006-0125-1.

Stein, P.S.G. 1978. Motor systems, with

specific reference to the control of locomotion.

Annu Rev Neurosci. 1:61-81. doi: 10.1146/

annurev.ne.01.030178.000425.

Westphal, A., Rulkov, N., Ayers, J., Brady, D.,

& Hunt, M. 2011. Controlling a lamprey-

based robot with an electronic nervous system.

Smart Struct Syst. 7(6):471-84.


P A P E R

A Hybrid Class Underwater Vehicle:Bioinspired Propulsion, EmbeddedSystem, and Acoustic Communicationand Localization SystemA U T H O R SMichael KriegPeter KleinRobert HodgkinsonDepartment of Mechanicaland Aerospace Engineering,University of Florida

Kamran Mohseni1

Departments of Mechanicaland Aerospace Engineeringand Computer Engineering,Institute for Cyber AutonomousSystems, University of Florida

A B S T R A C TInspired by the natural locomotion of jellyfish and squid, a series of compact

thrusters series is developed for propulsion and maneuvering of underwater vehi-cles. These thrusters successively ingest and expel jets of water in a controlledman-ner at high frequencies to generate propulsive forces. The parameters controllingthe performance of the thrusters are reviewed and investigated to achieve higherthrust levels. The thrusters are compact and can be placed completely inside a ve-hicle hull providing the desired maneuvering capability without sacrificing a sleekhydrodynamic shape for efficient cruising. The system design of a prototype hybridvehicle, called CephaloBot, utilizing these thrusters, is also presented. A compactand custom-developed embedded system is also designed for the CephaloBot. Keyfeatures of the system include a base set of navigational sensors, an acousticsystem for localization and underwater communication, Xbee RF transceiver forcommunication above water, and a LabVIEW programmed processing board.Keywords: AUV, thruster, bioinspired, communication

Introduction

Traditionally unmanned under-water vehicles fall into one of two cat-egories. One class of vehicles (torpedolike) are built to travel long distanceswith minimal energy and are usuallycharacterized by a long slender body,a rear propeller for propulsion and aset of fins to provide maneuveringforces. This type of vehicle is poorlysuited for missions requiring a high de-gree of positioning accuracy becausethe control surfaces provide little tono maneuvering force at low forwardvelocity. The other class of vehiclesimilar to remotely operated vehicles(ROVs) is designed to operate inthese situations, which do require

high positioning accuracy, and incor-porate several thrusters at various lo-cations to provide maneuvering forcesin all directions. However, this classof vehicle typically has a very highdrag coefficient due to the abundanceof external thrusters and cannot travelto remote locations without addi-tional support.

The abundance of remote marineresearch sites requiring high position-ing accuracy for inspection, as well asthe desire to create fully autonomousvehicle sensor networks, has inspiredsignificant research in a hybrid classof vehicles with the efficient cruisingcharacteristics of the torpedo classand the maneuvering abilities of theROV class. Some take a mechanicalapproach moving the maneuvering

propellers into tunnels which runthrough the hull of the vehicle(Mclean, 1991; Torsiello, 1994) orinto the fins themselves (Dunbabinet al., 2005). Others observe that na-ture’s swimmers have a healthy balanceof long-distance endurance and high-accuracy low-speed maneuvering. Ve-hicles have been designed to use finsfor both high-speed maneuvering aswell as mimic the low-speed flappingof turtles and marine mammals (Lichtet al. , 2004; Licht, 2008; Kato,2011); and some use tail fins as a pri-mary means of propulsion (Barrettet al., 1999). Our inspiration comesfrom the cyclical jet propulsion seenin jellyfish, scallops, octopus, squidand other cephalopod. Squid jet pro-pulsion produces the fastest swimming

1This work was started while the group was atthe University of Colorado.


velocities seen in aquatic invertebrates(O’Dor & Webber, 1991; Anderson& Grosenbaugh, 2005).

Jetting locomotion begins whenthe squid inhales seawater through apair of ostia behind the head, fillingthe mantle cavity (see Figure 1). Themantle then contracts forcing seawaterout through the funnel that rolls into ahigh-momentum vortex ring and im-parts the necessary propulsive force(Anderson & Grosenbaugh, 2005).The versatility of the system permitstwo distinct gaits, cruising and escapejetting (Bartol et al., 2008). Duringcruising, squid swim at nominalspeed with a greater efficiency than es-cape jetting, which involves a hyper-inflation of the mantle followed by afast powerful contraction to impartsignificant acceleration at the cost ofboth muscular and fluid dynamiclosses. Bartol et al. (2009) report cruis-ing mode efficiency at 69% (±14%)averaged over several species andswimming speeds and 59% (±14%)for escape jetting. Additionally, pro-pulsive efficiency was seen to rise ashigh as 78% in adult L. brevis swim-ming at high velocities and averaged87% (±6.5%) for paralarvae (Bartolet al., 2008), challenging the notionthat a low-volume high-velocity jetinherently negates a high propulsiveefficiency.

The locomotion of jellyfish tendsto be very similar to that of squidwith some key differences, primarilythat the refilling phase of jellyfish

swimming uses the same bell openingas the jetting phase. Despite the factthat squid do not use the funnel dur-ing refilling, the inlet vents are stillon the anterior side of the mantle cav-ity, meaning that locomotion for bothorganisms is quite different from tradi-tional pumping mechanisms. Jelly-fish use the cyclic jetting process forfeeding as well as locomotion as isevidenced by Lagrangian coherentstructures and particle tracer analysis(Lipinski & Mohseni, 2009; Wilsonet al., 2009); in addition, both squidand jellyfish utilize jetting for respi-ration, taking advantage of the largefluid flow rates. Both of these factorscan make it difficult to determinewhich swimming behaviors are opti-mized for propulsion versus secondaryfunctions. Similar to the different gaitsseen in squid locomotion, differentspecies of jellyfish generally fall intotwo categories of swimmers based onthe ‘quality’ of vortex ring they pro-duce. Jellyfish like moon jellyfishhave a very large bell opening, andthe jetting motion is similar to a pad-dling type motion. Box jellyfish andother faster swimming jellyfish havesmaller bell openings with nozzle-likeflaps and have a much more distinctjet. Jellyfish morphology during swim-ming has been digitally captured fromexperiment, and the body motionswere imported into numerical simu-lations to predict body forces on theswimming jellyfish, determining dras-tically different swimming efficiencies.

Froude propulsive efficiency of jelly-fish was directly calculated by Sahinand Mohseni (2008, 2009; Sahin et al.,2009) to be 37% for Aeqorea victoriaand 17% for Sarsia tubulosa. It shouldbe noted that both species of jelly-fish most likely do not use vortexgeneration for the sole purpose oflocomotion. Aeqorea victoria uses vor-tex generation for feeding and Sarsiatubulosa as an escape mechanism. Em-pirical data gathered through digitalparticle image velocimetry (DPIV)measurements of several species showssimilar efficiency characteristics forthe different swimming patterns(Dabiri et al., 2010).

The general concept of propellingwater craft by ejecting a high-velocitywater jet is centuries old, was hypoth-esized by both Bernoulli and BenjaminFranklin, and was utilized in a rudi-mentary sense in one of the first steamboat designs by James Rumsey (Allen,2010). Continuously pumped jets areused for propulsion in modern watercraft-like jet skis and bow thrustersof motorboats; however, this type ofjet propulsion is inherently differentfrom the propulsion of squid and jelly-fish, which create distinct vortex rings.The thrusters of this paper also pro-duce finite jets, which form arrays ofvortex rings, and should be consideredfundamentally different than continu-ous jet thrusters.

This paper showcases a completehybrid class vehicle that demonstratesadded maneuvering capabilities utiliz-ing a set of bio-inspired jet thrusters.The manuscript will focus on threeprimary systems of the vehicle: a bio-inspired thruster system and fun-damentals of thruster mechanics, anacoustic system, which serves thedual purpose of communication andlocalization, and a compact embeddedcontrol system. The manuscript is

FIGURE 1

Diagram of squid layout and locomotion.


organized as follows. The mechanicsof the thruster as well as the thrustdynamics are described in ‘VortexRing Thrusters’ section. A brief historyof thruster and vehicle prototypes isgiven in ‘Thruster and CephaloBotEvolution’. The ‘Hybrid Vehicle De-scription’ section gives basic require-ments for a hybrid class vehicle, and thesubsystem components of CephaloBotare described in more details in ‘Com-munication Localization System’,‘Embedded System’, and ‘Sensors’sections.

Vortex Ring ThrustersOur thruster inspired by jellyfish

and squid propulsion consists of an in-ternal fluid cavity, with a semi-flexibleplunger used to drive fluid motion,and a small circular orifice exposed tothe external fluid. See Figure 2 for adiagram of the thruster layout. Thecavity of the thruster provides thesame functionality as the squid mantleor the jellyfish bell, expanding andcontracting to cycle water in and outof the circular orifice (functionallysimilar to the squid funnel/siphon).Since the thruster generates propulsionby creating energetic vortex rings, it

is termed the ‘vortex ring thruster’(VRT).

Since VRTs are contained internalto the vehicle (with only a small open-ing on the vehicle surface), they do notsignificantly affect the vehicle’s for-ward drag profile, which means that avehicle equipped with a set of VRTsfor low-speed maneuvering and a rearpropeller for primary propulsion willhave a sleek aerodynamic shape allow-ing fast efficient cruising to a site ofinterest but still maintain full maneu-verability (even at zero forward speed)upon reaching that site of interest. SeeKrieg andMohseni (2010) for ‘parallelparking’ capability of an earlier versionof our vehicle. Additionally, sincethe VRT only needs a single opening(unlike tunnel thrusters or traditionalpumps, which extend from one endof the hull to the other), it allows fora greater degree of freedom for internalsystem arrangement.

The impulse generated by this typeof device can be modeled as if the jetacts like a solid slug of fluid with a uni-form velocity across the nozzle open-ing (Mohseni, 2004, 2006; Krieg &Mohseni, 2008). Properties of fluidslug have been investigated by severalgroups (Glezer, 1988; Gharib et al.,1998; Shariff & Leonard, 1992;Mohseni & Gharib, 1998; Krieg &Mohseni, 2008). The total impulsegenerated for a single pulsation underslug assumptions is

Islug tð Þ ¼ ρπ=4∫t0u2 τð ÞD2dτ ; ð1Þ

where u is the piston velocity (massflux across thruster opening dividedby nozzle area), D is the nozzle Diam-eter, ρ is the fluid density, t is the timeat which the impulse is evaluated, andτ is a dummy variable for time initial-ized at the beginning of pulsation.

Krueger and Gharib (2003) showedthat the impulse created by a cylinderpiston type vortex generator was con-sistently higher than the impulse pre-dicted by the slug model. This addedimpulse was attributed to a pressuregradient at the nozzle exit plane,referred to as ‘nozzle overpressure.’Adding Ip (the impulse due to over-pressure), we get an equation for theimpulse, in terms of the nozzle pres-sure, p (which is a function of timeand radial position), and the stagna-tion pressure, p∞.

I tð Þ ¼ Islug tð Þ þ Ip tð Þ

Ip tð Þ ¼ ∫t0∫A p r; τð Þ p∞½ dAdτ ð2Þ

We performed initial testing on athruster which periodically ingestedand expelled jets with a sinusoidal ve-locity program; which was chosen forsimplicity of fabrication. Assumingthat there is no net momentum trans-fer during the ingestion phase (fluidbeing taken into the cavity starts atrest outside of the thruster and endsat rest inside the cavity) and ignoringthe pressure impulse (which will beaccounted for with a coefficient termpost analysis), then the average thrustproduced over a full cycle can be calcu-lated in terms of the thruster frequencyto be (Krieg & Mohseni, 2008)

‐T ¼ ρπ3

16D4 L

D

2

f 2 ð3Þ

Here f is the frequency of actua-tion, and the term L=D is the jet strokeratio. If the jet maintained its shape asa solid cylinder the stroke ratio wouldbe the ratio of length to diameter ofthat cylinder (see Figure 2). The strokeratio has also been called the formation

FIGURE 2

Conceptual diagram of the thruster key com-ponents. Jet shown as hypothetical slug offluid.


time since it is equivalent to the timesince initiation of the jet flow scaledby jet velocity and nozzle diameter,t* ¼ L=D ¼ ∫t0u τð Þdt=D. The for-mation time is closely related to thejet formation dynamics. When a jet isexpelled into a stationary fluid, the vis-cous forces cause the initial portion ofthe jet to roll into a tightly wound vor-tex ring. As more fluid is expelled, itfeeds the growing vortex ring untila critical saturation point is reachedand the vortex ring can no longer sup-port the added circulation. At thispoint, the vortex ring separates fromthe remaining shear flow. The forma-tion time when the jet has achieved thesame circulation as the final vortex ringis known as the formation number.Gharib et al. (1998) demonstratedthat impulsively started vortex ringshave a universal formation number(≈3.6-4.2) independent of jet velocityand diameter. However, numericalstudies have shown the formationnumber to be drastically lower forjets created with a parabolic velocityprofile (Rosenfeld et al., 1998) or a2-D jet velocity like those producedin conical nozzles (Rosenfeld et al.,2009).

The slug model predicts that theaverage thrust is proportional to boththe square of the actuation frequencyand the square of the stroke ratio/formation time. To test this assertion,the thruster was placed in a static fluidreservoir and suspended from a loadcell, and the thrust output was mea-sured directly. This testing setup isshown in Figure 3 (Krieg & Mohseni,2008). It should be noted that if thevehicle is moving during thrusterpulsation there will inherently be anon-zero momentum transfer duringingestion; however, as was previouslymentioned these thrusters are primar-ily intended for low-speed maneu-

vering (vehicle velocities well belowjetting velocities), so that the momen-tum transfer during ingestion will stillbe negligible compared to the momen-tum transfer of the jetting phase. Inaddition, the expelled jet rolls into anisolated vortex ring which inherentlyproduces a large momentum transferassociated with the fluid impulseof the vortex ring itself. During inges-tion, the small internal cavity severelylimits vortex ring formation as well asmomentum transfer associated with it.

The thruster was tested over a widerange of stroke ratios and actuation fre-quencies. As can be seen in Figure 4,the thrust shows a square proportion-ality to frequency, within a certain fre-quency range. When producing jetswith low stroke ratio, this range thisis the entire sub-cavitation frequencyrange. However, when the jet strokeratio goes above the formation num-ber, the thruster exhibits a parabolicdependence on the actuation fre-quency after a short range of square de-pendence. To more clearly show thistrend, we define a scale factor, which

is a measure of the accuracy of theslug model for various operating con-ditions α = TExp/

‐T (Krieg &Mohseni,2008) where ‐T is the slug model pre-dicted average thrust from equation (3).This scale factor is plotted with respectto frequency for stroke ratios belowthe formation number in Figure 5aand for stroke ratios above the forma-tion number in Figure 5b. Note thatthe thruster of this study has a nozzlewhich is essentially a flat plate with acircular orifice in the middle. Thistype of nozzle produces a 2-D jetflow similar to a conical nozzle, mean-ing that the jet formation number iscloser to 3 (Rosenfeld et al., 2009).For a more in depth analysis of theshift in formation number due to the2-D aspect of the jet, see Krieg andMohseni (2011).

First, consider the thrust responseof the actuator operating below theformation number (Figure 5a). In thelow-frequency regime, the scale factoris higher than that predicted by theslug model due to the nozzle overpres-sure, reaching 1.4 times the predictedvalue. Krueger and Gharib (2003)showed that the pressure impulse canreach as much as 40% of the total

FIGURE 3

Thruster static testing environment.

FIGURE 4

Thrust versus frequency. Each set of markersshows thrust data for a different stroke ratio.The error bars plotted along with the thrustrelationship represent a single standard devia-tion of the thrust data.


impulse for low stroke ratio jets corre-sponding to a total impulse 1.6 timesthe predicted slug model impulse.However, as the frequency increases,the total thrust settles on the value pre-dicted by the slug model, meaning thatin this frequency range the impulsedue to overpressure during expulsionis equal to the impulse due to “under-pressure” during refilling so that thenet impulse transfer is that predictedby the slug model. Now considerthe thruster response when operatingabove the formation number, shownin Figure 5b. Again the low-frequencyranges exhibit an added impulse dueto the nozzle overpressure; albeit to alower extent as observed by Krueger andGharib (2003). But the high-frequencyrange exhibits an added loss in thrustwith respect to the slug model pre-diction. This relative loss is seen toincrease monotonically with both ac-tuation frequency and stroke ratio.This suggests that another assumptionmade in the slug model is no longervalid when operating above the forma-tion number. We assume that this lossin model accuracy is tied into the as-sumption made that all fluid beingingested between pulsations is at restoutside of the thruster. When a jet isejected with a stroke ratio above the

formation number, some of the shearflow is left behind in the trailingwake of the leading vortex ring. Thetrailing wake has a lower momentumthan the leading vortex ring and travelsat a much lower induced velocity butstill has a forwardmomentum substan-tially larger than the surrounding rest-ing fluid. Therefore, the loss in slugmodel accuracy could be explainedby the thruster ingesting some ofthe trailing wake during the refillingphase. Figure 6 shows successiveframes from a video of the thruster’sforming jet (at a high stroke ratio)where some of the trailing wake is in-gested back into the thruster.

It should also be noted that thescale factor results are only presentedabove an actuation frequency of 4 Hz,because the thruster was designed tooperate cyclically rather than generateindividual pulsations. The 2-D na-ture of the jet created by this type ofthruster (orifice nozzle) has an addedeffect on the nozzle overpressure notseen in the frequency ranges presented.This effect is fully explained (alongwith a more in depth description ofimpulse generation) for a single pulsa-tion with constant jet velocity in Kriegand Mohseni (2011). Despite therelative magnitude of the overpressure

impulse (with respect to the momen-tum impulse, Islug), it is only observedin low actuation frequencies, which arecoupled with low thrust output. Oftenvehicle mission scenarios will neces-sitate a large magnitude thrust output(higher actuation frequencies), wherethe overpressure is cancelled out, andthe slug model provides an accuratethrust measurement.

The vortex ring formation phe-nomenon plays a key role in the jetlocomotion process in squid as well.Bartol et al. (2009) observed that squidhave two distinct swimming gaits. Inthe efficient cruising gate jets are ex-pelled below the formation number,so that the majority of the jet rolls intothe primary vortex ring. Alternatively,in threatening situations the squid em-ploys a swimming technique referredto as escape jetting; which begins withthe hyperinflation of the mantle fol-lowed by a fast contraction expelling ajet well above the formation number.

FIGURE 5

Scale factor (slug model accuracy) versus frequency for stroke ratios below (a) and above(b) the formation number. Error bars shown on data points indicate a standard deviation ofmeasured values at that frequency (also taken in the scaled space).

FIGURE 6

Successive frames of jet flow showing thethruster re-ingesting wake flow.


Presumably this behavior indicates thatthis type of jet propulsion is most effi-cient when expelling jets below the for-mation number and that jetting abovethe formation number can achieve higherthrust at the expense of fluid losses.

Therefore, all subsequent vehiclethrusters have been designed with aset diameter resulting in a stroke rationear the formation number; to achievea maximum level of thrust, while stillbeing accurately described by the slugmodel. It should be noted that the for-mation number for a jet created on amoving vehicle will not be the sameas the formation number of the jet inthe static setup. Krueger et al. (2006)showed that vortex rings formed inthe presence of a uniform backgroundco-flow have a lower formation num-ber than vortex rings formed in a rest-ing fluid. Since a moving vehicle willinherently induce a co-flow with thejetting direction, this effect must betaken into consideration. However,the reduction in formation numberis proportional to the ratio betweenco-flow velocity and jet velocity; there-fore, the low-frequency (low jet veloc-ity) pulsation will experience pinch offat an earlier formation time, but thelower-frequency pulsation is also lesseffected by the dynamics of cyclic vor-tex ring formation.

One advantage of the VRT is thatit produces a desired level of thrustalmost instantaneously (Krieg &Mohseni, 2010). Propeller style thrust-ers suffer from a rise time associatedwith reaching the static thrust levelafter initiating rotation. This rise timeis inversely proportional to the desiredlevel of thrust and can be on the orderof several seconds for low thrust levels(Fossen, 1991; Yoerger et al., 1990).VRTs also have a rise time associatedwith reaching the desired level of thrust,which is inversely proportional to the

level of thrust. However, this rise timeis an order of magnitude smaller forVRTs. The exact thrust program as afunction of time is sinusoidal, due tothe nature of the thruster; however,using several thrust data sets to averageout the dynamic component and fittingthe average thrust to a basic logarithmiccurve the mean rise time can be ob-served. The fitted curves for several op-erational frequencies, and a stroke ratioof 4.3 is shown in Figure 7.

Along with a minimal rise time, theVRT is also immune to a thrust lag seenin tunnel thrusters (Mclean, 1991),where thrust continues to be exertedon the vehicle after the thruster hasbeen terminated.

Thruster and CephaloBotEvolution

The compact thrusters used in ve-hicle testbeds to provide validation ofstatic testing have taken a wide varietyof forms. The first generation utilizeda solenoid driving mechanism and aflexing diaphragm to expel fluid (Fig-ure 8a). The solenoid driving mecha-nism suffered from reduced strokeat higher actuation frequencies as thesolenoid stroke was load dependent.

All of the subsequent iterations haveused mechanical driving mechanismsfor better flexibility in adjusting the op-eration conditions. Gen. 2 (Figure 8b)used a completely flexible cavity in afitted mold, whereas Gen. 3 and 4switched to a semi-flexible cavitywhich was reinforced to ensure con-stant diameter but allow compressionin height. Gen. 3 (Figure 8c) used acomplicated encompassing cylindricaltube cam to drive cavity compression,but mechanical complexities and reli-ability issues caused us to simplify toa basic crank shaft design for Gen. 4(Figure 8d). In the figure, this thrusteris shown with an acrylic casing to allowthe components to be seen. The vehi-cle ready thruster uses an aluminumcasing (Clark et al., 2009).

Along with the thrusters them-selves the vehicle testbeds housingthe thrusters have evolved rapidly. Fig-ure 9 shows the evolution of vehicletestbeds used to demonstrate the feasi-bility of maneuvering using VRTs.Starting with the oldest vehicle at thebottom and successive generationsupwards, the first vehicle only hadfins to provide a standard for maneu-vering capabilities, 2nd, 3rd and 4thgeneration vehicles contain 1st, 2ndand 3rd generation thrusters, respec-tively, and increasing levels of auton-omy. The lessons learned from thesevehicles directly lead to the develop-ment of the most recent hybrid classvehicle described in the followingsection.

Hybrid Vehicle DescriptionThe newest generation of vehicle

is intended to be used in autonomoussensor network applications. There-fore, the vehicle must be able to travelon long range missions collecting data

FIGURE 7

Mean thrust produced at various frequencies(static desired thrust level) versus time. Risetime inversely proportional to thrust level.


but must also be capable of auto-nomously docking with permanentsupport structures. These structureswill be responsible for downloadingthe vehicle’s mission data, changingmission objectives and recharging ve-hicle batteries.

Autonomous docking with such astructure is a complicated problem,requiring not only high-accuracy ma-neuvering, but equally high-accuracylocalization and attitude determination.In addition vehicles in this type of en-vironment need to be capable of com-municating with each other at short

distances for coordinated missions. Ingeneral reduction in cost and internalstructure are also desirable to allowfor more vehicles in the network witha wider range of payload options.

The base vehicle (Figure 10) is sep-arated into three hull sections. Thefront and back sections house all ofthe actuators. Each section has twoVRTs and an active buoyancy controldevice (BCD). Each thruster has anoverall diameter of 7.6 cm (3 inches),a nozzle diameter of 1.8 cm (0.6 inch),has a stroke ratio of 4.5, and producesclose to 2N of thrust at 30 Hz beforecavitation starts to occur in the cavity.The back section also has the rear pro-peller motor. The primary batteriesand all the electronics except motorcontrol are housed in the center sec-

tion. An additional payload sectionmay be added between front and cen-ter to include additional sensors,devices, and/or batteries. Regulatedpower and digital communicationlines interface the payload to the cen-ter section. When put together, the0.15 m (6 inches) diameter, 0.92 m(36 inches) long vehicle weights a neu-trally buoyant 16 kg (36 lb). TheBCDs can change the buoyancy by±1% to dive or surface, and 2 kg ofinternal ballast are adjustable to bal-ance the pitch of the vehicle.

Communication/Localization System

Due to the physical properties ofwater, RF communication methods

FIGURE 9

The first four generations of our vehicle testbeds. Oldest vehicle at bottom, and successivevehicles placed in ascending order.

FIGURE 10

Fifth generation hybrid vehicle CephaloBot.

FIGURE 8

Successive generations of thrusters (a) utilize solenoid driver (7.5 cm/3-inch diameter), (b) utilize cavity and mold (10 cm/4-inch casing diameter),(c) have encompassing driving mechanism (12.5 cm/5-inch plate diameter) and (d) utilize simple crank shaft and semi-flexible ducting (10 cm/4-inchplate diameter).


are not practical for small unmannedunderwater vehicles. CephaloBot hasa joint acoustic communication andlocalization system which is ideallysuited for underwater sensor networkapplications. The communicationtechnique is based on binary frequencymodulation whereas the localizationmethodology is based on a time delayof arrival technique. The system con-sists of a specialized hydrophone array(fabricated in house as described laterin this section), which interpret acous-tic signals for both information anddirectional content. The receiving hy-drophone array consists of three piezoelectric ceramics spaced in a triangulararrangement parallel to the vehicleprinciple axis plane (Figure 11). Twopiezo electric ceramics are placed on aline parallel to the pitching axis, andthe 3rd is extended along the roll axisfrom the midpoint of the other two.The entire array is encased in urethanerubber with an acoustic impedancesimilar to water. The transmittingnode consists of a single high-powerpiezo electric ceramic encased in thesame type of urethane rubber as the

hydrophone array. Each vehicle isequipped with both a receiving andtransmitting node on the undersideof the vehicle (see Figure 12).

All three hydrophones in the re-ceiving node receive a signal from atransmitting node (either on anothervehicle or a docking station). Thephase lag (ϕ) between the signal com-ing from the hydrophone on the rollaxis and the signals coming from thetwo hydrophones on the pitching axisis measured and correlated with thesignal frequency ( f ) to determine thetime between when the hydrophonesreceived the source signal Δti = ϕi/f.The time lag is then multiplied bythe speed of sound in water to getrelative source distances in the vehicleframe and transformed into the inertialframe to get the azimuth and elevationof the receiving node with respect tothe source node.

The hybrid localization/commu-nication system is comprised of threemain phases: data sending, data receiv-ing and localization. Data receiv-ing and localization both use thesame incoming acoustic wave, thefirst 1000 cycles are dedicated to local-ization and the remaining cycles arefrequency modulated. The acousticwave propagates from a sending trans-ducer located at some other node and

propagates towards the vehicle. Oncethis wave is received by the hydro-phone array, it is conditioned by fil-tering and amplification circuitry.The conditioning circuitry contains a17-dB gain pre-amplifier followed bytwo stages of filtering and two stagesof amplification to bring the signalvoltage to 5 V. The electrical voltageis then passed through exclusive orgates with one of the other signals (inthe case of hydrophone 0, the signal isalso passed to demodulation circuitry).During the localization portion of re-ceiving an onboard microcontrollerreads and stores the output of each ofthe three exclusive-or gates and alsodetermines which of the three signalsarrives first. The duty cycle of theexclusive-or gates directly relates tothe phase difference of the signals. Thelocalization hardware calculates andpasses the azimuth and elevation angles(which define a cone of possible vehiclelocations with the source node at thevertex) to themain navigation processorwhich uses secondary positioning sen-sors (depth sensor and electronic com-pass) to determine a unique solution tothe position of the receiving node withrespect to the transmitting node.

After the localization portion ofreceiving is complete the output ofthe exclusive-or gates are ignored andthe microcontroller reads and stores theDC voltage level from a frequency tovoltage converter which directly repre-sents the frequency of the incomingfrequency modulated signal.

In order for the localization meth-odology to function properly the hy-drophones must be placed no morethan λ/2 apart where the wavelengthλ is equal to the speed of propagationdivided by the frequency (λ = c/f ).This leads to a maximum spacing of3 cm for a frequency of 25 kHz and1.875 cm for a frequency of 40 kHz.

FIGURE 11

Definition of principle submarine axes.

FIGURE 12

Transducer nodes on the vehicle. Receivingnode on the right and transmitting node onthe left.


Miniature hydrophones which are ap-proximately 1 cm in diameter are com-mercially available from Reson but areon the order of $1000 per unit. Half-inch diameter cylindrical piezo electricceramics (SMC14H12111) are avail-able from Steminc for less than $10/each. Placement of the three half-inch diameter piezo electric ceramicsin a triangle pattern yields a maximumnavigational frequency of 25 kHz.The sending and receiving transducerswere made in house using a methodsimilar to that in Li et al. (2010).The sending piezoceramic was chosenprimarily based on its resonant fre-quency of 22 kHz and cylindricalshape to provide an omni-directionalsignal. The receiving hydrophonearray and transmitter are shown placedon the belly of the CephaloBot inFigure 12.

Overall this customized commu-nication localization system places aminimal load on the vehicle. The cir-cuitry has a very small footprint (Fig-ure 13) compared to typical commercialacoustic modems. In a two-way com-munication mode, the power draw isless than 1 W, and in simple listen-ing mode, the power draw is less than0.25 W. The transducers are fabricatedin house so they can be customized to

any shape and placed at any locationon the vehicle to improve vehicle dragcharacteristics. The entire system isfabricated for under $300 in materials,making it a suitable option for sen-sor networks requiring several lowcost vehicles.

Embedded SystemThe embedded system was custom

designed for the CephaloBot. The ve-hicle will be used by researchers forvarious underwater sensor network-ing applications and multi-vehiclecoordination. In order to enable thisunderwater network to interact witha potential aerial sensor network,CephaloBot is also equipped with RFcommunication capabilities, sparecomputation power, and the abilityto quickly add new sensors. Each vehi-cle must be robust and easy to handleand operate. The embedded system isseparated into multiple printed circuitboards (PCB). A power distributionboard handles voltage regulation andbattery charging. An interface boardconnects the onboard devices and sen-sors with the power board and process-ing device. Smaller PCBs are locatedthroughout the vehicle to providespecific functionality such as motorcontrol, user interface, or simply wirerouting. Figure 14 shows the electron-

ics located in the center section, whereeverything except the motor control-lers and user interface is located.

ProcessingThe primary processing device on

the vehicle is a National InstrumentsSingle-board RIO (sbRIO). It has anon-board 400MHz processor runningreal-time LabVIEW software, 128MBof RAM, 256 MB of flash, and a40-MHz 2 M gate FPGA (field pro-grammable gate array), also pro-grammed in LabVIEW, providing110 digital I/O pins. The combinationof microprocessor and FPGA wasproven effective in the 4th generationvehicle where a Compact RIO wasused. All of the low-level communica-tion, interface, and control tasks arehandled on the FPGA, leaving themicroprocessor open to perform highlevel mission control. The core vehiclesoftware is located on the FPGA. Theprimary benefit is that the mission crit-ical software will always run at fullspeed and safety checks will be imple-mented that a vehicle user cannot eas-ily override. In this way, even if thealgorithms being tested fail, the vehiclewill remain safe to itself and surround-ings. The 400-MHz real-time proces-sor is little used by the core system andtherefore provides significant compu-tational power to the researcher. AnFIGURE 13

Communication and localization hardware.FIGURE 14

Hybrid vehicle embedded system mounted on battery pack.


additional 2 GB of flash storage isadded with a serial data logger andcan be used for mission review.

BatteriesThe power source is a four-cell lith-

ium polymer pack. The nominal volt-age of 14.8 V ranges from 10 to 16.8 V,depending on charge level. The lithiumpolymer chemistry was selected be-cause of its high-power density andexcellent discharge characteristics. Asingle four-cell battery pack was cho-sen because, during most of its dis-charge cycle, it will have a voltagebetween 14 and 16 V, which mini-mizes the voltage change of the high-current devices. The availability ofintegrated circuits and componentsfor four-cell batteries was found to bebetter than an eight-cell approach,which would have a maximum voltageof 33.6 V and require larger 35 V tol-erant components. Overall current re-quired dictates some trace thicknessesof approximately 8 mm. The capacityof each cell is 21 Ah, which provides atotal of 310 Wh for the pack. A com-mercial frontend battery circuit boardprotects the batteries from over-charging, over discharge, and shortcircuit. It also balances the cells to in-crease overall service life. The circuithas a very low resistance to have mini-mal impact on the efficiency. Thecharger built into the power distribu-tion can charge the batteries in approx-imately 12 h. The high capacity of thebatteries allows them to be dischargedat a rate lower than 0.5 C, which in-creases battery runtime and overalllifetime (Murphy et al., 1990).

Power DistributionThe power distribution is the most

complex custom-designed circuit boardin the vehicle made into a four-layer

10 × 17 cm PCB. It regulates andmonitors the voltages to powerthe rest of the electronics on thesubmarine.

The power distribution moduleuses a Microchip PIC18F45K22 asa microcontroller supervisor. Thismicrocontroller monitors and controlsvoltages and current draws and parts orthe whole system can be shut off ifexcessive current is drawn or voltagesare not in an acceptable range. Thepresence of power sources is also mon-itored, and the microcontroller cor-rectly decides which one to use andwhether a battery requires charging.The sbRIO can signal the supervisorto enable or disable certain regulatorson the vehicle to save power (i.e., wire-less bridge is off when submerged). Afunction of the microcontroller alsoallows the vehicle to enter a “sleep”mode where everything except the su-pervisor microcontroller is turned offfor a pre-determined period of timewhich reduces the total power usageto about 1 W. The microcontrollerhas an onboard analog-to-digital con-verter and a 16-channel multiplexeris used to read all of the requiredvoltages and currents. Voltages aremeasured using a resistor-divider net-work to reduce the voltage into themultiplexer, and therefore the micro-controller to between 0 and 5 V withsome error margin in case the voltagerises to more than intended. AllegroACS714 Hall effect current sensorsare used to provide very low loss methodcurrent measurements (0.06W at 5 A).

SensorsCephaloBot incorporates a mini-

mum set of sensors needed to maintaina heading and depth underwater.Acoustics provide the vehicle witha relative position to a static pinger.

Intervehicle communication preventscollisions between vehicles and thewalls of the pool. The vehicle is robustenough to withstand a collision if itdoes occur. When the vehicles aredeployed to an ocean environment,payload sensors may be added hetero-geneously to the vehicles and shared sothat each submarine has all requireddata to successfully navigate its en-vironment. To simplify the design,CephaloBot uses an all-in-one IMUsolution from VectorNav. The devicehas an onboard three-axis accelerome-ter, gyroscope, and magnetometer. Itperforms Kalman filtering and outputsquaternions, Euler angles, and the rawsensor data to the sbRIO FPGA. Theonboard filtering eliminates the needto develop or perform the computa-tions on the sbRIO. Stated accuraciesare less than 2°, and because a magne-tometer provides an absolute refer-ence, this accuracy will not degradewith time. A Honeywell pressure sen-sor provides fine resolution (0.01 m)measurements to 10-m depths. Thedevice outputs an analog voltage. Thetest pool for the vehicle is 5-m deep,and so a higher resolution device aschosen over one that may be used toa deeper depth.

ConclusionThe CephaloBot provides an ideal

low cost option for underwater sensornetworking and hybrid vehicle appli-cations. The vehicle has maneuveringcapabilities at zero forward velocity nec-essary for docking and high-resolutionsensing. This capability is provided byan array of novel squid and jellyfishinspired thrusters. The thrusters arelocated internal to the hull with onlya small orifice exposed to the outerflow minimizing the effect on for-ward drag and allowing for efficient


high-speed transit. Additionally thesethrusters only require the singleopening, allowing for greater systemfreedom internal to the vehicle in-between thrusters.

The embedded controller systemdesigned for the vehicle has a compactmodular design allowing for a widevariety of possible mission objectives.The microcontroller, which is oper-ated on an easily adaptable LabVIEWplatform, includes several open con-nections for future mission operations,on top of the base level vehicle opera-tion input/outputs. The vehicle also hasa wide variety of communication op-tions including a low cost and in housedeveloped acoustic communication/localization system for communica-tion between underwater vehicles andsupport structures. The vehicle has anRF system for communication withaerial vehicles while on the surface,and aWIFI bridge for communicationwith testers and data loggers while incontrolled laboratory environments.


S. Lawrence-Simon, Tyler Thomas,Ryan Delgizzi, Dan Ambrosio, ColinMiller, Mikhail Kosna, and MattRhode for their hours of working ondesign and fabrication of the vehicle.We would also like to thank the Officeof Naval Research (code 34) for fund-ing this research project.

Corresponding Author:Kamran Mohseni231 MAE-A, Department ofMechanical and AerospaceEngineeringUniversity of Florida, Gainesville, FLEmail: [email protected]

ReferencesAllen, D.G. 2010. James Rumsey: Steamboat

inventor or pioneer in steam navigation.

Magazine of the Jefferson County Historical

Society. 96:48-54.

Anderson, E., & Grosenbaugh, M. 2005.

Jet flow in steadily swimming adult squid.

J Exp Biol. 208:1125-46. doi: 10.1242/

jeb.01507.

Barrett, D., Triantafyllou, M.S., Yue, D.K.P.,

Grosenbaugh, M.A., & Wolfgang, M.J. 1999.

Drag reduction in fish-like locomotion.

J Fluid Mech. 392:183-212. doi: 10.1017/

S0022112099005455.

Bartol, I.K., Krueger, P.S., Stewart, W.J., &

Thompson, J.T. 2009. Hydrodynamics of

pulsed jetting in juvenile and adult brief squid

Lolliguncula brevis: Evidence of multiple jet

‘modes’ and their implications for propulsive

efficiency. J Exp Biol. 212:1889-903.

doi: 10.1242/jeb.027771.

Bartol, I.K., Krueger, P.S., Thompson, J.T.,

& Stewart, W.J. 2008. Swimming dynamics

and propulsive efficiency of squids throughout

ontogeny. Integr Comp Biol. 48(6):720-33.

Clark, T., Klein, P., Lake, G., Lawrence-

Simon, S., Moore, J., Rhea-Carver, B., …

Wu, A. 2009. Kraken: Kinematically roving

autonomously kontrolled electro-nautic.

In: Aerospace Sciences Meeting. Orlando, FL:

AIAA.

Dabiri, J.O., Colin, S.P., Katija, K., &

Costello, J.H. 2010. A wake based correlate of

swimming performance and foraging behavior

in seven co-occurring jellyfish species. J Exp

Biol. 213:1217-75. doi: 10.1242/jeb.034660.

Dunbabin, M., Roberts, J., Usher, K.,

Winstanley, G., & Corke, P. 2005. A hybrid

auv design for shallow water reef navigation.

In: Proceedings of the International Conference

on Robotics and Automation. pp. 2105-10.

Barcelona, Spain: IEEE. doi: 10.1109/

ROBOT.2005.1570424.

Fossen, T.I. 1991. Nonlinear modeling and

control of underwater vehicles. Ph.D. thesis,

Norwegian Inst. Technol., Trondheim,

Norway.

Gharib, M., Rambod, E., & Shariff, K. 1998.

A universal time scale for vortex ring formation.

J Fluid Mech. 360:121-40. doi: 10.1017/

S0022112097008410.

Glezer, A. 1988. The formation of vortex rings.

Phys Fluids. 31(12):3532-42. doi: 10.1063/

1.866920.

Kato, N. 2011. Swimming and walking of

an amphibious robot with fin actuators.

Mar Technol Soc J. 45(4):181-97.


characterization of pulsatile vortex ring

generators for locomotion of underwater

robots. IEEE J Oceanic Eng. 33(2):123-32.

doi: 10.1109/JOE.2008.920171.

Krieg, M., & Mohseni, K. 2010. Dynamic

modeling and control of biologically inspired

vortex ring thrusters for underwater robot

locomotion. IEEE T Robot. 26(3):542-54.

doi: 10.1109/TRO.2010.2046069.


enhancement from radial velocity in squid

inspired thrusters. OCEANS 2011, in press.

Kona, HI: MTS/IEEE.

Krueger, P.S., Dabiri, J., & Gharib, M. 2006.

The formation number of vortex rings formed

in a uniform background co-flow. J Fluid

Mech. 556(1):147-66. doi: 10.1017/

S0022112006009347.

Krueger, P.S., & Gharib, M. 2003. The

significance of vortex ring formation to the

impulse and thrust of a starting jet. Phys

Fluids. 15(5):1271-81. doi: 10.1063/

1.1564600.

Li, Y., Kastner, R., Faunce, B., Domond, K.,

Kimball, D., Schurgers, C., & Benson, B.

2010. Design of a low-cost underwater

acoustic modem. IEEE ESL. 2:58-61.

doi: 10.1109/LES.2010.2050191.

Licht, S. 2008. Biomimetic oscillating foil

propulsion to enhance underwater vehicle

agility and maneuverability. Ph.D. thesis,

Massachusetts Institute of Technology,

Cambridge, Massachusetts.

Licht, S., Polidoro, V., Flores, M., Hover, F.,

& Triantafyllou, M. 2004. Design and

projected performance of a flapping foil


auv. IEEE J Oceanic Eng. 29(3):786-68.

doi: 10.1109/JOE.2004.833126.

Lipinski, D., & Mohseni, K. 2009. A nu-

merical investigation of flow structures and

fluid transport with applications to feeding

for the hydromedusae Aequorea victoria and

Sarsiatubulosa. J Exp Biol. 212:2436-47.

doi: 10.1242/jeb.026740.

Mclean, M. 1991. Dynamic performance of

small diameter tunnel thrusters. Ph.D. thesis,

Naval Postgraduate School, Monterey,

California.

Murphy, T., Cason-Smith, D., James, S., &

Smith, P. 1990. Characterization of aa size

lithium rechargeable cells. In: Power Sources

Symposium, Proceedings of the 34th Inter-

national. pp. 176-80. Cherry Hill, NJ: IEEE.

Mohseni, K. 2004. Zero-mass pulsatile jets

for unmanned underwater vehicle maneuver-

ing. In: 3rd “Unmanned Unlimited” Tech-

nical Conference, Workshop and Exhibit.

pp. 2004-3686. Chicago, IL: AIAA.

Mohseni, K. 2006. Pulsatile vortex generators

for low-speed maneuvering of small under-

water vehicles. Ocean Eng. 33(16):2209-23.

Mohseni, K., & Gharib, M. 1998. A model

for universal time scale of vortex ring formation.

Phys Fluids. 10:2436-8. doi: 10.1063/1.869785.

O’Dor, R.K., & Webber, D. 1991. Inverte-

brate athletes: Trade-offs between transport

efficiency and power density in cephalopod

evolution. J Exp Biol. 160:93-112.

Rosenfeld, M., Katija, K., & Dabiri, J.O.

2009. Circulation generation and vortex ring

formation by conical nozzles. J Fluids Eng.

131, 091204:1-8.

Rosenfeld, M., Rambod, E., & Gharib, M.

1998. Circulation and formation number

of laminar vortex rings. J Fluid Mech.

376:297-318.

Sahin, M., & Mohseni, K. 2008. The nu-

merical simulation of flow patterns generated

by the hydromedusa Aequorea victoria using

an arbitrary Lagrangian- Eulerian formulation.

In: 38th Fluid Dynamics Conference and

Exhibit. pp. 2008-3715. Seattle: AIAA.

Sahin, M., & Mohseni, K. 2009. An arbitrary

Lagrangian-Eulerian formulation for the nu-

merical simulation of flow patterns generated

by the hydromedusa Aequorea Victoria.

J Comp Physiol. 228:4588-605. doi: 10.1016/

j.jcp.2009.03.027.

Sahin, M., Mohseni, K., & Colins, S. 2009.

The numerical comparison of flow patterns

and propulsive performances for the hydro-

medusae Sarsia tubulosa and Aequorea Victoria.

J Exp Biol. 212:2656-67. doi: 10.1242/

jeb.025536.

Shariff, K., & Leonard, A. 1992. Vortex

rings. Annu Rev Fluid Mech. 34:235-79.

doi: 10.1146/annurev.fl.24.010192.001315.

Torsiello, K. 1994. Acoustic positioning of

the NPS autonomous underwater vehicle

(AUV II) during hover conditions. Master’s

thesis, Naval Postgraduate School, Monterey,

California.

Wilson, M.M., Peng, J., Dabiri, J.O., &

Eldridge, J.D. 2009. Lagrangian coherent

structures in low Reynolds number swimming.

J Phys-Condens Mat. 21. Art. 204105.

doi: 10.1088/0953-8984/21/20/204105.

Yoerger, D., Cooke, J., & Slotine, J.J. 1990.

The influence of thruster dynamics on under-

water vehicle behavior and their incorporation

into control system design. IEEE J Oceanic

Eng. 15(3):167-78. doi: 10.1109/48.107145.


P A P E R

Modeling of Artificial Aurelia auritaBell DeformationA U T H O R SKeyur B. JoshiAlex VillanuevaColin F. SmithShashank PriyaCenter for Energy HarvestingMaterials and Systems,Center for Intelligent MaterialSystems and Structure,Virginia Polytechnic Instituteand State University

A B S T R A C TRecently, there has been significant interest in developing underwater vehicles

inspired by jellyfish. One of these notable efforts includes the artificial Aurelia aurita(Robojelly). The artificial A. aurita is able to swim with similar proficiency to theA. aurita species of jellyfish even though its deformation profile does not completelymatch the natural animal. In order to overcome this problem, we provide a system-atic finite element model (FEM) to simulate the transient behavior of the artificialA. aurita vehicle utilizing bio-inspired shape memory alloy composite (BISMAC) ac-tuators. The finite element simulation model accurately captures the hyperelasticbehavior of EcoFlex (Shore hardness-0010) room temperature vulcanizing siliconeby invoking a three-parameter Mooney-Rivlin model. Furthermore, the FEM incor-porates experimental temperature transformation curves of shape memory alloywires by introducing negative thermal coefficient of expansion and considers theeffect of gravity and fluid buoyancy forces to accurately predict the transient defor-mation of the vehicle. The actual power cycle used to drive artificial A. aurita vehiclewas used in the model. The overall profile error between FEM and the vehicle profileis mainly due to the difference in initial relaxed profiles.Keywords: autonomous undersea vehicle, Aurelia aurita, BISMAC, finite elementanalysis, transient dynamics

Introduction

J ellyfish have been in existencefor millions of years and are the ear-liest known metazoans that use mus-cles for swimming (Valentine, 2004).They are found at various oceandepths and possess the ability to sur-vive under hostile ocean environment.They exhibit colonial behavior andhave the ability to maintain certaindepth and certain distance from theocean shore (Albert, 2009). Jellyfishhave relatively simple biological formand muscle architecture, lacking ad-vanced sensors and a complex neuralnetwork possessed by many oceaniccreatures (Chapman, 1974; Gladfelter,1972, 1973). but they are still ableto survive and adapt in hostile envi-ronments. It maintains territorial exis-tence by swimming onminimal energyintake. These abilities have createdtremendous interest in the scientificcommunity to discover their structure-property-performance relationshipsand apply the learning towards creat-ing a jellyfish-inspired swimming ve-

hicle to perform various surveillanceand monitoring tasks.

There have been various effortsin literature on developing jellyfish-inspired robots by using smart material-based actuators. Inspired by the jetterclass of jellyfish that swim by creatinga jet of water by forcing it out of thebell, Villanueva et al. (2009) developedthe JETSUM. This prototype usedshape memory alloy (SMA) wires andcreated an actuating stroke of bellsegments attached to a passive neu-trally buoyant bell structure. Yanget al. (2007) used flappers with controlsurfaces made of ionomeric poly-mer metal composites (IPMC) tocontrol directionality of the vehicle.Tadesse et al. (2010a) used polypyrrole–polyvinylidene difluoride compositesto achieve bending actuation to create

a jellyfish robot. Larger jellyfish typi-cally use “rowing” locomotion (Colin& Costello, 2002) and are character-ized by formation of counter rotatingstarting and stopping vortex rings. In-teraction of these vortex rings reducesenergy lost in the wake and lendsrowers their superior swimming ef-ficiencies relative to jetters (Colin &Costello, 2002). Thus, with regard toenergy efficiency, rowers provide a bet-ter platform for larger vehicles. Yeomand Oh (2009) proposed an entirejellyfish made from IPMC actua-tors with segments cut such that, oncontraction, all the segments close toform a contracted bell shape. Recently,a significant breakthrough was madeby Villanueva et al. (2010b), who pro-posed the high-energy density bio-inspired shapememory alloy composite


(BISMAC) actuator that opened thepossibility of converting high-force gen-eration capability of SMA wires intohigh displacements. Using BISMAC,the design and implementation ofbiomimetic rowers became feasible.

In this study, we investigate thebell deformation of BISMAC-basedjellyfish robots using a finite ele-ment model (FEM, conducted usingANSYS) and identify the correlationwith the natural species. In order todo so, the first major challenge was aprecise implementation of SMA inFEM due to their giant aspect ratioand hysteretic temperature transfor-mation. We were able to successfullydemonstrate the deformation of SMAusing ANSYS by optimizing the mesh-ing technique, identifying the variabil-ity in thermal coefficient of expansion,and separating the total deformationcycle into individual heating and cool-ing curves. Building upon this success,we implemented the SMA in BISMACstructure and investigated its mechan-ics to optimize the BISMAC config-uration for mimicking the jellyfishprofile. The FEM model provides theunderstanding of mechanism for bend-ing strain amplification in BISMACactuators and clearly delineates theeffect of structural and thermal vari-ables. Using the FEM results, wewere able to identify the inaccuraciesthat can occur due to variability inprototyping of BISMAC. Further-more, our results provide importantinsight towards the development offeedback controller based on resistancechanges. Next, we introduce the de-sign of bell geometry in FEM for arti-ficial Aurelia aurita (later referred to as,A. aurita for convenience in this work)using radial arrangement of BISMACactuators. The objective was to developthe proper joint geometry that allowsBISMACs to provide maximum defor-

mation. Next, we describe the method for fabricating artificial A. aurita andexperimental characterization. Lastly, using the FEM simulations, we presenta comparative analysis between the biological A. aurita (swimming profile)and artificial A. aurita and FE simulation.

BISMAC ActuatorThe BISMAC actuator is a composite of an incompressible flexible metal strip

and SMA wires separated by distance d and embedded in silicone rubber. Ther-mal transformation from martensite phase into austenite upon Joule heating in-duces the contraction of SMA wires, which is resisted by the incompressible metalstrip. This introduces a tensile force f in the SMA that opposes contraction andreduces the strain in SMA wire by a small amount.

Figure 1 helps in explaining the mechanics of BISMAC bending de-formation (Figure 1(a)). Under Joule heating, as SMA transforms frommartensiteto austenite phase and tends to contract, but due to the structure of the BISMAC,the metal strip resists this contraction and induces tensile force f in the SMA wireand compressive force f of the same magnitude in the metal strip. This forcecouple being distance d apart generates effective moment in the BISMAC actuatorcausing it to bend. Figure 1(b) shows the deformed BISMAC geometry. Themechanics of BISMAC has been discussed in detail by Smith et al. (2011). Byusing constitutive relation for SMA, we can express the generated stress as

σ σ0 ¼ ESMA ɛ ɛ0ð Þ þ Ω ζ ζ 0ð Þ þΘ T T0ð Þ ð1Þ

where σ is the stress, ɛ is the strain, ESMA is the effective Young’s modulus ofSMA, Ω is the transformation coefficient, ζ is the martensite fraction, Θ is thethermoelastic coefficient, and T is the temperature. Subscript zero denotes theinitial state. Since transformation here is from fully martensite to fully austenitephase, ζ = 0, and ζ0 = 1. Pure thermoelastic expansion is negligible, and initialstress and strain states are zero. Using ESMA = Eaustenite , since SMA is completelyin austenite phase, Eq. (1) transforms into

σ ¼ Eausteniteɛ Ω ð2Þ

FIGURE 1

(a) Schematic of the force and moment in BISMAC. (b) Geometry of beam curvature.


If there was no BISMAC structure to resist the SMA contraction, σ = 0 and ɛ = ɛltransformation strain, providing

Ω ¼ Eausteniteɛ l ð3Þ

If there is a uniform tensile force applied onto the SMA wire while transformationtakes place,

σ ¼ Eausteniteɛ þ Eausteniteɛ l ð4Þ

Since SMA wires are thin with negligible bending stiffness, it is safe to assumeuniform distribution of axial stress. Thus, axial stress can be written as

σ ¼ fASMA

¼ ɛ f Eaustenite ⇒ ɛ f ¼ ɛ þ ɛ l ð5Þ

where f is the force generated by the SMA wire, ASMA is the total area of SMAcross section, and ɛf is the force induced tensile strain. If SMA length after con-traction is reduced from L to L′, from kinematics consideration, we can write

L0

L¼ 1þ ɛ ¼ R d

Rð6Þ

where R is the radius of curvature of the metal strip and d is the distance betweenSMA wire and the metal strip. Solving Eq.(5) and (6),

1þ ɛ f ɛ l ¼ 1 dR

ð7Þ

R ¼ dɛ l ɛ f

ð8Þ

Using Euler beam theory,

MI¼ E

R⇒

fdI¼ E ɛ l ɛ f

d

ð9Þ

ɛ f EausteniteASMAdI

¼ E ɛ l ɛ f

d⇒ ɛ f ¼

EI ɛ l ɛ f

EausteniteASMAd 2ð10Þ

ɛ f ¼ EIɛ lEausteniteASMAd 2 þ EI

ð11Þ

where EI is the total bending stiffness of the composite beam. For the particularconfiguration of the BISMAC used by Smith et al. (2011), Figure 2(a) shows the

relationship between the radius ofcurvature for the BISMAC and thedistance between SMA wires and flex-ible metal strip. Figure 2(b) revealsthat as the distance d decreases, thetensile force induced in SMA wiresincreases dramatically and attains a lim-iting value of 80 g force for 100-μm-thick SMA wire (BioMetal Fiber,TOKI Corporation) beyond whichhigh stresses cause austenite phase totransform into stress-induced martens-ite resulting in loss of transformationstrain and thus loss of performance,which is not accounted by Eqs. (1)-(11).

BISMAC Customizationand Bell Geometry

A. aurita curvature profiles in itsrelaxed and contracted states areshown in Figure 3(a) (Dabiri et al.,2005). Before any actuation, the bellis fully expanded and said to be inthe relaxed position. Fully contractedstate refers to the state correspondingto complete contraction of subumbrel-lar muscles andminimumbell volume.After actuation, the bell passivelyregains its original (relaxed) position.In earlier work (Villanueva et al.,2010b), we have presented the meth-odology used to customize BISMACconfigurations to mimic the naturalA. aurita curvature profile. For acurved beam, we can write

M ¼ EI sð Þ d Δθð Þds

ð12Þ

where M is the moment, EI(s) is thelocal bending stiffness at location s,Δθ(s) = θ(s) − θ0(s) is the change inslope at location s, and s is locationon curved profile length. Since theforce generated by SMA wires is uni-form, we ensure a constant momentM by maintaining the fixed distance


between SMA wires and the metalstrip. The bending stiffness was varied(i) to match jellyfish exumbrella andsubumbrella profiles in the relaxedstate and (ii) to manipulate the bend-ing stiffness EI(s) such that upon SMAactuation, d Δθð Þ

ds matches that of realA. aurita in order to ensure that wehave a good match between naturaland artificial vehicles in the contractedstate. The resultant BISMAC achievedclose similarity with A. aurita profile as

illustrated in Figure 3(b). The inflex-ion point in contracted A. aurita pro-file at ∼2.5 cm represents the fact thatat the center bell thickens and the pro-file becomes a little convex near thecenter and changes to concave at in-flexion point. Towards the end of theprofile, the BISMAC shows reductionin curvature due to passive material atthe tip end for protection from water.Similar reductions in contractedA. aurita curvature profile towards

the bell margin lead to discovery ofthe passive flap in A. aurita near thebell margin. The passive flap wasshown to improve the swimming per-formance of the artificial A. aurita sig-nificantly (Villanueva et al., 2010a).Figure 3(c) shows a schematic of ar-tificial A. aurita consisting of eightBISMACs radially distributed aroundthe bell, which is made of soft silicone.Artificial A. aurita has shown bell de-formation and kinematics as well as aswimming performance comparableto that of natural animal (Villanuevaet al., 2010a). Since most of the avail-able engineering materials have lowercompliance compared to that of natu-ral animal, the artificial A. aurita de-sign includes joint structures betweenBISMAC actuators to localize the ma-terial folding upon contraction. Thesejoint structures modify the originalaxisymmetric A. aurita bell shape andincreases similarity to theCyanea capil-lata bell shape. The joints are wedge-like cavities and are found on naturaljellyfish. The joint structure is de-scribed by Smith and Priya (2010)and is copied from the Polyorchis mon-tereyensis. Since the cross-sectionalshape of the joints varies with bellheight, we take the equation below torepresent the profile at midbell. Thejoint structure can be described as apiecewise function representing twosymmetric sides of a single function,mirrored about the y-axis,

y ¼ δ j k xð Þδk jx

ð13Þ

This function has been formed to de-scribe the joint shape across a 2-Dplane with height j, half-width k,and curvature δ as parameters. In thisway, the joint structure from a widevariety of species can be describedwith one basic equation by changing

FIGURE 3

(a) A. aurita profile in relaxed and contracted conditions. (b) Curvature comparison of BISMACmuscle with A. aurita after customization. (c) Schematic of the BISMAC placement in the artificialA. aurita.

FIGURE 2

(a) Radius of curvature vs. distance d. (b) Tensile force in the SMA wires vs. distance d.


the constants. Artificial A. aurita’sjoint structure was produced by usinga program that could manipulate theconstants in Mathematica. The valuesto match the shape of joints in P. mon-tereyensis were found to be height =2.15, half-width = 3, and curvature =

−2.75. The final shape was evolvedby sweeping circular section at belltip into the curve at mid-bell heightjoint section for a smooth blending.

Figure 4 shows the evolution oforiginal axisymmetric A. aurita bellshape into the final bell design of arti-

ficial A. aurita. Figure 5 shows the de-tails of the unigraphics model that wasused in finite element (FE) simulation.

Experimental SetupArtificial A. aurita consists of a

central mount that houses the electri-cal circuitry and clamps the BISMACactuators together (see Figure 6). Theradius of this hub was 25% of the belldiameter and covers a region whereminimal deformation is expected tooccur in both the natural and artificialA. aurita. Since the distance betweenthe SMA wires and the metal strip isvery crucial parameter in BISMAC de-sign, the SMA wires and the metalstrip are slide in position in smallacrylic supports, designed to maintainthis distance. The supports are thenplaced in the mold and silicone ispoured in to settle (Villanueva et al.,2010a). The actual vehicle has asmall portion (3-4 mm) of metal stripand SMA wires protruding out of thebell geometry. This is neglected in thiswork for model simplification. Thesimplification is justified by negligiblebending moment contribution fromthe protruded part towards the belldeformation. The experimental dataused for the deformation comparisonwere acquired by using the same artifi-cial A. aurita and experimental setup as

FIGURE 4

(a) Original axisymmetric A. aurita bell (b). Schematic of joint geometry (Smith & Priya, 2010).(c) Top and side view showing location in bell for which joint calculations were made. (d) Finalartificial A. aurita bell shape.

FIGURE 5

Unigraphics model of the artificial A. aurita.

FIGURE 6

Artificial A. aurita with uniform bell and flap,in the relaxed configuration.


described in previous work (Villanuevaet al., 2010a). Artificial A. aurita wassubmerged underwater and clampeddown by supports pressing on the topand bottom of the bell. The contactingarea between supports and bell cov-ered the region of the internal centralmount. This region is meant to un-dergo negligible deformation sinceBISMAC actuators do not directly de-form the bell at that location. The bellwas contracted by heating the SMAwires using a rapid heating control al-gorithm (Villanueva & Priya, 2010a).This controller uses SMA resistancefeedback to monitor the bell state ofdeformation. It sends high-currentpulses for rapid contraction and lowcurrent to maintain deformationallowing fast contraction while mini-mizing power consumption. The con-troller was developed in LabView(National Instruments), and the mea-surements were made using a NIcDAQ 9172 with NI-9215 andNI-9263 analog input and outputcards, respectively. A NF HAS 4052power amplifier was used to amplifythe DAQ output and actuate therobot. The bell profile was recordedduring the first actuation cycle. Thedeformation was captured using anIN250 high-speed camera from FastecImaging. The bell deformation wastracked by placing reflective beadsalong the profile and by processingthe images manually using ImageJ.This process included an error on theorder of ±1 cm.

FEM Setup and SolutionMaterial Properties

To ensure accuracy of the model inadequately representing the behaviorof various materials, we determinedthe properties experimentally. Thereare three materials critical to the simu-

lation: silicone matrix, flexible but incompressible metal strip and SMA wires(BioMetal Fiber).

Silicone RubberRoom temperature vulcanizing (RTV) silicone is a candidate material for bell

mesoglea. In addition to forming the main jellyfish body, it is also responsible formaintaining the required distance between spring steel and the SMA wires. Wehave tested several silicone rubbers with different shore hardness to evaluate theirmechanical properties. EcoflexTM (Smooth-On) with initial tensile Young’s mod-ulus of the order of 10,580 Pa was selected to construct the Artificial A. auritabecause it offered much less resistance to the BISMAC deflection. Figure 7(a)shows the tensile test, on a dog bone–shaped sample with gauge length of 7 mmand cross-sectional area of 2.45 × 2.8 mm2 for several cycles at room temperatureat 1 mm/s displacement rate up to 20 mm extension to ensure silicone propertiesdo not change with multiple loading cycles. Hysteresis is evident in the figure, andwe used the average of the two curves to generate our model. Figure 7(b) repre-sents completely defined stress-strain behavior of silicone including compressiontest data carried out on 16- mm-thick 25.4-mm diameter cylindrical specimen.

We generated a three-parameter Mooney-Rivlin model for silicone from testdata using ANSYS’s curve fitting tool, as a more conventional two-parameterMooney-Rivlin model failed to provide a good fit to the experimental data.The Mooney-Rivlin model for EcoflexTM is given by Eq. (14),

W ¼ c10 I1 3ð Þ þ c01 I2 3ð Þ þ c11 I1 3ð Þ I2 3ð Þ þ 1d

J 1ð Þ2 ð14Þ

where W is the strain energy function, I1, I2, I3 are the stretch invariants, J is thedeterminant of deformation gradient tensor, and c10 = 2307.1 Pa, c01 = −223.76 Pa,c11 = 142.83 Pa, d = 0 Pa−1 (compressibility parameter). Silicone thermal con-ductivity was measured to be 0.22 W/m K. Silicone properties used in theFEM are tabulated in Table 1.

SMA WireWe selected 100-μm diameter BioMetal fibers due to their superior perfor-

mance (Tadesse et al., 2010b). The temperature transformation of the wires wasmeasured experimentally as shown in Figure 8(a). Figure 8(b) represents the

FIGURE 7

(a) Tensile test for several cycles showing hysteresis. (b) Stress-strain curve used to modelEcoflexTM.


stress-strain relationship as measuredby tensile test to confirm the manufac-turer’s claims (Toki Corporation) thatthe wires can easily take 400 MPastress.

Other properties that were used inthe simulation are tabulated in Table 2.For temperatures other than trans-formation temperatures, the propertieswere interpolated according to mar-tensite fraction in the SMA. Martens-ite fraction was calculated based on thetemperature-strain curve.

In order to model the temperaturetransformation hysteresis of SMAwires, we defined two separate materialcurves, one for the heating profile andother for the cooling profile. Accord-

ingly, the elastic modulus EX and ther-mal conductivity KXX also followedtwo separate profiles as depicted in Fig-ures 9(a) and 9(b). Figure 9(c) showsthe artificially defined negative thermalcoefficient of expansion to achieve thetransformation strains with increase intemperature during heating and cool-ing, respectively.

Metal StripFor choosing the metal strip, two

criteria should be met: (i) incompress-ibility for BISMACmechanics to workwell and (ii) flexibility (low bend-ing stiffness) to obtain maximumdeformation with a small actuationmoment. We selected standard spring

steel (low carbon steel) as the suitablemetal strip material with propertiestabulated in Table 3.

Transient Heat Transfer ModelMeshing, Boundary Conditions,and Loads

We chose ANSYS as our simula-tion package. To simulate transientheat transfer, we built the modelby meshing the SMA, metal stripand one element thick silicone layeraround them with SOLID70 (8-nodebrick element). The rest of the siliconematrix was meshed with SOLID87(10-node tetrahedrons) and SOLID90(20-node brick elements) was used astransition element between SOLID70

TABLE 1

Silicone properties used in the FEM.

Property Value

Thermal Conductivity 0.22 W/m K

Elastic modulus Mooney-Rivlin model

Poisson’s ratio 0.49

Density 982 kg/m3

Specific heat 300 J/kg K

FIGURE 8

(a) BioMetal Fiber temperature transformation curve (As = Austenite start temperature, Af = Austenite finish temperature, Ms = Martensite starttemperature, Mf = Martensite finish temperature). (b) Stress-strain relationship of Martensite phase.

TABLE 2

SMA properties used in the FEM.

Property Martensite Austenite

Thermal Conductivity 8 W/m K 18 W/m K

Elastic modulus 28 MPa 75 MPa

Poisson’s ratio 0.33 0.33

Density 6450 kg/m3 6450 kg/m3

Specific heat 837.36 J/kg K 837.36 J/kg K


and SOLID87 elementmeshes as shownin Figure 10(a). All these elements haveTEMP degree of freedom.

We took advantage of the circularsymmetry and modeled only 1/8th ofthe bell segment with no loss of phys-ics. To reduce the meshing effortswe chose to mesh only half of the1/8th segment of the bell, reflectedthe mesh about the plane of symmetryand finally merged the nodes to createthe complete model. The boundary

conditions are depicted in Figure 10(b).Due to symmetry, the circular sym-metric sides (Figure 10(b), violet col-ored faces) of the model do not haveany temperature gradient; thus, theyare modeled as insulated boundaries.Exumbrella and subumbrella surfaces(Figure 10(b), yellow colored faces)are in contact with surrounding waterand conveys heat into the fluid, andthus, were modeled as convectiveboundaries. The heat transfer coef-ficient of both these surfaces weretaken to be 20 W/m2 K (Baker, 1972),and the ambient temperature wasfixed at 25°C. The thermal loadingresulting from electrical heating wasapplied using Villanueva et al.’s re-sistance feedback control algorithm(Villanueva & Priya, 2010a) to reducepower requirement. The internal heat-ing load was applied on the SMA wireas shown in Figure 11, which is consis-tently 1/8th of total power consumedby the vehicle’s eight segments. It con-sists of rapid heating pulses of high

FIGURE 9

(a) Variation of SMA Young’s modulus with temperature. (b) Variation of SMA thermal conduc-tivity with temperature. (c) Variation of thermal coefficient of expansion with temperature.

TABLE 3

Metal strip properties used in the FEM.

Property Value

Thermal Conductivity 47 W/m K

Elastic modulus 210 GPa

Poisson’s ratio 0.3

Density 7860 kg/m3

Specific heat 510 J/kg K

FIGURE 10

(a) Mesh detail for transient thermal model. (b) Boundary conditions.


current followed by a constant currentregime and finally reducing the magni-tude to idling minimum current gov-erned by need for resistance feedbackmeasurement. The three initial spikesseen in the curve correspond to thehigh-current impulse sent for rapid heat-ing. Multiple pulses are usually neededduring the first few actuation cyclessince the low current is not enough tomaintain the deformation. The materialsurrounding the SMA warms up andeventually the low current input isenough to maintain deformation.

Solution: Transient Thermal AnalysisWe first find the solution with

SMA wires having material propertiescorresponding to heating curve. Tocapture initial sharp temperature riseaccurately over the first 0.12 s, smalltime steps of 0.01 s were used to pro-vide sufficient time resolution. Thelow current period from 0.12 to 0.70 swas solved in 100 uniform time steps.After solving two more transient cur-rent time steps at 0.71 and 0.72 s inone time step each, we switched theSMA wire material to the one withproperties corresponding to coolingcurve. Temperatures of SMA wirescorresponding to t = 0.72 s being farbeyond Af ensures properties of boththe curves are same at this point. Fi-nally, relaxation phase of A. auritajellyfish vehicle from 0.72 to 2.00 swas solved in 200 uniform time steps.

Transient StructuralDeformation ModelMeshing, Boundary Conditions,and Loads

To simulate transient structuraldeformation, we reused the meshfrom transient heat transfer modelfor rapid model building. SOLID70(8-node brick element) of the SMA,metal strip and one element thick sili-

cone layer around them were trans-formed into SOLID185 (8-nodebrick element) with structural degreeof freedom UX, UY, UZ. Similarly,SOLID87 (10-node tetrahedrons)was transformed into SOLID187(10-node tetrahedrons) elements andtransition elements SOLID90 (20-node brick elements) were trans-

formed into SOLID186 (20-nodebrick elements) as shown in Fig-ure 12(a). Figure 12(b) displays thedisplacement boundary condition forthe model. The boundary conditionwas applied in cylindrical coordinatesystem defined using jellyfish vehiclecentral axis as the z-axis of the cylin-drical coordinate system (CSYS = 5).

FIGURE 11

Variation of internal heating load on SMA wire with time.

FIGURE 12

(a) Transient structural deformation model mesh. (b) Displacement boundary conditions. (c) Tem-perature and acceleration boundary conditions.


Both sides of 1/8th segment being circular symmetric were constrained frommoving in hoop direction (UY = 0). The line on the axis of the vehicle wasalso constrained from moving in radial direction (UX = 0). Also, to constrainany rigid body translation, we chose to constrain one node on rigid central hublying on the axis of the vehicle from moving in axial direction (UZ = 0).

We accounted for gravitation and buoyancy force of the water by applyinginertial acceleration corresponding to reduced gravitation under buoyancy fromEq. (15),

ACEL‐Y ¼ ΣNmat¼1Vmatρmat ρwaterΣ

Nmat¼1Vmat

ΣNmat¼1Vmatρmat

9:81

ms2

ð15Þ

whereACEL_Y is the inertial acceleration for the vehicle,mat is thematerial index,ρ is the density, and V is the volume. Earth’s gravitation acceleration was taken tobe 9.81 m/s2. Temperatures were applied as body forces and were read from pre-viously solved transient heat transfer model result. To account for the fluid dragwe have introduced damping by defining BETAD = 0.03. This creates a dampingmatrix [C ] = BETAD[K ], where [K ] = stiffness matrix of the FEM.

The overall FE system equation can be written as

M½ Üf g þ C½ :U þ K½ Uf g ¼ Ff g ð16Þ

Here, [M ], [C]and [K ] are displacement dependent on the mass matrix, thedamping matrix and the stiffness matrix, respectively, U is the displacementvector and F is effective force vector. The system defined by Eq. (16) is non-linear due to variable [M ], [C], and [K ] matrices.

Solution: Transient Structural Deformation AnalysisAs in transient heat transfer model, we start the modeling with SMAwires hav-

ing properties corresponding to heating curve (martensite to austenite tempera-ture transformation) for contraction phase of the cycle and during relaxation weuse material corresponding to cooling curve (austenite to martensite temperaturetransformation). In simulation, this was achieved by defining two separate SMAmaterial property curves corresponding to (martensite (M) to austenite (A) and aus-tenite to martensite) and switching frommaterial withM→ A curve to material withA→M curve after completion of contraction phase. As SMA wire is well above aus-tenite finish temperature Af , for which both the material curves (cooling and heating)have identical properties there’s no abrupt jump between these curves. We used thetime intervals as used in transient heat transfer model.

Results and DiscussionTransient Heat Transfer Analysis

Figure 13 summarizes all major results of transient heat transfer analysis. Fig-ure 13(a) shows overall temperature distribution at the end of heating phase. Itsuggests that practical region of interest is concentrated near SMA wire, and formost part, it does not change along SMA wire length. Near the edge of the bell,however, model predicts a little more temperature rise in the SMA wire compared

to rest of the area due to very thin sili-cone layer as same amount of generatedheat is absorbed by lesser siliconeavailable. Since rate of heat conductioninto silicone is higher compared to therate of heat being convected away atthe subumbrella surface, silicone tem-perature rises faster and temperaturegradient between SMA wire and sili-cone reduces, inhibiting heat conduc-tion from SMA wire to subumbrellasurface. Figure 13(b) shows tempera-ture history at a typical location atSMA center and in silicone 1 elementaway from SMA wire surface. It showstypical response of first order system toa given excitation. Initial high-currentpulses till t = 0.12 s result in fast tem-perature rise at SMA center whichkeep increasing at logarithmic ratefrom t = 0.12 s to t = 0.70 s. Temper-ature at SMA center decreases expo-nentially from t = 0.70 s to t = 2.00 sbut fails to return to starting tempera-ture of 25°C at the end of the cycle(T = 45°C at t = 2.00 s) suggestingnet heat accumulation in the model.It also suggests that we may be addingheat unnecessarily at higher rateduring low current constant heatingcycle as the temperature at the end ofhigh current pulsed heating is wellbeyond austenite finish temperatureAf . This additional heating also aggra-vates heat accumulation problem thateventually results in partial loss of per-formance of the actuator as insufficientcooling results into incomplete trans-formation into martensite phase. Tem-perature in silicone 1 element awayfrom SMA surface shows quite similartrend but has reduced temperature risepeaks during high-current pulse cycleas low conductivity dampens out thesharp peaks. Cooling cycle starts withsignificant temperature gradient be-tween SMA center and the aforemen-tioned silicone location but diminishes


exponentially fast as is evident from Fig-ure 13(b). Figure 13(c) compares tem-perature distribution near SMA lengthat typical location for key time points

t = 0.12 s (end of high-current pulseheating), t = 0.70 s (end of constantlow current heating), and t = 2.00 s(end of cooling cycle).

Transient StructuralDeformation AnalysisOverall Bell Deformation

Since, gravity and buoyancy forcesare accounted for, in the two equilib-rium positions (relaxed and contracted),where fluid pressure differential doesnot exist across subumbrella andexumbrella due to the FEM and theartificial A. aurita being held at thebell center; we can predict deforma-tion at these positions accuratelyat this location without worryingabout approximations involved influid drag. Figure 14(a) shows vari-ous views of FE simulation resultfor contracted state at t = 0.12 s.Original undeformed model is shownwith only black edges superimposedon deformed model for compari-son. Figure 14(a.1) represents bot-tom view of the contracted artificialA. aurita bell model, which qualita-tively matches the contracted bellsegments of C. capillata species (Fig-ure 14(d)) and experimental artificial

FIGURE 14

(a) FE result for artificial A. aurita bell contraction: (a.1) bottom view, (a.2) isometric view from bottom, (a.3) cross-sectional view at BISMAClocation, and (a.4) side view. (b) Biological A. aurita contraction. (c) Artificial A. aurita experimental contraction. (d) C. capillata bell segmentcontraction bottom view. (e) Artificial A. aurita experimental contraction bottom view (Villanueva, submitted).

FIGURE 13

(a) Overall temperature distribution at t = 0.70 s. (b) Temperature time history at SMA wirecenter and in silicone 1 element away from SMA surface. (c) Typical temperature distributionalong SMA wire length.


A. aurita contraction (Figure 14(e))that we have intended to model inour FE simulation. Note that due tomanufacturing imperfections not allthe bell segments deform the sameamount (Figure 14(e)). The compari-son of contracted profile with that ofC. capillata is justified as our artificialA. aurita possess joint structure andradial muscles that were inspired byC. capillata possessing similar ra-dial muscle arrangement (Gladfelter,1973). Figure 14(a.2) depicts iso-metric view of deformed model forbetter visualization. Figure 14(a.3)shows cross-sectional view taken alongBISMAC center line. Figure 14(a.4)exhibits side view of the contractedFEM also matching well with naturalA. aurita shown in Figure 14(b) andexperimental artificial A. aurita (Fig-ure 14(c)). It is evident that the bellcurves inwards at the BISMAC lo-cations radially and axially well andconfirms that the joint design pro-vided by Smith and Priya (2010) iseffective in assisting the bell defor-mation at the BISMAC locations.

This behavior was also confirmed inexperimental A. aurita deformation(Figure 14(c)).

Comparison of Deformationat the BISMAC Locationand at the Joint Location

Figures 15(a)-15(f ) reveal ra-dial displacements in the bell at theBISMAC cross-section and in the jointcross-section, respectively, at the keytime points (t = 0.12 s, t = 0.70 s andt = 2.0 s); each group uses a commoncolor legend for ease of comparison.It is obvious that the deformation isnegligible near the centre of the jelly-fish bell and maximum at the tip ofthe bell. At joint location, the bell ispractically undeformed.

Figure 16(b) plots the time historyof radial and axial tip displacement atBISMAC location, which conformsto second order system responseto step excitation. In fact, the high-power pulse heating is done at muchhigher frequency than the bell is ableto respond, thus inertia of the bellacts as low-pass filter for the pulsed

heating excitation. But as discussedin transient heat transfer results, SMAgoes through complete martensite toaustenite transformation in this pe-riod, contracting due to phase transfor-mation strain and achieves maximumdeformation of Uradial = 1.31 mmand Uaxial = 7.87 mm at t = 0.12 s.At the maximum deformation point,the inertial forces and elastic restoringforces are not in equilibrium. The bellhas a slight overshoot above equili-brium position due to inertia. Duringmoderate constant current excitationfrom t = 0.12s to t = 0.70s as tem-perature keeps rising, without anyadditional benefit to SMA transfor-mation strain the structure relaxesunder stored strain energy duringrapid contraction and after a small os-cillation around equilibrium conditionfinally settles to Uradial = 10.87 mmand Uaxial = 6.04 mm. During relaxa-tion phase from t = 0.70 s to t = 2.0 s,SMA goes through austenite tomartensite transition as temperaturedrops rapidly and returns to originalconfiguration after little oscillations as

FIGURE 15

(a-c) Deformed artificial A. aurita bell at BISMAC location at different time t = 0.12 s, 0.70 s, and 2.0 s (d-f) Deformed artificial A. aurita bell at jointlocation at different time t = 0.12 s, 0.70 s and 2.0 s.


is evident from Figures 16(b), 15(a)-15(b), and 17(a).

Figure 16(a) compares the re-laxed and contracted position of thesimulated A. aurita bell with naturalA. aurita and robotic A. aurita. Thenatural A. aurita with its muscle con-tracting in excess of 40% clearlyachieves the highest deformation,not being fully matched by eitherartificial A. aurita or FE simulation.Experimental deformation on artifi-cial jellyfish was measured throughimage processing thus only exum-

brella points are available for compar-ison. Experimental model, however,has changed its form during manufac-turing, which was designed to matchFEM in relaxed state. This results inapparent mismatch in FE predictionfrom the experimental deformation.Inspite of this mismatch, the overalldeformations are comparable. Fig-ure 16(b) shows comparison of timeresponse of the transient tip displace-ments between FE simulation and ar-tificial and natural A. aurita. It shouldbe noted that the natural A. aurita

swims freely in water moving forward,while artificial (experimental) A. auritaand the model are held at the bell cen-ter and does not move it water. Also,for the particular cycle for which datais obtained has cycle time of 1.7 s in-stead of 2 s. Natural A. aurita con-tracts and relaxes very smoothly anddoes not have any steady equilibriumposition. Artificial A. aurita achievesabout twice as much radial displace-ment as predicted by simulation butfollows similar trend of sharp rise(fall), overshooting above the equilib-rium position beyond t = 0.12 s, re-laxing to equilibrium position duringt = 0.12 s to t = 0.70 s due to storedstrain energy in the bell during con-traction cycle. In relaxation cycle, asSMA cools down and undergoes aus-tenite to martensite transition, thebell returns back to its original loca-tion slightly overshooting beyondequilibrium position before returningback to relaxed configuration. FE sim-ulation captures this entire physicsperfectly. Artificial A. aurita returnsto relaxed position slower than pre-dicted by model, suggesting thatheat gets conducted away from SMAmore slowly than we predict. How-ever, artificial A. aurita axial displace-ment being too small; gets affected byfinite resolution of image process-ing and shows inconclusive trend. FEsimulation predicts axial tip displace-ment time history similar to that ofradial tip displacement but reducedin magnitude.

Figure 17(a) exhibits the trace ofthe bell tip at BISMAC location andsuggests that bell tip does not followthe same path while relaxing to its orig-inal configuration than it did dur-ing contraction. It also illustrates theovershoot around the equilibrium po-sitions (relaxed and contracted). Tipdisplacement for the artificial A. aurita

FIGURE 17

Trace of tip displacement at BISMAC location by (a) FE simulation and (b) artificial A. auritaexperiment and (c) natural A. aurita.

FIGURE 16

(a) Comparison of the FE results with artificial A. aurita bell deformation and natural A. aurita.(b) Time history of A. aurita tip displacement at BISMAC location.


(Figure 17(b)) shows different trace patterns during contraction and relaxation aspredicted by the model. Since artificial A. aurita begins with different relaxationconfiguration, this is expected. Interaction with surrounding fluid would causethe tip displacement trace to change which are not accommodated in themodel. Oscillations in the trace of the experimental results are partly due to track-ing error. The magnitude of the oscillations is close to the predicted error from thetracking method. Figure 17(c) represents trace of tip displacement of naturalA. aurita scaled to the same bell diameter. Natural A. aurita is freely moving inwater. The direction of contraction and relaxation is quite similar to the one pre-dicted by the model (Figure 17(a)); however, relative location of the path duringcontraction and relaxation is reversed. We believe that this is caused by the fluidforces and the freely forward motion present in natural A. aurita.

For a more quantitative comparison, we calculated the curvature of the pro-files by calculating radius of circle passing through three consecutive points,

R P2ð Þ ¼ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffix1 x2ð Þ2þ y1 y2ð Þ2

q ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffix2 x3ð Þ2þ y2 y3ð Þ2

q ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffix3 x1ð Þ2þ y3 y1ð Þ2

q2 x2y1 x1y2 þ x3y2 x2y3 þ x1y3 x3 y1ð Þ

ð17Þ

ρ P2ð Þ ¼ 1R P2ð Þ ð18Þ

where R(P2) is the radius at point P2 (x2, y2), having previous point P1(x1, y1)and next P3(x3, y3). Except first and last point, radius and curvature of all pointscan be calculated by using Eq. (17) and (18). Equal interval of all the points isvery essential for this method, thus additional or fewer points were generatedfrom available FE nodes, experimental traces and A. aurita profile.

Figures 18(a) and 18(b) compare the curvature of the profile along thelength of the curved exumbrella surface in relaxed and contracted condition be-tween FE simulation, experimental and natural A. aurita. In relaxed condition,FEM curvature matches that of A. aurita. Experimentally as evident from Fig-ure 16(a) and Figure 18(a) it does not match A. aurita profile in the relaxed state.Figure 18(b) emphasizes that curvatures of FEM, experimental as well as naturalanimal, increases as the fish contracts. Artificial A. aurita does not have the samerelaxed state as the natural animal, but in contracted state it follows naturalA. aurita exumbrella deformation. It outperforms the curvature of the naturalA. aurita by achieving a maximum curvature of 67 m−1 at about 80% length.FEM and natural A. aurita achieve maximum curvature of 50 m−1 at 67% and45 m−1 at 85%, respectively, and show sharp increase at the tip, suggesting dy-namic overshoot of the passive flap that is hypothesized to help swimming per-formance. Figure 18(c) displays profile errors between various profiles andconditions,

ɛFER sð Þ ¼ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffixFEBR sð Þ xExpBR sð Þ 2 þ yFEBR sð Þ yExpBR sð Þ 2q

ð19Þ

The profile error ɛFER is defined as thedistance between correspondingpoints (x(s), y(s)) (subscript BR corre-sponds to the BISMAC location andrelaxed condition) of FEM and the ar-tificial A. aurita vehicle relaxed profilesat location s. This was calculated byEq. (19), and we observe that profileerrors between the FE simulation andartificial A. aurita experiment in re-laxed and contracted states, ɛFER andɛFEC, have similar trends (Figure 18(c)).This indicates that the difference be-tween FE simulation and artificialA. aurita is mainly due to the initial dif-ference in relaxed profiles. We hypo-thesize that if artificial A. aurita werere-designed to match the natualA. aurita’s relaxed state, the simulationwould match it more closely. The pro-file error does not increase beyond5.8mm which corresponds to 22%of the initial profile error in relaxedconfiguration.

The model developed in this studygives better understanding of the tem-perature rise in the SMA wires andtransient heat distribution in andaround it during entire cycle. It revealsthat in spite of complex geometry, heatdistribution is practically uniformalong SMA length and can be effective-ly captured by 1st order unsteady heatconduction equation. It explains deg-radation of performance of artificialA. aurita, over number of cycle dueto heat accumulation resulting frominsufficient cooling and suggestsopportunity of making the bio-inspired A. aurita vehicle more energyefficient by reducing oversupply ofheat to SMA wire and effectively con-trolling SMA temperature. The modelpredicts transient deformation behav-ior of the artificial A. aurita qualita-tively at BISMAC and fold locations,confirming effectiveness of the joindesign. It reveals that the transient tip


displacement response can be modeledas a second-order system; however, duetomismatch in initial profile, it predictsthe radial tip displacement response∼44% of the experimental curve. Themodel captures these physical aspectsof bell deformation accurately and is auseful tool to evaluate effectiveness ofdesign changes on performance of thevehicle without building one.

ConclusionWe introduce a customization pro-

cedure for BISMAC actuators tobe used as radial muscles in artifi-cial A. aurita such that we can matchthe relaxed and contracted profiles ofA. aurita at the BISMAC locations.By accurately modeling the hyper-elastic behavior of EcoFlex (Shore00-10) RTV silicone and using exper-imentally obtained temperature-straintransformation curves for SMA wires,

we provide a high-fidelity model thatcaptures most essential physics ofartificial A. aurita deformation. Themodel suggests a unique approachto overcome ANSYS limitation inmodeling SMA temperature transfor-mation by using negative thermal coef-ficient of expansion and two separatematerial curves for heating and cool-ing. This approach is generic enoughto be used in numerical modelingof the SMA temperature-dependenttransformation in any design. Thetransient heat transfer model providesbetter understanding of temperaturerise in SMA wire and heat distributionaround it during an entire contraction-relaxation cycle. This information iscrucial in designing and evaluatingSMA heating algorithm and cannotbe obtained from the prototype di-rectly. The model also reveals that inspite of the complex geometry, thetemperature distribution along the

length of SMA wires is uniform andwe could use a simple first order heattransfer model to study temperaturedistribution in and around SMAwires. It also exposes the reason fordegradation of the performance of arti-ficial A. aurita over a number of cyclesdue to heat accumulation and revealsan opportunity to make the artificialA. aurita more energy-efficient by re-ducing the amount of heat suppliedabove the austenite’s finish tempera-ture Af that does not contribute towardbell deformation. Transient structuralmodel accounts for the gravity andthe buoyancy forces and thus in equi-librium conditions (fully contractedand fully relaxed states), where fluidpressure differentials across sub-umbrella and exumbrella do notexist, we can predict the bell defor-mations fairly accurately. Transientstructural analysis captures all essen-tial behavior of artificial A. aurita atBISMAC and fold locations and con-firms effectiveness of the joint de-sign, though an exact match was notachieved due to initial profile mis-match. The model reveals that thetransient tip displacement responsecan be modeled as a second-order sys-tem; however, due to initial profilemismatch; the model predicts the ra-dial tip displacement response at∼44% of the experimental curve. Pro-file error analysis suggests that ini-tial error in simulation and artificialA. aurita relaxed profiles is a majorcause for the mismatch. The profileerror increases by a small amountwith an average of 0.00015 m (0.58%of ɛFER) and reaches a maximum of0.0058 m (22% of ɛFER). The modelcaptures these physical aspects of belldeformation physics accurately andis a useful tool to evaluate effective-ness of design changes on performanceof the vehicle without the need for

FIGURE 18

Curvature comparison at BISMAC and fold location between ANSYS, experiment, and biologicalA. aurita (a) relaxed condition, (b) contracted condition, (c) profile errors at BISMAC location-FER: between FE and experiment relaxed profile, FEC:FE and experiment contracted profile, AEC:A. aurita and experimental contracted profile, FAC:FE and A. aurita contracted profile.


time-consuming physical construc-tion. The modeling methodology isgeneric and could be used to modelother bio-inspired robots using similarconstruction technique.

AcknowledgmentThis research is sponsored by the

Office of Naval Research through con-tract N00014-08-1-0654.

Lead Authors:Keyur B. Joshi and Shashank PriyaCenter for Energy HarvestingMaterials and SystemsCenter for Intelligent MaterialSystems and StructureVirginia Polytechnic Instituteand State University310 Durham Hall, Blacksburg,VA 24061Email: [email protected];[email protected]

ReferencesAlbert, D.J. 2009. Aurelia labiata (Scyphozoa)

jellyfish in Roscoe Bay: Their spatial distri-

bution varies with population size and their

behaviour changes with water depth. J Sea

Res. 61(3):140-3. doi: 10.1016/j.seares.

2008.11.001.

Baker, E. 1972. Liquid cooling of micro-

electronic devices by free and forced con-

vection. Microelectron Reliab. 11(2):213-22.

doi: 10.1016/0026-2714(72)90704-4.

Chapman, D.M. 1974. Cnidarian histology.

In Coelenterate Biology, eds. Muscatine, L.,

& Lenhoff, H.M., 2-92. New York:

Academic Press.

Colin, S.P., & Costello, J.H. 2002. Morphol-

ogy, swimming performance and propulsive

mode of six co-occurring hydromedusae.

J Exp Biol. 205:427.

Dabiri, J.P., Colin, S.P., & Costello, J.H.

2005. Flow patterns generated by oblate

medusan jellyfish: Field measurements and

laboratory analyses. J Exp Biol. 208(7):1257.

doi: 10.1242/jeb.01519.

Gladfelter,W.G. 1972. Structure and function

of the locomotory system of Polyorchis monter-

eyensis (Cnidaria, Hydrozoa). Helgoland Wiss

Meer. 23(1):38. doi: 10.1007/BF01616310.

Gladfelter, W.G. 1973. Structure and function

of the locomotory system of the Scyphomedusa

Cyanea capillata. Mar Biol. 14(2):150.

doi: 10.1007/BF00373214.

ImageJ. Image Processing and Analysis in

Java. http://rsbweb.nih.gov/ij/. (accessed 12

July 2011).

National Instruments. LabView. http://

www.ni.com/labview/. (accessed 12 July 2011).

Smith, C., Villanueva, A., Joshi, K., Tadesse,

Y., & Priya, S. 2011. Working principle of

bio-inspired shape memory alloy composite

actuators. Smart Mater Struct. 20(1):012001.

doi: 10.1088/0964-1726/20/1/012001.

Smith, C.F., & Priya, S. 2010. Bio-inspired

Unmanned Undersea Vehicle. Behavior and

Mechanics of Multifunctional Materials and

Composites 2010, March 8, 2010 - March

11, 2010, The Society of Photo-Optical

Instrumentation Engineers (SPIE); American

Society of Mechanical Engineers. San Diego,

CA: SPIE.

Tadesse, Y., Brennan, J., Smith, C., Long,

T.E., & Priya, S. 2010a. Synthesis and

characterization of polypyrrole composite

actuator for jellyfish unmanned undersea

vehicle. 764222-11. San Diego, CA: SPIE.

Tadesse, Y., Thayer, N., & Priya, S. 2010b.

Tailoring the response time of shape memory

alloy wires through active cooling and pre-

stress. J Intel Mat Syst Str. 21(1):19-40.

doi: 10.1177/1045389X09352814.

Toki Corp. Biometal Fiber. http://www.toki.

co.jp/biometal/english/qanda.php. (accessed

12 July 2011).

Valentine, J.W. 2004. On the Origin of

Phyla. Chicago: University of Chicago Press.

Villanueva, A., Bresser, S., Chung, S.,

Tadesse, Y., & Priya, S. 2009. Jellyfish

inspired underwater unmanned vehicle.

Proc SPIE. 7287:72871G. doi: 10.1117/

12.815754.

Villanueva, A., & Priya, S. 2010. BISMAC

control using SMA resistance feedback.

In: SPIE Conference Proceedings, SPIE

NDE Conference. 76421Z-12. San Diego, CA:

SPIE.

Villanueva, A., Priya, S., Anna, C., & Smith,

C. 2010a. Robojelly bell kinematics and

resistance feedback control. In: IEEE

International Conference on Robotics and

Biomimetics (ROBIO). 1124-9. Tianjin,

China: IEEE (Institute of Electrical and

Electronics Engineers).

Villanueva, A., Smith, C., & Priya, S.

submitted. Under review, Paper Ref no.

“BB/378339/PAP/255146”. Biomimetic ro-

botic jellyfish (Robojelly) using shape memory

alloy. IOP Bioinspir Biomim.

Villanueva, A.A., Joshi, K.B., Blottman, J.B.,

& Priya, S. 2010b. A bio-inspired shape

memory alloy composite (BISMAC) actuator.

Smart Mater Struct. 19(2):025013.

doi: 10.1088/0964-1726/19/2/025013.

Yang, Y., Ye, X., & Guo, S. 2007. A new

type of jellyfish-like microrobot. In: IEEE

International Conference on Integration

Technology. 673-8. Rourkela, India: IEEE

(Institute of Electrical and Electronics

Engineers). doi: 10.1109/ICITECHNOLOGY.

2007.4290404.

Yeom, S.-W., & Oh, I.-K. 2009. A biomi-

metic jellyfish robot based on ionic polymer

metal composite actuators. Smart Mater

Struct. 18(8):085002. doi: 10.1088/0964-

1726/18/8/085002.


P A P E R

Swimming and Walking of an AmphibiousRobot With Fin ActuatorsA U T H O RNaomi KatoGraduate School of Engineering,Osaka University

A B S T R A C TWith the goal of automatic monitoring of environments along natural coastal

areas and tidal flats, researchers designed and developed an amphibious robotequipped with fin actuators called “RT-I” that mimics the locomotion of both a tor-toise and a sea turtle. Experiments were carried out using a forearm with 4 degreesof freedom, which can reproduce the walkingmotions of tortoises and sea turtles onsand, to evaluate the walking performances of a robotic tortoise and a robotic seaturtle. It was clarified that the arm for a robotic tortoise is more suitable for use onsoil compared with the arm for a robotic sea turtle. The advantages of both seaturtles and tortoises were adopted in a robotic turtle, namely, the lift-based swim-mingmode sea turtles use and the quadrupedal locomotion tortoises use. The pres-ent amphibious robot consists of four main components: (i) leg units, (ii) a controlunit pressure hull, (iii) a buoyancy adjusting device, and (iv) a fairing cover. To re-alize not only swimming motion with the combination of flapping, rowing, andfeathering, but also tortoise-like walking motion, three motors were set up at theacromioclavicular joint using a differential gear mechanism to independently pro-duce the three types of motion, and one motor was set up to produce elbow jointmotion. A buoyancy-adjusting device was installed to realize walking on land and inwater as well as swimming in shallow water. The swimming and walking perfor-mances of the amphibious robot in water were evaluated by measuring the forwardswimming speed, backward swimming speed, speed of turning, and speed of de-scending vertically as the indexes of the maneuverability of the robotic turtle, andthe walking speed and propulsive efficiency with the crawl gait for various walkingpatterns in still water and in waves.Keywords: locomotion, tortoise, sea turtle, experiment

Introduction

I t has recently been clarified thatnatural coastline areas and tidal flatsplay an important role in preservingocean environments. To protectcoastal environments, regular moni-toring of these areas is important.Monitoring has previously beendone by humans on foot or usingboats, but this work can be dangerousbecause of breaking waves and ripcurrents. Monitoring on foot is lim-ited because it cannot be carried outin deep waters, while monitoring byboat is limited to the areas accessibleby water. Automatic monitoring byan amphibious robot is therefore ex-pected to eliminate the safety threatsto human monitors and improve op-erational efficiency. However, if anamphibious robot moves by gainingtraction with screws and caterpillars,it will not be able to move about inareas such as marshes and will damagethe environments of the areas inwhich it moves. An environmentallyfriendly amphibious robot is thusneeded.

Several studies have reported thedevelopment of amphibious robots.An amphibious snake-like robot, theACM-R5 (Yamada et al., 2005), canoperate both on ground and in waterby undulating its long body. TheACM-R5 uses special paddles and

wheels mounted around its body topropel itself through water and overground in a snake-like fashion, gener-ating propulsive force that allows it toglide freely in the tangential direction.The biomimetic amphibious soft cordrobot (Wakimoto et al., 2006), whichis made of Mckibben actuators andplastic plates, can move both on theground and in water, undulating itslong body. A spinal cord model andits implementation in an amphibioussalamander robot (Ijspeert et al.,2007) were studied to demonstratehow a primitive neural circuit for

swimming can be extended usingphylogenetically more recent limb os-cillatory centers to explain the abilityof salamanders to switch betweenswimming and walking. The AQUA(Dudek et al., 2007), an amphibiousrobot, can swim and walk along theshore and on the bottom of the oceanby moving its fins. The AQUA usessix paddles, which act as control sur-faces during swimming and as legswhile walking. An amphibious walk-ing robot was developed by Tanakaand Shirai (2006) to perform a shore-line survey. It has six legs, each of


which has three joints. It was success-ful in obtaining the distribution ofground levels from land to shallowwater. An amphibious robotic turtlewas built by Low et al. (2007) to im-itate the locomotion of Cheloniidae,both in water and on land, to performvarious operations. The crawling andlift-based swimming gaits were ana-lyzed and implemented in the proto-type. However, all of these robots arenot operated in practice to monitorthe coastal environment on land andin water by using the multiple func-tions of walking and swimming.

For the field operation of an am-phibious robot, a rigid fuselage withan adequate payload is necessary sothat a control system and sensors formonitoring the environment can beinstalled.We have been studying a bio-mimetic underwater robot equippedwith mechanical pectoral fins fromthe viewpoint of high maneuverabilityunder disturbances such as waves andwater currents (Kato et al., 2006;Suzuki & Kato, 2005; Kato & Liu,2003). Taking the field operationand application of our experiences onthe biomimetic underwater robotinto account, this study focuses on anamphibious robotic turtle, which notonly can swim in the sea but alsowalk on the land to perform environ-mental monitoring of natural coastand tidal flat areas.

Turtles that can walk and swim aregenerally categorized into sea turtlesand tortoises. Sea turtles have goodswimming ability but poor walkingability because they drag their bodieson the land, which causes frictionagainst the sand. Tortoises, on theother hand, cannot swim smoothly,but they can walk better than seaturtles can because of their quadrupe-dal locomotion capability. In thisstudy, we attempted to adopt the ad-

vantages of these two turtles into a ro-botic turtle.

This paper presents (1) a descrip-tion of the walking performance ofan arm with 4 degrees of freedom(DOF) that can reproduce the walkingmotions of sea turtles and tortoisesfrom the viewpoints of mobility andterrain trafficability, (2) details of thedesign and development of an am-phibious robot with fin actuators,and (3) an evaluation of the walkingand swimming performance of therobot in a laboratory environment.Here, mobility is defined as the walk-ing performance of a vehicle de-pending on motor torque and vehicleconfiguration. Trafficability is thesoil-bearing capacity of a vehicle.

Locomotion of Tortoisesand Sea TurtlesTerrestrial Locomotion

Walker (1971) studied the walkingof a tortoise, Chrysemys picta, usingcinephotography and X-rays andclarified the relationship between thestructure of the skeleton and themove-ments of the fore and hind limbs. Thestructure of the skeleton of the fore-limb is the same as that of the hindlimb. The forelimb consists of anacromioclavicular joint with 3 DOF,a humerus, an elbow with 1 DOF, a

forearm, a wrist with 2 DOF, and ahand. Wyneken (1997) explainedthat in the locomotion of sea turtleson land, clutching movements areseen in adult Chelonia mydas, Natatordepressus, and Dermochelys coriacea,while adults of other cheloniid speciesemploy quadrupedal gaits in which di-agonally opposite feet move as a pair.Sea turtles support themselves on thecarpus and the anterior edge of thehand rather than on the palmarsurface.

To design a robotic turtle, we ana-lyzed quantitative information on themovement of the forelimb joints of acaptured tortoise, Chinemys reevesii,with the following dimensions: lengthof shell × length of forelimb × lengthof hindlimb = 230 mm × 40 mm ×40 mm. The markers and body-fixedcoordinates (x, y, z) on the tortoisewere set up as shown in Figure 1.The origin of the body-fixed coordi-nates was set on the acromioclavicularjoint. First, movies of the walkingmotions of the tortoise were takenusing two CCD cameras. The moviesfrom the top view and side view weretaken at the same time. Second, themarked points were tracked usingsoftware that computed the two-dimensional coordinates of the points.Third, the three-dimensional coor-dinates of the elbow and wrist onthe body-fixed coordinates were

FIGURE 1

Top view and side view of tortoise Chinemys reevesii.


computed. Figures 2 and 3 show thetwo-dimensional tracks in the x-yplane and x-z plane of the motion ofthe wrist and the elbow, respectively,during walking in the case of a walkingspeed of 0.13 m/s, a period of walkingof 1.3 s, and a stance phase of 0.78.Here, the stance phase denotes the frac-tion of time during which the forelimbis set on the ground during the walkingperiod.

If the forelimb is considered to bean arm, the joint angles of the forelimb

derived from inverse kinematics of thearm can be obtained. Here, an arm isassumed to consist of an acromioclavi-cular joint with 3 DOF, a humerus, anelbow with 1 DOF, a forearm, and awrist. The (x, y, z) coordinates arefixed at the acromioclavicular joint, asshown in Figure 4. The rowingmotionis defined as rotational motion aroundthe z axis. The (x′, y′, z′) coordinatesare defined as coordinates rotatedaround the z axis by the rowing mo-tion. The feathering motion is defined

as rotational motion around the y′ axis.The (x″, y″, z″) coordinates are definedas the coordinates rotated around they′ axis by the feathering motion. Theflapping motion is defined as rota-tional motion around the x″ axis.The coordinate system of the armwas set up as shown in Figure 4,where n, s, and a, denoting unit vec-tors fixed on the forearm, were set par-allel to x, y, and z, respectively, whenall of the joint angles were zero. Thejoint angles θ1, θ2, θ3, θ4 are definedas the angle of rowing motion, angleof feathering motion, angle of flappingmotion, and angle of bending of fore-arm, respectively. Lengths of the armsl1, l2 are defined as the length of thehumerus and the length of the fore-arm, respectively.

Figure 5 shows time variations ofthe joint angles. We can see that thefeathering angle varies from 5° to 49°during the power stroke from 0 to0.6 s, and the bending angle of theforelimb varies between −75° and−115° during the power stroke,which indicates that the forelimb pro-duces forward thrust by kicking theground and positioning the forearm al-most vertically on the ground. Duringthe recovery stroke from 0.6 to 1.3 s,the bending angle of the forelimbreaches −20°, which indicates that theforearm is raised from the ground. Theflapping angle does not vary muchduring the entire stroke.

Aquatic LocomotionSea turtles swim in water using

their forelimbs to provide thrust. Thesynchronous sweeping of the flippersintroduces a stable heading direction.Wyneken (1997) explains lift-basedmechanisms of thrust production inwhich the locomotor apparatus of asea turtle acts as a wing to generate

FIGURE 2

Trajectories of the elbow and the wrist in x-y plane.

FIGURE 3

Trajectories of the elbow and the wrist in x-z plane.


lift forces during large portions of thepowerstroke. Isobe et al. (2010) clari-fied that the feathering motion offore flippers of a sea turtle influencesthe thrust production by measuringthe 3-D motion of fore flippers of asea turtle in a water circulating tankand conducting numerical simulations

based on quasi-steady wing elementtheory.

Wyneken (1997) showed drag-based propulsion by a swimming semi-aquatic turtle. Diagonally oppositelimbs are protracted, then retracted to-gether, and act as paddles. The distalelements of the limbs are flexed to

minimize surface area as they arebrought forward.

Experiment on theWalking Performanceof an Arm

We constructed an arm that couldreproduce the walking motion of a tor-toise and a sea turtle to analyze thewalking performance from the view-points of mobility and trafficability.

ArmThe robotic arm we developed

makes the motions of rowing, feather-ing, and flapping, and the forearm canalso be bent. The armmoves on rails tomake these motions on sand. The an-gles of the motions of rowing, feather-ing, flapping, and bending weremeasured by four potentiometers, thedistance of the movement along therails was measured by a potentiometer,and the forces that were applied to thehand were measured by a six-axes forcesensor, as shown in Figure 6. The hu-merus and forearm are each 150 mmlong. The rowing angle, featheringangle, flapping angle, and bendingangle of the forearm vary within thefollowing ranges, respectively: ±70°,

FIGURE 5

Time variations of joint angles.

FIGURE 4

Coordinate system and definitions.

FIGURE 6

Picture of manipulator.


±70°, ±50°, and 0° to −110°. The rowingmotion, featheringmotion, flappingmotion, and bendingmotion were produced byfour motors independently. Because frictional force works between the rails and the arm, a force corresponding to the staticfrictional force was applied to the arm along the rail using a pulley and a weight. In the experiments on the walking perfor-mance of a sea turtle, an amount of force was subtracted from the static frictional force to simulate the friction a sea turtlegenerates on sand. The arm was connected to the weight by a string. Toyoura sand, which has almost constant particlediameter and known physical parameters, was used in the experiments.

Figure 7 shows the hand shapes. The shape of the end of the hand of the model simulating a tortoise (T ) is a 70 mm ×70 mm square. That of the model simulating a sea turtle (S) is a 150 mm × 125 mm rectangle.

Kinetic Relations Between External Force and Joint TorqueTo discuss the propulsive efficiency in terms of walking performance of the arm, we need to know the torque of each joint

of the arm. We then consider the kinetic relations between external force and joint torque.First, we refer to the Jacobian matrix in the kinetics of the arm. We define θ(θ1, θ2, θ3, θ4)

T as the rotational angles ofjoints. The relations between the rotational velocities of the joints,

:θ

:θ1;

:θ2;

:θ3;

:θ4

T , translational velocity of the end of the

forearm (see Figure 6), (:Pr

:x; :y; :zð ÞT ), and rotational velocities of the end of the forearm, (:Φrð

:ϕx;

:ϕy;

:ϕzÞT ), are described as

follows:

:Pr:Φr

¼ J

:θ ð1Þ

J ¼ s1× Pr P1ð Þ s2× Pr P2ð Þ s3× Pr P3ð Þ s4× Pr P4ð Þs1 s2 s3 s4

ð2Þ

where J, Pi , and si denote the Jacobian matrix, position vector, and vector of rotational direction, respectively (refer toFigure 8).

Next, we will explain the kinetic relations between the external force, F(Fx , Fy , Fz)T, moment, M(Mx , My , Mz)

T, andjoint torque, Q(q1, q2, q3, q4). The moment, Mi , that is applied to each joint is

Mi ¼ Pr Pið Þ× FþM ð3Þ

and thus the torque of each joint is

qi ¼ si ⋅Mi ð4Þ¼ si⋅ Pr Pið Þ× Fþ si ⋅M

¼ si × Pr Pið Þf g⋅Fþ si⋅M

and the relation between eternal forces, moments, and joint torque is represented using a transposed Jacobian matrix asfollows:

Q ¼q1q2q3q4

2664

3775 ¼

s1 × Pr P1ð Þ s1s2 × Pr P2ð Þ s2s3 × Pr P3ð Þ s3s4 × Pr P4ð Þ s4

2664

3775 F

M

¼ JT

FM

ð5Þ


In the present study, we discusspropulsive efficiency from the view-points of mobility and trafficability.We define propulsive efficiency,η, asfollows:

η ¼ ∫T F xvd t

∑i¼1

4 ∫T qi:θidt

ð6Þ

Here, v denotes the translational speedalong the x axis. The numerator is thevalue of the work that the arm appliesto the sand. The denominator is thevalue of the total input work by thejoints.

Comparison of WalkingPerformance of an ArmBetween a Robotic Tortoiseand a Robotic Sea Turtle

The walking performance of anarm for a robotic tortoise using thehand models (T ) was measured. Thefollowing three types of walking

motion of a robotic tortoise were con-figured with reference to the experi-mental motion analysis of a tortoisein locomotion of tortoises and seaturtles. Type (a) motion includesmovement of the arm vertically duringthe power stroke. It has a symmetrictrajectory in the anteroposterior direc-tion (see Figure 9). Type (b) motion isbased on the pulling motion duringthe power stroke. It has a larger ante-

rior trajectory component within thewhole trajectory, which is analogousto the observed trajectory of the tor-toise. Type (c) motion is based onthe kicking motion during the powerstroke. It has a larger posterior trajec-tory component within the whole tra-jectory. Figures 10, 11, and 12 showthe time variations of the joint anglesof type (a), type (b), and type (c), re-spectively. The range of movement of

FIGURE 7

Hand models.

FIGURE 8

Vectors of coordinates.

FIGURE 9

Trajectories of wrist and elbow of type (a).


the wrist along the x axis and that alongthe z axis were taken as 162 and70 mm, respectively. The range ofmovement of the wrist along they axis was taken as 103 mm. The pe-riod of motion was changed from 7.5to 15.0 s with an interval of 2.5 s.The sinkage of the end of the forearmbelow the surface of the sand was alsovaried; values of 10, 15, and 20 mmwere used. Although the sinkage of arobotic turtle changes with time ac-cording to the weight of the body,the type of arm motion and the condi-

tion of the soil, it was treated as an in-dependent parameter affecting thewalking performance of the arm inthis study so we could evaluate it to-gether with other parameters fromthe viewpoint of fundamental walkingperformance. It was controlled bya feedback controller located in thearm controller using the measured an-gles of the motions of rowing, feather-ing, flapping, and bending by fourpotentiometers, after the initial atti-tude of the arm was set up in relationto the surface of the sand.

Figure 13 shows the averaged pro-pulsive forces, Fx, in the direction ofthe x axis during one period for thethree types of arm motion with themotion period of 10 s. Figure 14shows the averaged vertical forces, Fz,in the direction of the z axis during oneperiod for the three types of arm mo-tion with the motion period of 10 s.The fact that the averaged verticalforces, Fz, for type (b) motion are neg-ative can be attributed to the posture ofthe forearm in the first half of the timeperiod. Namely, the flat plate at theend of the forearm digs into the sandby positioning the forearm anteriorlyin the first half of the time period.This action leads it to carry sand onthe flat plate, which causes a negativevertical force during one cycle. Onthe other hand, the forearm in type(a) motion is positioned almost verti-cally on the sand during one period,and the forearm in type (c) motion ispositioned posteriorly so as to kicksand. Although it is difficult to dis-criminate among the three motiontypes in terms of the propulsive effi-ciencies beyond the sinkage of 15 mm,type (a) shows the largest values forthe averaged propulsive force. The

FIGURE 10

Time variations of joint angles of type (a).

FIGURE 11

Time variations of joint angles of type (b).

FIGURE 12

Time variations of joint angles of type (c).


same relation among the three motiontypes can be applied to the averagedvertical forces. Type (a) arm motion issuitable not only from the viewpointof mobility, but also trafficability.

Because sea turtles use the anterioredge of the hand for locomotion onland, we used type (d) of walking

motion, in which the flipper moves in-clined at an angle of 45°, in the designof the robotic sea turtle. Figure 15shows the time variations of the jointangles of type (d) with the motion pe-riod of 10 s. The sinkage of the bottomof the flipper below the surface of thesand was varied among three different

levels: 10 mm, 15 mm, and 20 mm.The walkingmotion of a sea turtle gen-erates friction on the sand because theturtle drags its body on the land as itwalks. Tomodel this situation, frictionon the carriage of the manipulator wasadded. We compared the walking per-formance of these two types of motionunder the conditions including fric-tion on the carriage of the manipulatorset at the following levels: 2.94, 4.90,and 6.86 N.

Figure 16 shows the averaged pro-pulsive force, Fx, in the direction of thex axis during one period against sink-age of the end of the forearm belowthe surface of the sand and frictionon the carriage of the manipulator(the key to the symbols expresses fric-tion on the carriage of themanipulator).As the friction on the rail increases, theaveraged propulsive force, Fx, also in-creases. Figure 17 shows the averagevertical force during one period againstsinkage. As the sinkage increases, themaximum vertical force, Fz, alsoincreases. The dependency of themaximum vertical force, Fz, on the

FIGURE 15

Time variations of joint angles of type (d).

FIGURE 13

Averaged propulsive force Fx in the direction of the x-axis during oneperiod for the three types of arm motions against sinkage of the endof the forearm below the surface of the sand using the hand model (T).

FIGURE 14

Averaged vertical force Fz in the direction of the z-axis during one period for the three types of armmotions against sinkage of the end of the forearm below the surface of the sand using the handmodel (T).


friction on the rail is clear. When weconsider the walking motion of a ro-botic sea turtle on the beach, frictionbetween the sand and the body de-pends on the vertical force subtractingthe vertically upward force from theweight of the robotic sea turtle.

When comparing type (a) walkingmotion of a robotic tortoise, whichshowed the best performance amongthe three types, with type (d) walkingmotion of a robotic sea turtle, it can beseen that the propulsive efficiency oftype (d) is influenced by friction andthe averaged vertical force acting onthe hand plate in the case of the armfor a robotic tortoise is larger than inthe case of the arm for a robotic sea tur-tle. An arm for a robotic tortoise is suit-able for moving heavy payloads over avariety of terrains. To discuss the traf-ficability of the arm for a sea turtle, wedefine M as the mass of the body, g asgravitational acceleration, k as the fric-tional coefficient (Setouchi & Shinjo,2001) between the soil and the ventralsurface of the robotic sea turtle, Fxm as

themean value of propulsive force, andFzm as the mean value of vertical forceagainst the soil. Assuming that a pair offore flippers and a pair of rear flippersexert thrust forces simultaneously, weobtain the following equation of staticequilibrium for walking on soil:

M⋅g − 4⋅Fzmð Þ⋅k ¼ 4⋅Fxm ð7Þ

If we assume Fxm = Fzm for thetype (d) arm motion based on Fig-ures 16 and 17, we obtain the follow-ing relation:

Fzm ¼ M⋅g ⋅k

4 1þ kð Þ < M⋅g=4 ð8Þ

If we set the length of the humerus as l1and the length between the elbow jointand center of the hand plate as l2 usingthe arm structure shown in Figures 6and 8, the torque around the x axisand that around the z axis are Fzm ∙(l1 + l2). For a robotic tortoise using acrawling gait with slow speed, we obtainFzm = M · g/4 during the time when

the 4 feet are placed on ground. If weassume that the robotic turtle usestype (a) arm motion and that Fxm isnearly equal to Fzm/25 based on Fig-ures 13 and 14, we obtain the torquearound the x axis of Fzm∙l1 and atorque around the z axis of Fzm · l1/25.The total torque on the arm of thesea turtle robot, QS, is expressed asfollows:

QS ¼ 2⋅M⋅g⋅k

4 1þ kð Þ ⋅ l1 þ l2ð Þ: ð9Þ

The total torque on the arm of thetortoise robot, QL, is expressed asfollows:

QL ¼ 26=25⋅M⋅g=4⋅ l1 ð10Þ

The ratio of QS to QL γ is

γ ¼ 2513

⋅k

1þ k⋅ 1þ ℓ2

ℓ1

ð11Þ

Because the friction ratio of sand(Setouchi & Shinjo, 2001) is in the

FIGURE 17

Averaged vertical force Fz in the direction of z-axis during one periodfor type (d) of arm motion against sinkage of the end of the forearmfrom the surface of sand using the hand model (S).

FIGURE 16

Averaged propulsive force Fx in the direction of x-axis during one pe-riod for type (d) arm motion against sinkage of the end of the forearmbelow the surface of sand using the hand model (S).


range of 0.4-0.6 and l2/l1 is larger than1.0 for a sea turtle, we find that γ tendsto approach 1.0 if k becomes smaller,and that γ tends to become largerthan 1.0 if k becomes larger. Becausesmaller torques acting around thejoints are needed from the viewpointo f the mechan i ca l de s i gn o f aturtle robot, the arm for the robotictortoise is suitable on soil from theviewpoint of trafficability.

Design and Developmentof an Amphibious RobotWith Fin Actuators

A previously designed mechanicalpectoral fin (Kato & Liu, 2003) has adrag-based swimming mode and a lift-based swimming mode. The former ischaracterized by the rowing actionforming a high angle to the horizontalaxis of the body, while the latter ischaracterized by the flapping actionforming a small angle to the horizontalaxis. It was revealed through the opti-mization of motion of the mechanicalpectoral fin that the lift-based swim-ming mode rather than the drag-based swimming mode is suitable forgeneration of propulsive force in uni-form flow, while the drag-based swim-ming mode rather than the lift-basedswimming mode is suitable for genera-tion of propulsive force in still water.To realize both the drag-based swim-ming mode and the lift-based swim-ming mode, a combination of rowingmotion, flapping motion, and feath-ering motion is needed. The hydrody-namic characteristics of the drag-basedswimming mode and the lift-basedswimming mode are discussed in de-tails elsewhere (Suzuki et al., 2007).Walking locomotion using fin actua-tors is of two types, imitating the mo-tion of a tortoise and that of a sea

turtle. Sea turtle-like walking motionhas the following characteristics:

It is dynamically stable.Terrain condition strongly affectsthe attitude of the body.Friction between the soil and thebody necessitates additional thrustforce.

Tortoise-like walking motion has thefollowing characteristics:

It is dynamically more unstablecompared with the sea turtle-likewalking motion.The attitude of the body has moreDOF compared with the sea turtle-like walking motion.

The body weight is vertically appliedto the foot.

In this study, we adopted thetortoise-like walking motion, whichhas better trafficability on soil thanthe sea turtle-like walking motion, asdiscussed in the previous section.

The amphibious robot we designedconsists of the following four maincomponents (see Figure 18):

(1) leg units,(2) a control unit in a pressure hull,(3) a buoyancy adjusting device in apressure hull, and(4) a fairing cover

Each pressure hull was designed to re-sist the pressure at the water depth of10 m. We used “Solidworks” for 3-DCAD software to design the hard-ware and “LabVIEW” to developthe control software. Four DOFs areneeded to realize not only threetypes of swimming motion but alsothe tortoise-like walking motion.Therefore, three motors were set upat the acromioclavicular joint using adifferential gear mechanism to inde-pendently produce flapping motion,rowing motion, and feathering mo-tion, and one motor was set up at

FIGURE 18

Components of Robotic Turtle (RT-I) and fin actuator with 4 DOF.


the elbow joint to produce bendingmotion of the forearm (see Figure 18).Motors and reduction gears wereselected according to the simulationresults on walking. An open dynamicengine of the kinetics calculation li-brary with open sources was used forthe simulation. A fuselage with the di-mensions (W × H × L = 0.50 × 0.20 ×0.80 m) was used. The length of thehumerus was set at 0.25 m, and thelength of the forearm was set at0.15 m. There are various walkinggaits for quadrupedal locomotion.We selected a crawling gait for lowspeed where the legs are lifted up oneby one. The fin actuators of thisrobot have two functions, walking andswimming, so we had to design theshape of the fin with both walkingand swimming in mind. For swim-ming, it is desirable to have a largefin area. However, the fin may inter-fere with the base of the leg unitduring walking. The fin shape is de-signed to minimize the interference.Figure 19 shows the form of the finwith the root chord of 0.195 m, thespan of 0.15 m, and the maximumthickness of 0.32 m.

Figure 20 shows the outline ofthe control unit’s electric circuit. Thecontrol unit consists of a CPU,motor drivers, an azimuth sensor,

three-axes rate gyros, a pressure sensor,a GPS, and related minor parts. Bat-teries are used separately for motorsand the CPU with sensors. Nickelhydride batteries with the capacity of13.2 V, 8.0 Ah are used for motorsby arranging a series circuit of twoand a parallel circuit of two. It is alsopossible to provide the control unitwith electric power from outside.

The amphibious robot has to real-ize walking in which the buoyancy isless than the weight and swimmingin which the buoyancy is equal to orgreater than the weight, because themission of the robot includes travelin shallow water with breaking waves.To realize these capabilities, the robotmust be equipped with a buoyancyadjusting device. The buoyancy ad-

justing device with the buoyancy ca-pacity of ±0.35 kg was designed usinga pair of pistons arranged symmetri-cally in the longitudinal direction soas not to have an effect on the attitudeof the robot.

It is desirable to cover the bodyof the robot with streamlined fairing.The general-purpose computer fluiddynamics software “FLUENT” wasused to estimate the hydrodynamicdrag on the fairing cover and to designa fairing cover with low hydrodynamicdrag. The cover was made of glassfiber-reinforced plastics.

Figure 21 shows a photograph ofthe amphibious robot equipped withfin actuators, named “RT-I.” Theprincipal dimensions are shown inTable 1.

FIGURE 19

Form of fin.

FIGURE 20

Electric circuit of the control unit.


Swimming and WalkingPerformance by theRobotic Turtle RT-ISwimming Performance

A towing tank test was carried outto measure drag forces acting on thebody of the robotic turtle in the towingtank of Osaka University. The forcesapplied to the body were measuredby the force sensor at various towingspeeds. From this experiment, the rela-tionship between the drag forces andthe fluid velocities was obtained to es-timate the thrust forces produced bythe fins. The swimming speeds infree swimming condition of the ro-botic turtle were measured in thesame towing tank for the estimationof the swimming performance. Thel i f t -based swimming mode wasadopted in the experiments. It usesthe flapping motion and the featheringmotion in horizontal plane of therobot; on the other hand, it uses therowing motion and the feathering mo-tion in vertical motion of the robot.The speeds of the towing carriagewere determined by operating it par-allel to the robot for several values ofamplitudes of flapping motion and

feathering motion. From the experi-mental result of the swimming speeds,the thrust forces were estimated byusing the relationship between dragforces and swimming speeds. Fig-ure 22 shows the thrust forces for theamplitudes of feathering motion of30° and 45° and the amplitudes of flap-ping motion of 20° and 30°. From thisfigure, we can see that the thrust forceincreases as the amplitude of theflapping motion increases. Figure 23shows the thrust forces for amplitudesof a flapping motion of 30° versus am-plitudes of the feathering motion.From this figure, we can see that thethrust force becomes the largest at the

amplitude of the feathering motion of30°. Judging from these results, thelargest thrust force is made at the am-plitudes of flapping motion and feath-ering motion of 30°.

The swimming speeds of forwardmotion, backward motion, turningmotion, and vertically descending mo-tion were measured as performances ofthe maneuverability of the robotic tur-tle. Table 2 shows the swimmingspeeds of these motions with fixed am-plitudes of joint angle. Unlike the for-ward motion, the swimming speed inbackward motion with the amplitudesof feathering motion of 45° is largerthan the case of 30°.

TABLE 1

Specification of RT-I.

Total mass [kg] (without buoyancy control device and battery) 90

Depth of pressure resistant [m] 10

Walking speed in the water [m/s] (experimental value) 0.025

Swimming speed [m/s] (experimental value) 0.168

Dimension [m] Body width 0.73

Body height 0.55

Body length 1.68

Forearm length 0.26

Humerus length 0.2

FIGURE 22

Thrust forces in swimming motion against amplitudes of flapping motion and feathering motion.

FIGURE 21

Photograph of the RT-I amphibious robotequipped with fin actuators.


Walking Performancein Still Water

The crawl gait was adopted as thewalking gait of the robotic turtle be-cause it has a high level of stability.During the walking motion using thecrawl gait, the robot moves each armindividually. This means the robotsupports the body with at least threearms at all times. In operation of therobot in sea water, it is supposed thatthe seabed is not flat and there existdisturbances such as waves and watercurrents. Therefore, the robotic turtleneeds more stability under such condi-tions. For that reason, we introduced amodified walking motion by consider-

ing the movement of the center ofgravity in the normal crawl gait, asshown in Figure 24. The movementof the center of gravity makes the sta-bilitymargin larger, which is defined asthe distance from the side of the sup-port polygon to the projection of thecenter of gravity on land, as shown inFigure 25.

Two types of walking patterns weremade. The first walking pattern con-sists of four steps to move an arm, fora total of 16 steps to move four armsduring a cycle. The trajectory of theend of an arm in this pattern forms arectangle. The second walking patternconsists of three steps to move an arm,for a total of 12 steps to move four

arms during a cycle (Figure 26). Thetrajectory of the end of the arm ofthe second walking pattern forms aright triangle. In this case, it is expectedthat walking will be faster because ofthe reduction in the number of walk-ing steps and that the energy consump-tion will be smaller because of thereduction of total motor speed.

The walking speed and the motorcurrents for each walking patternwere measured while the robot walkedon the bottom of the pool (L × B ×H =4.5 m × 2.0 m × 0.8 m) (see Figure 27)to evaluate the walking performancesin still water. Here the length of thestride of each walking pattern was setat 0.2 m. The joint torque was derivedfrom the value of the motor currentmeasured by the experiment, thetorque constant, the reduction ratio,and the transmission efficiency, asshown in Eq. (12).The consump-tion energy was derived from thevalue of the derived joint torque, asfollows:

qij ¼ Iij ⋅Kt ⋅ i⋅η′ ð12Þ

Here, Iij denotes the motor current ofthe joint Pij , Kt is the torque constant,i is the reduction ratio, andη′ is thetransmission efficiency. Figure 28shows a comparison of walking speedsbetween measurement and simulationby the walking simulator for the twowalking patterns. From this figure,we can see that the walking speed forthe second walking pattern is greater.This is attributed to the small periodof the second walking pattern for onecycle because of the reduction in walk-ing steps. We can see that the experi-mental value of the walking speed foreach walking pattern is smaller thanthe calculated value. This is becausethe end of the arm slips on the bottom

FIGURE 23

Thrust forces in swimming motion against amplitudes of feathering motion.

TABLE 2

The swimming speeds for various swimming motions.

Swimming PatternFlapping (Rowing)Amplitude (°)

FeatheringAmplitude (°)

SwimmingSpeed

Forward motion 30 30 0.168

Forward motion 30 45 0.15

Backward motion 30 30 0.088 m/s

Backward motion 30 45 0.102 m/s

Turning motion 30 30 9.33°/s

Turning motion 30 45 10.30°/s

Descending motion 30 (rowing) 45 0.05 m/s


of the pool when the robot walks instill water.

Walking Performance in WavesWe estimated the walking perfor-

mance of the robotic turtle in wavesthrough a series of experiments. Fig-

ure 29 shows a schematic view of theexperiment for estimating the walkingperformance in waves. Two types ofexperiments were carried out in thetowing tank of Osaka University.The first type measured the walkingperformance by changing the walking

patterns in waves and in still water.The second type measured the walkingperformances of the second walkingpattern in waves by changing thewave conditions, namely, the waveheight, the wavelength, and thedepth of the virtual ground. For thefirst type of experiment, the walkingspeeds of the first and second normalwalking patterns were measured forthe stride lengths (L) of 0.2 m and0.3 m under the wave height of 0.1 m,the wavelength of 2 m, and the waterdepth of 0.9 m. The walking speedsof the first and second modified walk-ing patterns were also measured. Fig-ure 30 shows a comparison of thewalking speeds for various walking pat-terns in waves and in still water. Fromthis figure, we can see that the walkingspeeds for all cases decreased in wavescompared to the performance in stillwater. The walking speed reaches itsmaximum value in the case of thelength of the stride of 0.3 m and thesecond walking pattern, regardless ofcrawl gait. Figure 31 shows a compar-ison of the consumption energy forvarious walking patterns in still waterand in waves. From this figure, wecan see that the consumption energyof the first walking pattern in wavesis larger than that in still water in thecase of the length of the stride of

FIGURE 25

Definition of stability margin.

FIGURE 26

Trajectories of the arm end of the first and second walking patterns.

FIGURE 27

Picture of Robotic Turtle in walking in a pool.

FIGURE 24

Schematic view of crawl gait and modified crawl gait.


0.3 m, although that of the secondwalking pattern in waves is smallerthan that in still water. Regardingthat the walking speeds of the firstand second walking patters in wavesbecome smaller than those in stillwater, we find that the propulsive effi-ciency of the first walking pattern inwaves becomes worse than that of thesecond walking pattern in waves inthe case of the length of the stride of0.3 m. This may be attributed to theinteraction between the hydrodynamicforces acting on the arms and the pathsof the arms, namely, larger energy isconsumed in the first walking patternduring one cycle of the arm motionin waves by the hydrodynamic forcesthan in the second walking.

Figure 32 shows the variation ofwalking speed against the wave heightsfrom 0 m to 0.15 m with the wave-length of 2 m, the water depths of0.6 m and 0.9 m, and normal andcrawl modified gaits for the secondwalking pattern with the length ofthe stride of 0.3 m. From this figure,we can see that the walking speed is de-creased as the wave height increasesand that the walking speed becomessmaller and the rates of the decrease be-come larger as thewater depth decreasesfrom 0.9 m to 0.6 m. Figure 33 showsthe variation of consumption energyagainst the wave heights under thesame condition for Figure 30. Fromthis figure, we can see that the con-sumption energy is almost constantagainst the wave height, although theconsumption energies are smallerthan those in still water, and that thepropulsive efficiency in waves becomeworse if the water depth decreases be-cause the rate of decrease of walkingspeed for d = 0.6m is larger than ford = 0.9 m. This is attributed to largerwater current induced by waves inshallow water, which produces larger

FIGURE 28

Comparison of walking speed in water.

FIGURE 29

Schematic view of experiment for walking performance in waves.

FIGURE 30

Comparison of walking speeds for various walking patterns in still water and in waves.


hydrodynamic drag force on thevehicle.

SummaryIn this paper we discussed the

mechanisms of swimming and walkingof turtles and examined how thosemechanisms could be applied to anamphibious robot, with the goal ofdesigning a robot that can performautomatic monitoring of environ-ments along natural coastal areas andtidal flats. Using a manipulator with4 DOF, we discussed the character-istics of walking of sea turtles and tor-toises from the viewpoints of mobilityand trafficability. The advantages ofsea turtles in swimming and tortoisesin walking were adopted in the roboticturtle. The experiments in a water tankrevealed that the mobility and the traf-ficability of the amphibious robot atthe bottom of the water were greatlyaffected by waves in shallow water.The author will investigate the track-ing performance of the amphibiousrobot on tidal flats, moving fromland to sea and vice versa.

AcknowledgmentsThis research was funded for

3 years beginning in 2008 by the Min-istry of Education, Culture, Sports andTechnology, Japan (grant 20246130)under the project title “Establishmentof basic technology of a biomimeticunderwater vehicle and its applications.”

Author:Naomi KatoGraduate School ofEngineering, Osaka UniversityYamadaoka 2-1, Suita,Osaka 565-0871, JapanEmail: [email protected]

FIGURE 31

Comparison of consumption energy for various walking patterns in still water and in waves.

FIGURE 32

Variation of walking speed against wave height.

FIGURE 33

Variation of consumption energy against wave height.


ReferencesDudek, G., Giguere, P., Prahacs, C.,

Saunderson, S., Sattar, J., Torres-Mendez,

L.-A., … Georgiades, C. 2007. Aqua: An

amphibious autonomous robot. IEEE

Comput Mag. 40(1):46-53.

Ijspeert, A.J., Crespi, A., Ryczko, D., &

Cabelguen, J-M. 2007. From swimming to

walking with a salamander robot driven by

a spinal cord model. Science. 315(5817):

1416-20. doi: 10.1126/science.1138353.

Isobe, Y., Kato, N., Suzuku, H., Senga, H.,

Matsuda, Y., Kawamura, K., … Oshika, T.

2010. Motion analysis of sea turtle with

prosthetic flippers. In: The 20th (2010) Int.

Offshore and Polar Eng. Conf. pp. 335-42.

Beijing: International Society of Polar and

Offshore Eng.

Kato, N., Ando, Y., Shigetomi, T., &

Katayama, T. 2006. Biology-inspired precision

maneuvering of underwater vehicles (Part 4).

Int J Offshore Polar. 16(3):195-201.

Kato, N., & Liu, H. 2003. Optimization

of motion of a mechanical pectoral fin.

JSME Int J C. 47(4):1356-62. doi: 10.1299/

jsmec.46.1356.

Low, K.H., Zhou, C., Ong, T.W., & Yu, J.

2007. Modular design and initial gait study

of an amphibious robotic turtle. In: paper

presented at the International Conference

on Robotics and Biomimetics, 2007. 535-40.

Sanya, China: IEEE.

Setouchi, H., & Shinjo, T. 2001. The effect

of water on frictional characteristics between

calcareous sands and a steel plate. In: Report

of Faculty of Agriculture. University of the

Ryukyus. 48:103-11.

Suzuki, H., & Kato, N. 2005. A numerical

study on unsteady flow around a mechanical

pectoral fin. Int J Offshore Polar. 15(3):161-7.

Suzuki, H., Kato, N., & Suzumori, K. 2007.

Load characteristics of mechanical pectoral

fin. Exp Fluids. 44(5):759-71. doi: 10.1007/

s00348-007-0432-x.

Tanaka, T., & Shirai, T. 2006. Development

of automated shoreline surveying system using

amphibious walking robot—The design

concepts and the 1st field experiment.

In: paper presented at the International

Symposium on Automation and Robotics

in Construction 2006 (ISARC 2006). 46-51.

Tokyo: IEEE.

Wakimoto, S., Suzumori, K., & Kanda, T.

2006. A bio-mimetic amphibious soft cord

robot. JSME Int J C. 72(714):171-7.

Walker, W.F., Jr. 1971. A structual and

functional analysis of walking in the turtle

Chrysemys picta marginata. J Morphol.

134:195-213. doi: 10.1002/jmor.1051340205.

Wyneken, J. 1997. Sea turtle locomotion:

Mechanisms, behavior, and energetics. In:

The Biology of Sea Turtles, eds. Lutz, P.L.,

& Musick, J.A., 165-98. New York: CRC

Press.

Yamada, H., Chigisaki, S., Mori, M., Takita,

K., Ogami, K., &Hirose, S. 2005. Development

of amphibious snake-like robot ACM-R5.

In: paper presented at the 36th International

Symposium on Robotics(ISR 2005). Tokyo:

Japan Robot Association.


P A P E R

Marine Applications of the BiomimeticHumpback Whale FlipperA U T H O R SFrank E. FishWest Chester University

Paul W. WeberApplied Research Associates, Inc.

Mark M. MurrayUnited States Naval Academy

Laurens E. HowleDuke University

A B S T R A C TThe biomimetic approach seeks technological advancement through a transfer

of technology from natural technologies to engineered systems. The morphology ofthe wing-like flipper of the humpback whale has potential for marine applications.As opposed to the straight leading edge of conventional hydrofoils, the humpbackwhale flipper has a number of sinusoid-like rounded bumps, called tubercles, whichare arranged periodically along the leading edge. The presence of the tuberclesmodifies the water flow over the wing-like surface, creating regions of vortex gen-eration between the tubercles. These vortices interact with the flow over the tubercleand accelerate that flow, helping to maintain a partially attached boundary layer.This hydrodynamic effect can delay stall to higher angles of attack, increases lift,and reduces drag compared to the post-stall condition of conventional wings. Asthe humpback whale functions in the marine environment in a Reynolds regimesimilar to some engineered marine systems, the use of tubercles has the potentialto enhance the performance of wing-like structures. Specific applications of the tu-bercles for marine technology include sailboat masts, fans, propellers, turbines, andcontrol surfaces, such as rudders, dive planes, stabilizers, spoilers, and keels.Keywords: tubercles, delayed stall, Megaptera novaeangliae, leading edge,bio-inspired design

Introduction

Life began with water. The high-density and viscous nature of waterhas imposed a strong evolutionary se-lection pressure on the design of ani-mals that move through this aqueousmedium. Over the course of millionsof years, different phylogenetic linesof animals have, in effect, experi-mented with various combinationsof morphologies and behaviors to en-hance locomotor performance. Thegreat diversity of body shapes, surfacetextures, and propulsive mechanismsexhibited by aquatic animals has pro-duced a variety of biomechanical solu-tions for the reduction of drag, increasein thrust production and efficiency,maintenance of stability, and enhance-ment of maneuverability. By emulat-ing these biological characteristics inthose instances where animal perfor-mance is superior to manufactured de-vices, the performance of engineeredmarine systems may be improvedthrough the field of biomimetics.

The cetaceans (whales, dolphins,porpoises) have been the focus of inspi-ration for technological development in

the marine environment. The ceta-cean lineage dates back 55 million years.The intense selection pressures for afast swimming, maneuverable marinepredator have culminated in a highlystreamlined body with advanced sen-sory capabilities that is propelled by ahighly efficient propulsion mechanism.

There are a number of exampleswhere cetaceans have been the inspira-tion for the development or improve-ments of marine technology. Thecetacean body shape was used byCayley (circa 1800) as a solid of least-resistance for the development of air-plane fuselage and boat hull designs(Gibbs-Smith, 1962). The famousbut erroneous “Gray’s paradox” ledto examination of special drag reduc-tion mechanisms (Gray, 1936; Fish& Rohr, 1999; Fish, 2006), including

the biomimetic development of com-pliant coatings for viscous dampen-ing (Kramer, 1960; Riley et al., 1988;Carpenter & Pedley, 2003). The com-pactness and high resolution of theecholocation system of dolphins pro-vides a benchmark for the improve-ment of SONAR systems (Au, 1993).The thrust performance of oscillating,wing-like systems, such as the flukes ofdolphins, has been considered superiorto screw propellers (Peterson, 1925;Liu & Bose, 1993; Triantafyllou &Triantafyllou, 1995). The flexibilityof the oscillatory flukes can allow oper-ation at high efficiency over an ex-tended speed range without cavitation(Fish & Lauder, 2006; Iosilevskii &Weihs, 2007).

This report focuses on a uniquemor-phology of a highly derived aquatic


mammal that has general biomimeticapplications for marine systems. Theflipper of the humpback whale andits bumpy leading edge provide anovel approach to enhance the hydro-dynamic performance of wing-likestructures for operation in water. Aprincipal attribute of using the whaleas a model to construct a biomimeticsystem is the scale. As the whale is ofa large size and swims at speeds thatcompliment the scale and operationof engineered marine systems, applica-tion for marine technologies can bereadily undertaken.

Humpback Whaleas Inspiration

The humpback whale (Megapteranovaeangliae) has the longest flipperof any cetacean (i.e., whale, dolphin,porpoise), with regard to both abso-lute and relative size (Figure 1; Fish& Battle, 1995). The flippers are in-volved with the underwater maneuversperformed by the species that are asso-ciated with their mode of feeding(Friedlaender et al., 2009; Hazenet al., 2009). Humpback whales arethe only baleen whales (e.g., bluewhale, fin whale, minke whale, rightwhale) that rely on tight, rapid turnsto capture prey (Fish & Batt le,1995). The humpback whales usetheir flippers as biological hydroplanesto achieve tight circles to corral andengulf prey (Fish et al., 2011).

The humpback whale flippers areunique because of the presence oflarge tubercles along the leading edge,which gives this surface a scallopedappearance (Figure 1). The distancesbetween tubercles decrease distally,although these distances remain rela-tively constant at 7-9% of span overthe midspan of the flipper (Fish &Battle, 1995).

The planform and cross-sectionalviews of the humpback whale flipperare shown in Figure 1.Whereas typicalwing-like structures have a straightleading edge without the presence ofirregularities or perturbations, thehumpback flipper defies conventionwith prominent rounded bumps thatare regularly spaced along the leadingedge of its high-aspect ratio flip-per (Fish et al., 2008). Humpbackwhale flippers closely resemble the21% thick, low drag NACA 634-021foil in cross section (Abbott & vonDoenhoff, 1959; Fish & Battle,1995). Furthermore, the flippers havehigh mobility (Edel & Winn, 1978).

The elongate flippers function aswings to generate the forces necessaryfor turning maneuvers (Fish et al.,2011). Turning is important in thecapture of elusive prey. Humpbackwhales feed on shoals of small fishand krill. The preys are forced into atight ball by the whales striking thewater surface with their flukes or cir-cling the prey from underneath whileemitting bubbles. The bubbles rise tocorral the prey as a bubble net. In ei-ther case, the whale maneuvers underthe prey for engulfment by executinga rapid turn. Lift generated by the flip-pers is used to produce a centripetalforce for the turn. The tubercles pro-duce vortical flows over the surface ofthe flipper and control lift characteris-tics at high angles of attack and delaystall (Fish & Battle, 1995; Miklosovic

et al., 2004; Fish& Lauder, 2006; Fishet al., 2011).

Hydrodynamic Effectof Tubercles

The prominence of the tubercleson the leading edge of the humpbackwhale flippers and the swimming pat-tern of the whale suggests that thesenovel structures have a distinct hydro-dynamic effect (Figure 2). In addition,the pattern of barnacle distribution onthe flippers (i.e., barnacles are confinedto the peak of the tubercle and donot occur between tubercles; Fish &Battle, 1995) indicates that the flowover the flipper is affected by tubercles.This section reviews potential hydro-dynamic advantages, flow control,and limitations of the tubercles onwing-like structures.

Hydrodynamic AdvantagesThe presence of leading-edge tuber-

cles on a wing-like structure can havea positive influence on the hydro-dynamic performance. Wind tunneltests showed that wings with tuberclesimproved maximum lift by over 6%,increased the ultimate stall angle by40%, and decreased drag by as muchas 32% (Figure 3; Miklosovic et al.,2004). The tubercles, when facing

FIGURE 1

Flippers on humpback whale (left), showingscalloped pattern of tubercles (center), and flip-per cross-section (right) (from Fish & Lauder,2006).

FIGURE 2

Idealized humpback flipper with leading-edgetubercles.


into the free stream flow, alter thefluid flow over wing-like structures(Bushnell & Moore, 1991; Fish &Battle, 1995; Fish et al., 2011). Further-more, the lift to drag (L/D) ratio,which represents the aerodynamic ef-ficiency, displayed a greater peak L/Dfor a wing geometry with tubercles(Miklosovic et al. , 2004, 2007;Hansen et al., 2009).

The position and number of tuber-cles on the flipper suggest analogueswith specialized leading edge con-trol devices that improve the hydrody-namic performance of wings. Theoccurrence of “morphological com-plexities” on a lifting body could reduceor use pressure variation at the tip to de-crease drag and improve lift to preventtip stall (Bushnell &Moore, 1991). Al-ternatively, various biological wingsutilize leading-edge control devices tomaintain lift and avoid stall at high at-tack angles and low speeds.

The function of the tubercles maybe analogous to strakes used on air-

craft. Strakes are large vortex gen-erators that change the stall charac-teristics of a wing (Hoerner, 1965;Shevell , 1986; Bertin & Smith,1998). Stall is postponed because thevortices exchange momentum withinthe boundary layer to keep it partiallyattached over the wing surface. Lift isthus maintained at higher angles of at-tack with strakes compared to wingswithout strakes, although maximumlift is not increased by strakes (Shevell,1986). Another leading-edge device areslots or slats that delay stall or movethe angle of attack of maximum liftto a lower value (Wegener, 1991;Bertin & Smith, 1998). The moveableslats create a space anterior of the fixedwing to allow higher-pressure fluid torise from the underside of the wing.The movement of fluid from thehigh-pressure side to the low-pressureside of the wing improves mixing andhelps to maintain the boundary layerand delay stall (Wegener, 1991). How-ever, the tubercles have distinct ad-

vantages over slats. Tubercles arepassive structures that have no dragpenalty when designed onto wings(Miklosovic et al., 2004), whereasslats are actively deployed and incur in-creased drag (Hoerner, 1965).

Vortex GenerationThe mechanism for enhanced hy-

drodynamic performance due to thepresence of tubercles appears due tothe specific pattern of vortex genera-tion over the surface of the flipper.Flow visualization experiments onwavy bluff bodies showed periodicvariation in the wake width across thespan (Owen et al., 2000). A wide wakewith two simultaneous vortices oc-curred where the body protrudeddownstream and a narrow wake oc-curred where the body protruded up-stream. The flow in the wake of thewavy body was different from thewake of a straight cylinder, which ex-hibited a typical von Karman VortexStreet of alternating vortex pairs. Abluff body with a spanwise sinusoidalform could reduce drag by at least30%, compared to equivalent straightbodies (Bearman & Owen, 1998).

The vortices produced by a wingsection with tubercles are shown inFigure 4. The tubercles generateseparated, chordwise vortices in thetroughs at high angles of attack.These vortices are formed as the flowstrikes the leading edge of the trough.As the flow does not strike the leadingedge normally, the flow is sheared intothe trough’s center to generate thevortices. These vortices are convectedalong the chord. The spanwise ar-rangement of the vortices is in a pairon each side of the tubercle crest withopposite spins (Hansen et al., 2010).The flow directly over the tubercle in-teracts with the vortices located down-stream and lateral to the tubercle crest.

FIGURE 3

Humpback whale flipper models and results of wind tunnel experiments. The models (left) withand without tubercles were machined from clear polycarbonate, based on a symmetrical NACA0020 foil section. Lift and drag data (right) for the flipper models were obtained from tests in awind tunnel. The solid lines in a, b, and c show the average of the data for the flipper model with-out tubercles, and open triangles are for the model with tubercles. Lift coefficient CL (a), drag co-efficient CD (b), and aerodynamic efficiency L/D (c) are plotted against angle of attack, α (fromMiklosovic et al., 2004).


The tangential velocities of the inwardfacing flows of the pair of vortices aredirected toward the trailing edge ofthe wing section. The flow from thetubercle peak is accelerated posteriorlydue to the interaction with the vortexpair. These effects prevent the localboundary layer downstream of the tu-bercles from separating and push thestall line further posterior toward thetrailing edge. When integrated overthe entire structure, the wing with tu-bercles will stall at a higher angle ofattack than a wing without tubercles.

Flow experiments conducted on amodel wing section with leading-edgetubercles at low speeds showed flowseparation from the troughs betweenadjacent tubercles but attached flowon the tubercles (Johari et al., 2007).Flow separation pattern and surfacepressure was dramatically altered bythe tubercles. For regions downstreamof tubercle crest, separation was de-layed almost to the trailing edge. Al-though this flow pattern did notresult in improved lift generation,drag reduction or delay in the stallangle of attack, the post-stall character-

istics were greatly smoothed ( Johariet al., 2007; Saadat et al., 2010),which could provide important perfor-mance benefits for systems that rou-tinely operate beyond the stall point.

The vortices produced from thetubercles re-energize the boundarylayer by carrying high-momentumflow close to the flipper’s surface(Figure 5; Wu et al., 1991; Pedro &

Kobayashi, 2008; Hansen et al.,2010). The flow dynamics are im-proved also by confining separationto the tip region. Tubercles delay stallby causing a greater portion of the flowto remain attached on a wing, with theattached flow localized behind the tu-bercle crests (Weber et al., 2011).

The size and frequency of the tu-bercles along the leading edge influ-ences the performance of a wing.Small amplitude tubercles show thebest performance with regard to liftand stall characteristics ( Johari et al.,2007; Hansen et al., 2009, 2011).The wavelength and thus frequencyof tubercles was found, however, tohave little effect on performance(Johari et al., 2007).

The tubercle effect is further en-hanced when sweepback is added tothe wings (Murray et al., 2005).Wings with sweep angles of 15° and30° required higher angles of attackto achieve stall than nonswept wingsand showed superior drag performanceover most of the range of α comparedto models without tubercles (Murrayet al., 2005). Flow tests on delta

FIGURE 4

Pressure contours and streamlines atα = 10° for NACA 63-021with straight leading edge (left) andwith tubercles (right). An unsteady Reynolds-averaged Navier-Stokes (RANS) simulation wasused. A separation line is shown on the wing section without tubercles. For the wing with tuber-cles, large vortices are formed posterior of the troughs along the leading edge and flow posterior ofthe tubercles is shown as straight streamlines without separation. Images courtesy of E. Paterson.

FIGURE 5

Vorticity computed from detached eddy simulation (DES) for flippers with (a) and without (b) tu-bercles at an angle of attack of 15°. Vortices re-energize boundary layer to delay separation andstall. Images courtesy of H. T. C. Pedro and M. H. Kobayashi.


wings with a sweep of 50° showed thatat high angles of attack large-scalethree-dimensional separation oc-curred for the wing with a straightleading edge (Goruney & Rockwell,2009). However when tubercles areadded, the flow is radically trans-formed. Tubercles with amplitude of4% of wing chord can completelyeradicate the negative effect of theseparation and foster re-attachment.

Experiments performed on flap-ping wings with tubercles (Figure 6)showed an affect on the spanwiseflow (Ozen & Rockwell, 2010). Typi-cally a straight wing, whether flapping

or static, will develop a spanwise flowdue to the pressure differential that de-velops between the upper and lowersurfaces. This spanwise flow becomesmanifest as a wing tip vortex, which in-creases the drag and reduces the effi-ciency of a wing. A flapping wingwith tubercles does not produce a pro-nounced region of spanwise flow, butthe wing tip vortex generation is unaf-fected (Ozen & Rockwell, 2010).

Limitations of TuberclesEnhanced performance due to the

presence of the tubercles is not uni-versal. There exist limitations to the

advantages of the tubercles. Tuberclesimprove performance when in concertwith wing geometries that are charac-terized by the combination of finitespan, swept wing, tapered planform,and thick foil.

Foil sections with no wing tip thatemulate infinite wings do not demon-strate reduced drag and increased lift, al-though stall is still delayed (Miklosovicet al., 2004; Johari et al., 2007; VanNierop et al., 2008). Tip effects occuras a consequence of lift generationwhen a fully three-dimensional wingis canted at an angle of attack to an in-cident flow. Induced drag is producedin lift generation from kinetic energyimparted to the fluid from pressure dif-ferences between the two surfaces ofthe wing as there is leakage of fluidfrom high pressure to low pressurearound the distal tip of a lifting surfaceresulting in spanwise flow and the for-mation of tip vortices (Vogel, 1981).The flow pattern set up by the tuber-cles helps to maintain a chordwiseflow and reduce the induced drag dueto tip vortices. This effect is only real-ized for finite wings.

The tubercles require wings withthick sectional geometries to function.The section must have a prominentnose radius. In part, this is due to thenecessity to contour the tubercles three-dimensionally into the leading edgeto avoid any flat surfaces. Leading-edge tubercles were tested on foilsbased on NACA 0021, 634-021 and65-021 sections (Miklosovic et al.,2004; Johari et al., 2007; Custodioet al., 2010; Hansen et al., 2009,2011). These designs approach thecross-sectional geometry of the hump-back whale flipper (Fish & Battle,1995).

It is necessary to have a relativelysteady flow to maintain the pattern ofthe vortices and incur the hydrodynamic

FIGURE 6

Comparison of flapping plate at Reynolds number of 1300 at angle of attack of 8° with and withouttubercles from Ozen and Rockwell (2010). The pattern of the flow structure for the plate with tu-bercles produces a series of spanwise vortices that limit spanwise flow compared with the flappingplate without tubercles. Image courtesy of D. Rockwell.


advantages (Stanway, 2008). Foilswith tubercles that were oscillated inroll and pitch demonstrated that thetubercles did not improve hydrody-namic performance (Stanway, 2008).Flapping degraded performance bythe redirection of energy to tubercle-generated vortices from the vorticesof the wake, which are necessary forthrust production during flapping.Furthermore, if the period of os-cillations is too rapid, there may beinsufficient time to allow the full devel-opment of the vortices over a wing. It isnecessary to have a relatively steadyflow to maintain the pattern of thevortices and incur the hydrodynamicadvantages (Stanway, 2008).

Tubercle TechnologiesApplication of natural technologies

into biomimetic-engineered systemshas a number of problems that are in-herent due to differences in biologicaland engineered systems (Fish, 2006).Engineered systems are relatively largein size, are composed of dry rigidmaterials, including metals and ce-ramics, use rotation motors, and arecontrolled by computational systemswith limited sensory feedback. Biolog-ical structures associated with animalsystems are relatively small in size, arecomposed of wet compliant materials,including composites of ceramics,polysaccharides and proteins, moveby translational displacements gener-ated by muscles, and are controlledby complex neural networks with mul-tiple sensory inputs and fine scalemotor outputs.

The humpback whale tuberclesprovide an ideal solution for applica-tion to engineered marine systems.The size of the whale and its flippersoperate near or at the same scale assome marine vehicles. The mature

whales have a maximum length of17 m and weigh 40,000 kg (Clapham& Mead, 1999). Flipper length is ap-proximately one-third the length ofthe animal and can be over 5 m. Thewhale cruises between 1.1 and 4.0 m/sand is able to burst to a speed of 7.5 m/s(Fish & Rohr, 1999). The flow experi-enced by andmodified by the tuberclesis within the same Reynolds regimethat coincides with a large array ofengineered applications. The Reynoldsnumber of a flipper is 1.6 × 106 whenthe whale is lunge feeding (2.6 m/s).Thus, the flipper and tubercles are op-erating in a turbulent flow regime,which is the standard operating condi-tion for most marine systems. Further-more, the tubercle functions passivelyto modify flow and maintain favorablehydrodynamic conditions. Therefore,control systems can be simplified.

Control SurfacesThere are few other passive means

of altering fluid flow around a wing-like structure that can delay stall andboth increase lift and reduce drag atthe same time. As a result, the applica-tion of leading-edge tubercles for pas-sive flow control has potential in thedesign of marine technologies. Theubiquity of wing-like structures withmarine applications for stability andmaneuverability presents an opportu-nity to enlist tubercles to improve per-formance. Included in such structuresare fixed surfaces, such as keels, finsand skegs, and mobile control surfaces,such as rudders and dive planes.

Delay of stall by tubercles on bothfixed and mobile control surfacesprovides a benefit in tight turning sit-uations. To produce the required cen-tripetal force to effect a turn, a lift forcethat is directed toward the center of theturn is generated by the control sur-face. The magnitude of the lift is

directly dependent on the angle of at-tack with a higher angle providing agreater centripetal force. As stall canbe delayed with tubercles, the controlsurface can operate at higher angles ofattack producing a tighter turn radiuswith more control. If stall were tooccur, the control surface would notbe able to generate the centripetalforce to maintain the turn. In effect,it would be like driving a car along acurved road and slipping on a patchof ice. The reduced friction and cen-tripetal force between the ice and thetires would cause the car to drive offthe road tangential to the curve, ratherthan following the original curvedtrajectory.

A low-aspect ratio rudder withtubercles (Figure 7) and an unsweptleading edge generated more lift at an-gles of attack above 22° compared to asmooth rudder at a Reynolds numberof 200,000 (Weber et al., 2010). Athigher Reynolds numbers, this effectdiminishes and the tubercles acceleratethe onset of cavitation. A human-powered submarine, Umpty Squash,utilized tubercled dive planes and rud-ders (Figure 8). Students of the Sussex

FIGURE 7

Rudder with leading-edge tubercles.


County Technical High School,Sparta, NJ, constructed the subma-rine. In 2005, the submarine com-peted in the International SubmarineRaces held at the David TaylorModel Basin in Bethesda, MD. Thesubmarine was capable of making a90° turn within 25 feet. The designersat Feadship De Voogt developedBreathe, a concept superyacht thatincorporates biomimicry into thedes ign (www.feadship.nl) . Thestabilizers and steering fins werebased on the humpback whale flipperwith tubercles (Figure 9).

Commercially, the company FluidEarth markets a surfboard skeg with

leading-edge tubercles (Figure 10;Anders, 2009). The addition of tuber-cles would provide enhanced controlduring a cutback, when a surfer rapidlychanges direction by 180° to maneuverthe surfboard opposite to the directionof the wave’s braking motion.

Application of tubercles to the mastof a sailboat (Figure 11) could beuseful during close reach maneuvers.A close reach would have the sail set

with a high angle of attack to theapparent wind. The lack of a thickcross-section by the sail itself may pre-clude any advantage in lift and stall.However, the presence of tubercleson the mast, representing a bluffbody, could have advantages in termsof drag. Bluff bodies, like cylinders,

FIGURE 8

Human-powered submarine (left) with rudder and dive planes with tubercles (right). Courtesy ofChris Land and the Sussex County Technical School.

FIGURE 9

Conceptual yacht incorporating biomimeticstructures. Stabilizers are shaped like the flip-pers of the humpback whale. Image courtesyof Feadship.

FIGURE 10

Surfboard skeg with leading-edge tubercles.

FIGURE 11

Iceboat with leading-edge tubercles on themast supporting a sail.


can experience lowered drag when theleading edge has a sinusoidal design(Bearman & Owen, 1998).

Propellers and TurbinesThe use of tubercles can effectively

be employed in the generation ofpower by wing-like structures. Propel-lers operating in a marine system havethe potential to be improved by theaddition of tubercles (Figure 12).The effective angle of attack of a pro-peller blade can be increased by in-creasing the blade angle (Larrabee,1980). A higher angle of attack canproduce more lift to derive greaterthrust and increase the effective pitchof the propeller. Prevention of stallby the tubercle effect would reduceflow-induced vibrations. Furthermore,suppression of tonal noise is possibleby the addition of tubercles to a pro-peller (Hansen et al., 2010). Tonalnoise is most effectively reduced bylarge amplitude and smaller wave-length tubercles. As propellers producea particular noise signature, reductionof noise would be advantageous forstealth in naval operations. Similarly,a reduction of noise pollution by com-mercial marine traffic could be benefi-cial to marine organisms, although theuse of radiated noise by whales to avoidcollisions with ships may be negativelyimpacted.

Tubercle modified blades were alsofound to be effective in power genera-

tion of a marine tidal turbine at lowflow speeds (Murray et al., 2010). Tu-bercles were placed on the distal 40%of the three turbine blades. Comparedto blades with smooth leading edges,blades with leading-edge tuberclesdemonstrated enhanced performance.The marine tidal turbine is analogousto wind turbines. A variable pitchwind turbine with retrofitted bladeswith tubercles demonstrated increasedelectrical generation at moderate windspeeds compared to unmodified blades(Howle, 2009; Wind Energy Instituteof Canada, 2008).

ConclusionsA passive means of altering fluid

flow around a wing-like structure thatcan delay stall and both increase liftand reduce drag simultaneously ishighly novel. This performance byleading-edge tubercles, therefore, haspotential application for passive flowcontrol in the design of various marinetechnologies. The flow is modified bythe formation of paired vortices in thetroughs between tubercles. These vor-tices interact with the flow over the tu-bercles to keep the flow attached to thewing surface and delay stall. The tuber-cles perform best when designed intotapered wings with finite span, sweptplanform, and with a thick foil sectionthat have limited oscillatorymovement.Such applications for marine technol-ogy include fins, rudders, dive planes,water turbines and propellers. The fu-sion of these marine devices and tuber-cles can produce biomimetic designsthat can exhibit superior performanceto conventional engineered systems.

AcknowledgmentsWe thank the technical sup-

port staff of the United States Naval

Academy. This work was supportedby the National Science Foundation(IOS-0640185) to FEF and the Na-tional Defense Science and Engineer-ing Graduate (NDSEG) Fellowshipto PWW.

Lead Author:Frank E. FishDepartment of BiologyWest Chester UniversityWest Chester, PA 19383Email: [email protected]

ReferencesAbbott, I.H., & von Doenhoff, A.E. 1959.

Theory of Wing Sections. New York: Dover.

Anders, M. 2009. Fins I like. Surfer’s Path,

April/May 2009:116.

Au, W.W.L. 1993. The Sonar of Dolphins.

New York: Springer-Verlag.

Bearman, P.W., & Owen, J.C. 1998.

Reduction of bluff body drag and suppression

of vortex shedding by the introduction of

wavy separation lines. J Fluid Struct.

12:123-30. doi: 10.1006/jfls.1997.0128.

Bertin, J.J., & Smith, M.L. 1998. Aero-

dynamics for Engineers. Upper Saddle River,

NJ: Prentice Hall.

Bushnell, D.M., & Moore, K.J. 1991. Drag

reduction in nature. Annu Rev Fluid Mech.

23:65-79. doi: 10.1146/annurev.

fl.23.010191.000433.

Carpenter, P.W., & Pedley, T.J. 2003.

IUTAM Symposium on Flow Past Highly

Compliant Boundaries and in Collapsible

Tubes. Dordrecht: Kluwer.

Clapham, P.J., & Mead, J.G. 1999.

Megaptera novaeangliae. Mamm Spec.

604:1-9.

Custodio, D., Henoch, C., & Johari, H.

2010. The performance of finite-span

hydrofoils with humpback whale-like leading

edge protuberances. 63rd Ann Mtg APS Div

FIGURE 12

Two views of marine propeller designs withtubercles.


Fluid Dyn, 55 (16), Nov. 21, 2010, Long

Beach, CA.

Edel, R.K., & Winn, H.E. 1978. Observa-

tions on underwater locomotion and flipper

movement of the humpback whale Megaptera

novaeangliae. Mar Biol. 48:279-87.

doi: 10.1007/BF00397155.

Fish, F.E. 2006. The myth and reality of

Gray’s paradox: Implication of dolphin drag

reduction for technology. Bioinspir Biomim.

1:R17-25. doi: 10.1088/1748-3182/1/2/R01.

Fish, F.E., & Battle, J.M. 1995. Hydrody-

namic design of the humpback whale flipper.

J Morphol. 225:51-60. doi: 10.1002/

jmor.1052250105.

Fish, F.E., Howle, L.E., & Murray, M.M.

2008. Hydrodynamic flow control in marine

mammals. Integr Comp Biol. 211:1859-67.

Fish, F.E., & Lauder, G.V. 2006. Passive

and active flow control by swimming fishes

and mammals. Annu Rev Fluid Mech.

38:193-224. doi: 10.1146/annurev.

fluid.38.050304.092201.

Fish, F.E., & Rohr, J. 1999. Review of

Dolphin Hydrodynamics and Swimming

Performance. San Diego, CA: SPAWARS

TR 1801.


Howle, L.E. 2011. The humpback whale’s

flipper: Application of bio-inspired tubercle

technology. Integr Comp Biol. 51:203-13.

doi: 10.1093/icb/icr016.

Friedlaender, A.S., Hazen, E.L., Nowacek,

D.P., Halpin, P.N., Ware, C., Weinrich,

M.T., … Wiley, D. 2009. Diel changes in

humpback whale Megaptera novaeangliae

feeding behavior in response to sand lance

Ammodytes spp. behavior and distribution.

Mar Ecol Prog Ser. 395:91-100. doi:

10.3354/meps08003.

Gibbs-Smith, C.H. 1962. Sir George

Cayley’s Aeronautics 1796-1855. London:

Her Majesty’s Stationery Office.

Goruney, T., & Rockwell, D. 2009.

Flow past a delta wing with a sinusoidal

leading edge: Near-surface topology and flow

structure. Exp Fluids. 47:321-31.

doi: 10.1007/s00348-009-0666-x.

Gray, J. 1936. Studies in animal locomotion

VI. The propulsive powers of the dolphin.

J Exp Biol. 13:192-9.

Hansen, K.L., Kelso, R.M., & Dally, B.B.

2009. The effect of leading edge tubercle

geometry on the performance of different

airfoils. In: 7thWorld Conf ExpHeat Transfer,

Fluid Mechanics and Thermodynamics.

28 June-03 July 2009. Krakow, Poland.

Hansen, K.L., Kelso, R.M., & Dally, B.B.

2011. Performance variations of leading-edge

tubercles for distinct airfoil profiles. AIAA

J Aircraft. 49:185-94.

Hansen, K.L., Kelso, R.M., & Doolan, C.J.

2010. Reduction of flow induced tonal noise

through leading edge tubercle modifications.

In: 16th AIAA/CEAS Aeroacoustics Conf.

7-9 June 2010. Stockholm, Sweden.

Hazen, E.L., Friedlaender, A.S., Thompson,

M.A., Ware, C.R., Weinrich, M.T., Halpin,

P.N., & Wiley, D.N. 2009. Fine-scale

prey aggregations and foraging ecology of

humpback whales Megaptera novaeangliae.

Mar Ecol Prog Ser. 395:75-89.

doi: 10.3354/meps08108.

Hoerner, S.F. 1965. Fluid-dynamic Drag.

Published by author. Brick Town, NJ.

Howle, L.E. 2009. WhalePower Wenvor

blade. A report in the efficiency of aWhalePower

Corp. 5 meter prototype wind turbine blade.

BelleQuant Engineering, PLLC.

Iosilevskii, G., & Weihs, D. 2007. Speed

limits on swimming of fishes and cetaceans.

J R Soc Interface. 5:329-38. doi: 10.1098/

rsif.2007.1073.

Johari, H., Henoch, C., Custodio, D., &

Levshin, A. 2007. Effects of leading-edge

protuberances on airfoil performance. AIAA

J. 45:2634-42. doi: 10.2514/1.28497.

Kramer, M.O. 1960. The dolphin’s secret.

New Sci. 7:1118-20.

Larrabee, E.E. 1980. The screw propeller.

Sci Am. 243:134-48.

Liu, P., & Bose, N. 1993. Propulsive perfor-

mance of three naturally occurring oscillating

propeller planforms. Ocean Eng. 20:57-75.

doi: 10.1016/0029-8018(93)90046-K.

Miklosovic, D.S., Murray, M.M., &

Howle, L.E. 2007. Experimental evaluation of

sinusoidal leading edges. J Aircraft. 44:1404-7.

doi: 10.2514/1.30303.

Miklosovic, D.S., Murray, M.M., Howle,

L.E., & Fish, F.E. 2004. Leading-edge

tubercles delay stall on humpback whale

(Megaptera novaeangliae) flippers. Phys

Fluids. 16:L39-42. doi: 10.1063/1.1688341.

Murray, M.M., Fish, F.E., Howle, L.E., &

Miklosovic, D.S. 2005. Stall delay by leading

edge tubercles on humpback whale flipper

at various sweep angles. In 14th Int Symp

Unmanned Untethered Submersible Tech.

Durham New Hampshire: Autonomous

Undersea Systems Insitute.

Murray, M., Gruber, T., & Fredriksson, D.

2010. Effect of leading edge tubercles on

marine tidal turbine blades. 63rd Ann Mtg

APS Div Fluid Dyn, 55(16), Nov. 21, 2010,

Long Beach, CA.

Owen, J.C., Szewczyk, A.A., & Bearman,

P.W. 2000. Suppression of Karman vortex

shedding. Phys Fluids. 12:S9.

Ozen, C.A., & Rockwell, D. 2010. Control

of vortical structures on a flapping wing

via a sinusoidal leading-edge. Phys Fluids.

22:021701. doi: 10.1063/1.3304539.

Pedro, H.T.C., & Kobayashi, M.H.

2008. Numerical study of stall delay on

humpback whale flippers. In: 46th AIAA

Aerospace Mtg Exhibit. 7-10 Jan. 2008.

Reno, Nevada.

Peterson, C.J.G. 1925. The motion of

whales during swimming. Nature. 116:327-9.

doi: 10.1038/116327a0.

Riley, J.J., Gad-el-Hak, M., & Metcalfe,

R.W. 1988. Compliant coatings. Annu Rev

Fluid Mech. 20:393-420. doi: 10.1146/

annurev.fl.20.010188.002141.

Saadat, M., Haj-Hariri, H., & Fish, F. 2010.

Explanation of the effects of leading-edge


tubercles on the aerodynamics of airfoils

and finite wings. 63rd Ann Mtg APS Div

Fluid Dyn, 55(16), Nov 21, 2010, Long

Beach, CA.

Shevell, R.S. 1986. Aerodynamic anomalies:

Can CFD prevent or correct them? J Aircraft.

23:641-9. doi: 10.2514/3.45356.

Stanway, M.J. 2008. Hydrodynamic effects

of leading-edge tubercles on control surfaces

and in flapping foil propulsion. M.S. thesis,

Massachusetts Institute of Technology.

Triantafyllou, G.S., & Triantafyllou, M.S.


Sci Amer. 272:64-70. doi: 10.1038/


Van Nierop, E.A., Alben, S., & Brenner,

M.P. 2008. How bumps on whale flippers

delay stall: An aerodynamic model. Phys

Rev Lett. 100:054502. doi: 10.1103/

PhysRevLett.100.054502.

Vogel, S. 1981. Life in Moving Fluids.

Boston: Willard Grant Press.

Weber, P.W., Howle, L.E., & Murray, M.M.

2010. Lift, drag, and caviation onset on

rudders with leading-edge tubercles. Marine

Technol Sname N. 47(1):26.

Weber, P.W., Howle, L.E., Murray, M.M.,

& Miklosovic, D.S. 2011. Computational

evaluation of the performance of lifting

surfaces with leading edge protuberances.

J Aircraft. 48:591-600. doi: 10.2514/

1.C031163.

Wegener, P.P. 1991. What Makes Airplanes

Fly? New York: Springer-Verlag.

Wind Energy Institute of Canada. 2008.

WhalePower Tubercle Blade Power Perfor-

mance Test Report. North Cape, PE:

Wind Energy Institute of Canada.

Wu, J.Z., Vakili, A.D., & Wu, J.M. 1991.

Review of the physics of enhancing

vortex lift by unsteady excitation.

Prog Aerosp Sci. 28:73-131. doi: 10.1016/

0376-0421(91)90001-K.


P A P E R

Shark Skin Separation Control MechanismsA U T H O R SAmy LangUniversity of Alabama

Philip MottaMaria Laura HabeggerUniversity of South Florida

Robert HueterMote Marine Laboratory

Farhana AfrozUniversity of Alabama

A B S T R A C TDrag reduction by marine organisms has undergone millions of years of natural

selection, and from these organisms biomimetic studies can derive new technolo-gies. The shortfinmako (Isurus oxyrinchus), considered to be one of the fastest andmost agile marine predators, is known to have highly flexible scales on certain loca-tions of its body. This scale flexibility is theorized to provide a passive, flow-actuatedmechanism for controlling flow separation and thereby decreasing drag. Recent bi-ological observations have found that the shortfin mako has highly flexible scales,bristling to angles in excess of 50°, particularly on the sides of the body downstreamof the gills. High “contragility,” which is explicitly defined here as the ability tochange or move in a new or opposing direction while already in a turn, wouldoccur if form drag were minimized. This would thus indicate the potential controlof flow separation on body regions aft of the point of maximum girth or in regions ofadverse pressure gradient. Thus results are consistent with the hypothesis thatscale bristling controls flow separation. This scale flexibility appears to be a resultof a reduction in the relative size of the base of the scales as well as a re-organization of the base shape as evidenced by histological examination of theskin and scales. Probable mechanisms leading to separation control are discussed.Keywords: shark skin, flow separation, drag reduction

Introduction

Natural selection, operating forhundreds of millions of years, hashoned fast swimming marine organ-ism forms to reduce energy expen-diture through streamlining. Amongthe fastest swimming fishes, certainspecies of sharks have converged ona suite of adaptations to reduce drag.Consequently, fast swimming sharkshave been studied for insight into po-tential drag-reducing mechanisms forwell over three decades. Drag on asubmerged, swimming body consistsof three sources. These include (i) formdrag due to a difference in pressurearound the body, (ii) drag due to lift,and (iii) skin friction due to boundarylayer formation (Bushnell & Moore,1991). At low Reynolds numbers(Re) skin friction predominates, whileat higher Re pressure drag can domi-nate if not minimized. It is not sur-prising then to consider the fact thataquatic organisms have evolved tominimize drag (Fish, 1998), withthe primary decrease coming froma streamlined body shape to reduceflow separa t ion and thus formdrag. Aquatic organisms that swim at

high Re (>103) have a variety ofshapes and structures to reduce drag,which we often attempt to duplicate(Vogel, 2003; Fish & Lauder, 2006).It has been deduced that a crescent taildesign could decrease induced drag onthe order of 8%, and not surprisinglythis lunate tail design is found onmany fast swimming marine animals(Fish, 1998; Donley et al., 2004).Several researchers (e.g., Andersonet al., 2001) have observed that swim-ming fish experience more frictiondrag than the same r ig id bodytowed. This higher friction is attrib-uted to motion of the body as itswims to produce thrust. The undulat-ing body motion can result in mea-surably thinner boundary layers andthus higher skin friction (Fish, 2006).The argument has been made thatbecause of this higher power outputrequirement, to overcome drag and

maintain a certain speed, swimmingdrag reduction due to various mor-phological mechanisms is extremelyprobable (Schultz & Webb, 2002).

Separation of the boundary layerfrom a body typically occurs in vicin-ities where the flow is deceleratingalong a curved body after the point ofmaximum thickness, resulting in anadverse pressure gradient. As a resultseparation typically occurs in areasposterior of the maximum body thick-ness. Incipient separation is charac-terized by regions of decreasing skinfriction approaching zero, and conse-quent reversal of the flow at the surface(Doligalski et al., 1994). Swimmingkinematics in thunniform fish such astunas and mako sharks are characterizedby cyclically repeating motions andsmall linear and angular accelerations(Blake, 2004). Most fast swimmingsharks, such as the shortfinmako Isurus


oxyrinchus, are thunniform swimmerswhere oscillations are for the mostpart limited to the posterior end ofthe body. It has been reported that in-cipient separation (inflected boundarylayer profile) is often observed duringswimming movements, and this mo-tion may be tuned by the fish to takeadvantage of the lower shear stress ina nearly separating boundary layer;yet separation must be avoided to re-duce form drag and increase thrustproduced by the caudal fin (Andersonet al., 2001).

The skin of sharks is covered byminute scales, called dermal denticlesor placoid scales, which originallyevolved as a hard, protective cover-ing to the animal (Raschi & Tabit,1992). The bases of these hard scalesare embedded in the superficial collag-enous layer of the skin (dermis) termedthe stratum laxum, with the crownsof the scales exposed to the water. It isthe unique geometry observed on fastswimming sharks for these tooth-likescales that is of interest from a drag re-duction standpoint. Beginning in the1970s researchers became interestedin the small , streamwise ridges,or keels, located on the top of eachcrown (see scanning electron micros-copy (SEM) pictures of shark scalesshown in Figure 1). Now labeled asriblets, these ridges were found to re-sult in a reduction in turbulent skinfriction drag of up to 9.9% whensized correctly (Bechert et al., 1997).Even as early as the 1980s, ribletswere utilized on boat hulls competingin the Olympics and America’s Cupbut later were banned (Gad-el Hak,2000).

However, the scales on some fastswimming sharks exhibit a differentproperty that is flexibility or capabilityto bristle, and this is the focus of ourcurrent work. Previous work (Bechert

et al., 2000) investigating this aspect ofthe shark skin found no advantagefrom a skin friction reduction stand-point, and only greatly increased fric-tion drag if the scales were allowed toremain bristled. Thus, a new, pas-sive flow-actuated separation controlmechanism is proposed that is inspiredby the scale flexibility found on theshortfin mako shark.

This investigation chose to focuson the skin of the shortfin makobased on several factors. First, of fast

swimming pelagic sharks, the shortfinmako is considered by most to be thefastest and most agile (Stevens, 2009).It is also one of the more derivedspecies of shark and is recognized asmaking its appearance roughly 55 mil-lion years ago in the line of sharkevolution dating back more than400 million years (Naylor et al.,1997). Next, it is one of two speciesof shark previously reported in litera-ture, the other being the smooth ham-merhead Sphyrna zygaena, as having

FIGURE 1

The average scale erection angles for 16 regions on the shortfin mako shark (Isurus oxyrinchus)together with scanning electron micrographs (top row) and histological sections (middle row) ofthe skin for three regions on the body. The flank region, including B2, has the most flexible scales,which are characterized by long backwardly projecting crowns and relatively short bases. The scalebases are embedded in the superficial part of the dermis. The surface of the scales has three ribletsvisible on the scanning electron micrographs. The scales in any region are oriented along the lon-gitudinal axis of the shark, and anterior is to the left. The erection angles noted on the figure are theangles to which the scales in that region can be manually manipulated without damage to theirattachment and which remain at that angle after release by the needle used to erect them.


flexible scales over large portions of itsbody (Bruse et al., 1993). Finally, theshortfin mako is readily obtainable offthe Atlantic coast of the United Statesand is not currently listed as overfishedor otherwise in a vulnerable statefor conservation purposes (NMFS,2010).

Materials and MethodsTo investigate scale structure and

erection, we acquired two subadultshortfin makos (female: total length,192 cm; fork length, 171.5 cm; male:total length, 158 cm; fork length,150 cm) from commercial and recrea-tional fishers in the coastal waters offMontauk, New York. The frozenspecimens were shipped to the Univer-sity of South Florida in Tampa. There,following Reif (1985), scales at 16 re-gions along the body were marked inorder to sample the scales under avariety of flow regimes (Figure 1).Three 1 cm2 samples from each loca-tion were removed, two for histologicalanalysis and one for SEM. Becauseswimming sharks have superambientsubcutaneous pressure ranging ashigh as 100-200 kPa increasing skinstiffness (Wainwright et al., 1978;Martinez et al., 2002), scale erectionangles were recorded with and withoutsubcutaneous pressure, as follows.

We measured scales erection angleunder a dissecting microscope at amagnification of 135×. In seven ofthe 16 regions (B1, B2, B4, B5, A1,A2, A3; Figure 1), an aneroid sphyg-momanometer was placed under theskin and underlying muscle and thepressure elevated and held at 15 psi(103 kPa), the maximum possiblewithout damaging the underlyingmuscle tissue. Scale erection was notnoted with the increase in subcutane-ous pressure, leading us to suspect a

passive mechanism. While the skinwas pressurized, we used a fine acu-puncture needle to gently manipulatefive haphazardly selected scales totheir maximum erected position with-out tearing them from the skin. Releas-ing the scales, we allowed them tosettle at an erected angle, which wethen measured by calculating thechange in length of the scale crownwhen viewed from above. The inversecosine of the apparent crown length di-vided by the resting or true crownlength provided the angle of erection.Because the individual scales could ac-tually be easily erected past this restingangle, we were in essence calculating aminimal erection angle. The pressurewas then released and the angle simi-larly calculated on the flaccid skin.Our a priori test determined thatstretching the skin in this manner didnot affect the non-pressurized erec-tion angle.

At the other regions (H2, B3, B6,P1, P2, P3, C1, C2, C3; Figure 1)where subcutaneous pressure was notpossible, such as in the fins and overthe body cavity, the erection angleswere measured by gently erecting thescales (if possible) with the acupunc-ture needle and measuring theirpre- and post-erection crown length.Finally, to get a more global pictureof scale flexibility over the entirebody, 35 equidistant sampling loca-tions encompassing the entire dorsal,left lateral, and ventral surfaces ofeach shark were marked and scales ineach area manually erected as beforewithout subcutaneous pressure. Theerected scales were simply recorded asgreater or less than 50°, an angle thatappeared to the approximate maxi-mum. We also measured the crownand base length of scales in selectareas, as well as the spacing of the rib-lets on the scales.

To understand the attachment ofthe scales to the skin, we prepared his-tological sections of the skin and scales,decalcified the scales, stained the sam-ples to reveal the fibrous attachment,and examined the sections at 20× and40× with a compound microscope. Fi-nally, surface pictures of the scales, atall studied regions, were prepared byexamining the skin at 100× and 200×under a SEM.

FindingsThe placoid scales of sharks have

a pulp cavity and a hard enameloidcovering over dentine and are an-chored at the base of the scale to thestratum laxum collagenous layer ofthe dermis (Figures 1 and 2). The ex-posed crowns overlap each other onthe shark’s surface, and the majorityof scales have small riblets or keels ontheir surface, oriented in the stream-wise direction of the flow (Figures 1and 2). On the fast swimming shortfinmako, the flank scales (e.g., areas B2,B5, and A2) have a crown length of ap-proximately 0.18 mm, and each crowntypically has three keels each having aheight of 0.012 mm and a spacing of0.041 mm. This differs from otherslower swimming sharks such as theblacktip shark (Carcharhinus limbatus),whereby preliminary data indicate theflank scales are typically 0.32 mmin length with each crown typicallyhaving five keels with a height of0.029 mm and a spacing of 0.065 mm.When considering shark species as awhole, length of the scales is typicallyfixed for specific regions of the bodywithin a species but differs among re-gions and species. Similarly, the num-ber of keels per scale is also consistentper body location for a species.

Scale flexibility on the shortfinmako varies considerably across the


body and fins, with average erectionangles varying from 0° on the leadingedge of the pectoral fin to approxi-mately 50° on the widest part of thebody just behind the gill region (Fig-ure 3). However, only certain portionsof the body have very flexible scales.The most flexible scales are found onthe flank of the body extending be-hind the gills to the tail; here scalesare found to be easily erected with slightmanipulation on dead specimens to an-gles of approximately 50° or greater(Figure 3). The lateral scales (B2, B5,A2) had significantly greater erectionangles (mean angle = 44° ± 1°) thanboth the dorsal (mean angle = 26° ±1° SE) and ventral regions (mean

angle = 25° ± 2° SE). Erection an-gles for the dorsal region did not dif-fer from that of the ventral region(Kruskal-Wallis one-way ANOVA,Tukey’s pairwise test; H = 53.173,

df = 2, P ≤ 0.001). Highly flexiblescales are also found at the trailingedge of the pectoral fins. The scaleson the trailing edge of the pectoral(P3) had the highest erection anglescompared to the leading edge (P1)and the central region of the fin (P2),and all the regions were significantlydifferent from each other (Kruskal-Wallis one-way ANOVA, Tukey’spairwise test; H = 26.289, df = 2,P ≤ 0.001). Conversely, for the caudalfin the mean angles were not signifi-cantly different among the three re-gions (ANOVA; F = 0.0614, df = 2,P = 0.941).

At least two factors appear to con-trol scale flexibility on the body. Thefirst is a reduction in the length ofthe base relative to the length of thecrown over certain regions of thebody such as the flank in the short-fin mako (Table 1). The flank scaleshave a greater ratio of crown lengthto base length compared to the dorsalsca les (Kruskal-Wal l i s one-wayANOVA, Tukey’s pairwise test; H =18.104, df = 2, P < 0.001) due to a sig-nificantly shorter base on the flankscales (ANOVA,Holm-Sidakmethod;F = 45.967, df = 2, P = 0.001). Theflank scales have a relatively wide basecompared to its length, whereas thebase of the dorsal scales is more uni-form in shape (Table 1, BL/BW ratio;Figure 4). The ventral scales of theshortfin mako are smaller than thedorsal and flank scales as they areshorter in crown (ANOVA, Holm-Sidak method; F = 45.967, df = 2,P < 0.001) and base length (ANOVA,Holm-Sidak method; F = 42.916,df = 2, P < 0.001) overall. Secondly, rel-ative changes in the length of the lead-ing and trailing edges of the scale baseaffect its anchoring in the dermis, sim-ilar to the root system of a tree. Morefirmly anchored scales have a more

FIGURE 2

Scanning electron micrographs of the scales and histological sections of the embedded scalesfrom three regions of the pectoral fin, P1, P2, and P3 of the shortfin mako (Isurus oxyrinchus).The scales on the leading edge of the fin cannot erect and lack riblets, whereas those on the trailingedge can erect to greater angles than either the scales on the leading edge or the midregion of thefin. The bases of the scales (B) are anchored in the dermis (D), and the thin epidermis (E) is visiblebetween the scales. The pulp cavity (PC) is visible in some of the scales, and not all scales aresectioned through their center, resulting in some crown (CR) lengths appearing shorter thanothers. Ceratotrichia or fin rays (CE) are visible in region P3 because this part of the fin is verythin. Anterior is to the left.

FIGURE 3

Outline of a representative shortfinmako, Isurusoxyrinchus, showing the approximate region(lines) on the flank with the most flexible scalescapable of erection to at least 50°. This regionapproximates the region of flow separation.


evenly distributed base like a tree witha broad but evenly distributed shallowroot system (Figure 4B). The moreerectable scales on the flank have a nar-rowbase on the trailing edge (Figure 4A).In this manner, the scales can pivot up,analogous to a tree with a narrow rootsystem on one side that is blown overaway from this side. The scales returnto their resting position with the helpof elastic fibers that anchor its base(stained black in the histology sections).A sideview picture of the skin with thescales in the foreground manuallybristled is shown in Figure 5.

Effects on FluidFlow Patterns

The findings of scale flexibility andangle of erection on the shortfin mako

have led to a working hypothesis cur-rently being investigated throughhydrodynamic testing. Our discoveryof highly flexible scales on the flankand trailing edges of the pectoral finfits with the hypothesis that the scalesact as a means of controlling flow sepa-ration. First, these results indicate thatthe erection of the scales is most likelyinitiated by a passive, flow-actuatedmechanism, i.e., in a region of turbu-lent flow separation the flow consists ofmoments when the flow close to thewall is both in the main direction aswell as reversed. Pressurizing the skinhad no effect on scale erection. Flowreversal over a large region indicates

that global separation from the surface(body) occurs, resulting in increasedpressure drag. In the case of the sharkthis would not only increase drag butalso inhibit contragility or the abilityto change direction quickly and to alarge degree. As the shark swims andturns, the body is bent laterally withregions of greatest curvature occur-ring on the flank. From the nose tothe point of maximum girth (aroundthe location of the gills), the flow willexperience a favorable pressure gradi-ent, and unfavorable pressure gradientregions will be located downstream ofthis point. Thus, the findings that themost flexible scales are found on theflank of the body and downstream ofthe gills corroborate the hypothesisthat these scales are working to controlflow separation. Likewise, the delay offlow separation over the pectoral fincan lead to increased performance intheir ability to act as lifting surfacesduring high-speed swimming ma-neuvers. The generation of high liftby the pectoral fins is important forquick upward maneuvers while attack-ing prey, as has been observed in videoevidence of a shortfin mako in pursuitof a towed baitfish. This same videoevidence shows the shark’s ability toturn in one direction and then changedirection before the body completesthe initial turn. This type of turningbehavior, also defined as contragility,requires not only large muscular effortbut also low form drag (Frank Fish,

TABLE 1

Means and standard errors for scale crown length (CL), base length (BL), base width (BW), ratio of CL/BL, and ratio BL/BW for three body areas ofshortfin mako (Isurus oxyrinchus).

Body Position CL (mm) BL (mm) BW (mm) CL/BL BL/BW

B1, B4, A1 Dorsum 0.173 ± 0.004 0.145 ± 0.005 0.161 ± 0.005 1.201 ± 0.029 0.938

B2, B5, A2 Flank 0.179 ± 0.004 0.104 ± 0.003 0.161 ± 0.005 1.745 ± 0.076 0.625

B3, B6, A3 Ventrum 0.128 ± 0.004 0.091 ± 0.005 0.125 ± 0.003 1.472 ± 0.102 0.692

FIGURE 4

Representative scales of the shortfin mako(Isurus oxyrinchus) from (A) the flexibleflank area B5, and (B) the less flexible areaB4. The scales in (A) have a relatively shortbut wide base compared to those of (B) witha more uniformly shaped base.

FIGURE 5

Coronal section through the shortfin makoskin showing the scales in the foregroundthat have been manually erected from locationB2. Not all scales are erected to the samedegree because of the individual manual erec-tion. Flow would normally pass over the skinfrom left to right and reversed flow, as occursduring separation, is believed to cause bris-tling as shown.


personal communication, February18, 2008).

Flow visualization images (Fig-ure 6) of particles illuminated with alaser sheet show characteristic flow sce-narios found in a turbulent boundarylayer undergoing separation. In thiscase, separation is induced on a flatsurface via the presence of a rotatingcylinder located above the wall onwhich a turbulent boundary layer isformed. The free stream flow moves

at 17 cm/s, giving a local Reynoldsnumber, based on distance from theleading edge of 0.3 m, in the bound-ary layer of 5 × 104. In comparison, ashark’s boundary layer will have a Re of∼106-107 when swimming at 10 m/s.However, the characteristics of the tur-bulent flow will be similar; the flow ona shark will be faster and the boundarylayer thinner resulting in shorter timeand length scales than the water tunnelexperiments. Under these conditions,there are moments when the flow isfor the most part attached (Figure 6a),developing into a large region of sepa-ration with reversed flow near the wall(Figure 6b), and finally times whenlarge vortex bursting occurs as the sep-arated region becomes unstable result-ing in a shedding of vortices (Figure 6c).A time trace of the velocitymeasured ata location adjacent to the wall wherethe flow is reversed about 50% of thetime is shown in Figure 7. Here, thecyclic nature of a separating turbulentboundary is clearly evident. Regions in

the plot where the velocity is negative in-dicate moments of reversed flow,which would actuate scale bristlingon the shark. It is transitions in thetime trace of the velocity where theflow moves (Figure 7) from positive(u > 0) to negative (u < 0) (or frompoint a to point b labeled in Figure 7)when scale actuation would be initi-ated and potentially disrupt the evolv-ing flow that leads to the formationof a separation bubble as shown inFigures 6b and 6c.

Cassel et al. (1996) provide a de-scription of the process leading toflow separation. Because of the suc-tion pressure upstream (adverse pres-sure gradient) the region closest tothe wall, where the flow has the lowestmomentum, is where flow reversalis first initiated. This patch of fluidmoves upstream and thickens, ulti-mately leading to large scale flow sep-aration from the surface. It is ourhypothesis that on the shark’s bodyin the region close to the wall where

FIGURE 6

Flow visualization generated by particle streak-ing in a plane parallel to the free stream flow(16.5 cm/s) illuminated by a laser sheet. A ro-tating (40 RPM) cylinder (5.1 cm diameter) islocated above the image with its center, 6.4 cmabove the wall, to induce boundary layer sep-aration. Boundary layer thickness prior to thetest area was approximately 1 cm and an areaof approximately 3 cm × 1.5 cm is imaged.Main flow moves left to right. Distinguishingmoments in time for this unsteady flow areshown when (a) the flow is attached, (b) theseparation process is initiated with reversedflow close to wall, and (c) separated flow char-acterized by vortex bursting away from thewall is observed, with a large region of re-versed flow near the wall.

FIGURE 7

Velocity (u) measured as a function of time at a point close to the wall corresponding to approx-imately two-thirds the distance downstream in Figure 6. Measurements were made using time-resolved digital particle image velocimetry (TR-DPIV), which sampled the flow at 1 kHz. At thislocation for about 50% of the time, flow reversal is occurring, as can be seen by the regions whereu < 0. This is a point in the vicinity where flow reversal is initiated in the turbulent boundary layerand develops into a large-scale region of reversed flow due to the presence of the adverse pressuregradient induced by the rotating cylinder above. Points labeled (a), (b), and (c) aremoments in the flow typified correspondingly to those shown in Figure 6.


flow reversal begins to occur, the scalesare actuated by the flow to erect, therebydisrupting the unsteady flow sep-aration process. Future experimentswill investigate this hypothesis throughhydrodynamic testing using mod-els and real specimens of shark skin.This aspect of shark skin resulting ina surface with a preferred flow direc-tion is likely key to its ability to controlflow separation.

Previous experiments over a bris-tled shark skin model confirmed thepresence of embedded vortices form-ing between replicas of the scales(Lang et al., 2008). Thus, if flow isinduced to form between the scaleswhen bristled, there are two additionalmechanisms that may aid to controlthe flow. The formation of embeddedvortices, similar as occurs with dimpleson a golf ball, would allow the flow topass over the skin with a resulting par-tial slip condition, thereby leading tohigher momentum adjacent to the sur-face. Secondly, with a turbulent flowforming in the boundary layer abovethe cavities, there may be additionalmomentum exchange whereby highmomentum fluid typically locatedaway from the surface is induced at agreater rate to move towards the sur-face and into the cavities. This lattermechanism, resulting in turbulenceaugmentation (Gad el-Hak, 2000), isanother potential means to increasethe momentum overall in the flow ad-jacent to the wall. These three mecha-nisms may be working in conjunctionto inhibit flow separation over thesurface of the shark.

This method has advantages, aboveand beyond other methods currentlyin use to control flow separation, inthat it is passive with no energy inputrequired. Also it causes no additionaldrag penalty when not in use in that

scale bristling would be controlled byon-demand erection of the scales in-duced by regions of flow reversal asoccurs under conditions of incipientflow separation. This obviates the useof bristled scales to act as vortex gen-erators as a means of separation con-trol, as previously theorized by Bechertet al. (2000). Vortex generators, whichconsist of small, typically V-shapedprotrusions, require careful placementand protrusion into the boundary layerflow upstream of the point of flow sep-aration and work by mixing highermomentum flow down towards thesurface (Lin, 2002). Vortex generatorsalso result in a drag penalty due to theirprotrusion into the flow (Gad-el-Hak,2000). Our findings suggest that thescales are bristled passively and are ac-tivated in a region of flow reversal thatoccurs downstream of the point of sep-aration and is thus a different method-ology from that of vortex generatorscurrently in use today. Finally, thisnew passive, flow-actuated mechanismmay in fact go to the initial root causeof the separation, that of flow reversaladjacent to the wall, and disrupt itprior to growth into a fully separatedregion. Our ultimate aim is to gain afundamental understanding of howthe flow-actuated bristling of sharkskin scales can control flow separationso that bio-inspired surfaces can beengineered for greater flow control inmarine and other applications.

SummaryFast swimming shortfin mako

sharks Isurus oxyrinchus have high-scale flexibility on the flank andtrailing edges of the pectoral fin, withbristling angles up to a range of ap-proximately 50° on the flank. These re-gions correspond to those on the body

where flow separation control is likelyto be most beneficial. In the case of theflank region, high curvature of thebody will result from the shark’s lateralswimming motion, and this is alsoin the vicinity of the point of maxi-mum girth; both conditions indicatethe presence of an adverse pressuregradient. In the case of the pectoralfin, it is hypothesized that the flexiblescales can lead to the control of flowseparation which indicates high dragand loss of lift. Control of the flow inboth regions will lead to increased swim-ming speeds with high contragility forthe shark. Increased scale flexibility inthis shark appears to be due to a reduc-tion in the relative size of the scale basecompared to the crown and changesin the shape of the base where it is an-chored into the dermis. Future workwill lead to the manufacturing of bio-inspired surfaces based on shark skinmicrogeometry whereby a passive,flow-actuated surface patterning canbe used for applications where flowseparation control is required.

AcknowledgmentsFunding for this work received

through collaborative NSF grants(0932352, 0744670, and 0931787)to A. Lang, P. Motta, and R. Hueterto support both the engineering andbiological work is gratefully acknowl-edged. We also thank Jessica Davisfor assisting in the sharkmeasurementsand Candy Miranda for preparing thehistological samples. Finally, we wishto express our gratitude to Paul andJane Majeski and crew, Captain MarkSampson, Captain Al VanWormer,Philip Pegley, Lisa Natanson, JackMorris, and Mote Marine Labora-tory for assistance in obtaining sharkspecimens.


Lead Author:Amy LangDepartment of AerospaceEngineering and MechanicsUniversity of AlabamaBox 870280Tuscaloosa, AL 35487Email: [email protected]

ReferencesAnderson, E., McGillis, W., & Grosenbaugh,

M. 2001. The boundary layer of swimming

fish. J Exp Biol. 204:81-102.

Bechert, D.W., Bruse, M., Hage, W., &

Meyer, R. 2000. Fluid mechanics of biological

surfaces and their technological application.

Naturwissenschaften. 80:157-71.

Bechert, D.W., Bruse, M., Hage, W.,

Van der Hoeven, J., & Hoppe, G. 1997.

Experiments on drag-reducing surfaces and

their optimization with an adjustable

geometry. J Fluid Mech. 338:59-87.

Blake, R. 2004. Fish functional design and

swimming performance. J Fish Biol. 65:1193-222.

Bruse, M., Bechert, D., van der Hoeven, J.,

Hage, W., & Hoppe, G. 1993. Experiments

with conventional and with novel adjustable

drag-reducing surfaces. In: Proc. of the Int.

Cong. On Near-Wall Turbulent Flows.

719-738. Tempe, AZ: Elsevier Science

Publishers.

Bushnell, D., & Moore, K. 1991. Drag

reduction in nature. Annu Rev Fluid Mech.

23:65-79.

Cassel, K., Smith, F., & Walker, J. 1996. The

onset of instability in unsteady boundary-layer

separation. J Fluid Mech. 315:223-56.

Doligalski, T., Smith, C., & Walker, J. 1994.

Vortex interactions with walls. Annu Rev

Fluid Mech. 26:573-616.

Donley, J.M., Sepulveda, C.A., Konstantinidis,

P., Gemballa, S., & Shadwick, R.E. 2004.

Convergent evolution in mechanical design of

lamnid sharks and tunas. Nature. 429:61-65.

Fish, F.E. 1998. Imaginative solutions by

marine organisms for drag reduction. In:

Proceedings of the International Symposium

on Seawater Drag Reduction, ed.Meng, J.C.S.,

443-50. Newport, Rhode Island.

Fish, F. 2006. The myth and reality of

Gray’s paradox: Implication of dolphin drag

reduction for technology. Bioinspir Biomim.

1:17-25.

Fish, F., & Lauder, G. 2006. Passive and

active flow control by swimming fishes and

mammals. Annu Rev FluidMech. 38:193-224.

Gad-el-Hak, M. 2000. Flow Control: Passive,

Active and Reactive Flow Management.

Cambridge, UK: Cambridge University

Press. 421 pp.

Lang, A., Hidalgo, P., Motta, P., & Westcott,

M. 2008. Bristled shark skin: A micro-

geometry for boundary layer control?

Bioinspir Biomim. 3:046005.

Lin, J. 2002. Review of research on low-

profile vortex generators to control boundary-

layer separation. Prog Aerosp Sci. 38:389-420.

Martinez, G., Drucker, E., & Summers, A.

2002. Under pressure to swim fast. Integr

Comp Biol. 42(6):1273-4.

Naylor, G.,Martin, A.,Mattison, E., &Brown,

W. 1997. The inter-relationships of lamniform

sharks: Testing phylogenetic hypotheses with

sequence data. In: Molecular Systematics of

Fishes, eds. Kocher, T.D., & Stepien, C.,

199-217. New York: Academic Press.

NMFS. 2010. Final Amendment 3 to the

Consolidated Atlantic Highly Migratory

Species Fishery Management Plan. Silver

Spring, MD: National Oceanic and Atmo-

spheric Administration, National Marine

Fisheries Service, Office of Sustainable Fish-

eries, Highly Migratory Species Management

Division. 632 pp. Public Document.

Raschi, W., & Tabit, C. 1992. Functional

aspects of placoid scales: A review and update.

Aust J Mar Fresh Res. 43:123-147.

Reif, W. 1985. Squamation and ecology of

sharks, no. 78, Courier Forschungs-Institut

Senckenberg, Frankfurt am Main.

Schultz, W., & Webb, P. 2002. Power

requirements of swimming: Do new methods

resolve old questions? Integr Comp Biol.

42:1018-25.

Stevens, J. 2009. The biology and ecology of

the shortfin mako shark, Isurus oxyrinchus.

In: Sharks of the Open Ocean: Biology,

Fisheries and Conservation, eds. Camhi,

M.D., Pikitch, E.K., Babcock, E.A., 87-94.

Oxford, UK: Blackwell Publishing Ltd.

Vogel, S. 2003. Comparative Biomechanics:

Life’s Physical World. Princeton: Princeton

University Press.

Wainwright, S., Vosburgh, F., & Hebrank, J.

1978. Shark skin: A function in locomotion.

Science. 202:747-9.


P A P E R

Can Biomimicry and Bioinspiration ProvideSolutions for Fouling Control?A U T H O R SEmily RalstonGeoffrey SwainFlorida Institute of Technology

A B S T R A C TBiomimicry, modeling biological systems to find engineering methods, and bio-

inspiration, improving upon or repurposing the biological model, may provide directionfor the development of new antifouling solutions. Despite being subject to constantpressure from foulers, many organismsmaintain a clean surface. The challenge liesin selecting the most effective and reproducible antifouling mechanisms fromnature and mimicking or modifying them to provide a realistic engineered solution.Keywords: natural antifouling, marine coatings, biomimicry, bioinspiration, anti-fouling, foul release

Introduction

Antifouling is changing. The banon tributyltin (TBT), arguably themost effective and yet environmentallydamaging antifouling precipitated anincrease in research to develop newtechnology (Swain, 1999; Omae,2003). The immediate response wasto return to copper as the antifoulingof choice. However, high levels of cop-per in many ports and harbors has ledto concern and even the banning ofcopper based antifouling (i.e., SanDiego, Washington State; Carsonet al., 2009; Nehring, 2001; Qianet al., 2010; Thomas et al., 2001).The use of toxic coatings may also belinked to the transport of biocide toler-ant non-indigenous species (Daffornet al., 2008). Some users have switchedto biocide-free silicone antifoulingcoatings for improved hydrodynamicand environmental performance. How-ever, these coatings may foul, whichnecessitates cleaning, and there maybe an increased risk of transportingnon-native species. They are mechan-ically weak and can be damaged moreeasi ly than tradit ional coatings(Chambers et al., 2006; Nehring, 2001;Swain, 2010; Swain et al., 2007; Yebraet al., 2004). An ideal coating will behydraulically smooth and control foul-ing for the lifetime of the vessel, be

environmentally compliant, controlinvasive species, be easily applied, re-paired and maintained, be compatiblewith materials and methods of hullconstruction and decommissioning,and be cost effective (Swain, 2010).

In the decades since a ban on TBTwas proposed, research has been di-rected towards new solutions to thefouling problem. Some of the discov-eries have been in the form of newbooster biocides to improve the perfor-mance of copper based antifoulingagainst biofilms (slimes). These in-clude Irgarol 1051, Seanine 211, diu-ron, pyrithiones, and many others.Unfortunately, some of these boostersmay parallel TBT in terms of environ-mental impact. For example, Irgarol1051 is a photosynthesis inhibitorthat reduces plants’ ability to createenergy causing growth to slow, repro-duction to stop and eventually death ofthe plant. It has a long half-life anddoes not partition into sediment so itis found primarily in the water col-umn. It has been found to occur in es-tuaries, ports and harbors at elevatedlevels and is highly toxic to non-targetmarine plants (Hall et al., 1994;Thomas & Brooks, 2010; Thomas

et al., 2001; Yebra et al., 2004). Re-cently, it has been detected in watersamples collected from the Caribbean,Bermuda, Florida and Australia in thevicinity of seagrass beds and coral reefswhere the effects could be devastating(Carbery et al., 2006; Owen et al.,2002; Scarlett et al., 1999). Research-ers must be careful to avoid any chem-istry that may impact the environmentand non-target organisms. One sourceof new technology is to understandhow organisms prevent fouling andthen mimic or draw inspiration fromthese biological models to create anengineering solution.

Organisms that are unfouled oronly lightly fouled provide insightsinto the mechanisms that have evolvedto prevent surface colonization or epi-biosis (Wahl, 1989). These “clean” or-ganisms are the focus of research intobiomimetic and bioinspired solutionsto fouling on human structures. Bio-mimicry refers to the study of thestructure and function of biologicalsystems as models for the design ofengineering solutions (dictionary.com)while bioinspiration expands on bio-mimicry by not only copying or imitat-ing nature but also improving them or


repurposing the biological model foran idealized engineering solution.

This review will discuss the meth-ods by which nature controls foulingand characterize natural antifoulingand engineered solutions in terms ofchemical, physical, mechanical, behav-ioral and combined mechanisms.

Natural AntifoulingOrganisms have evolved several

different strategies to prevent fouling.These natural antifouling methods in-clude chemical, physical, mechanical,behavioral and a combination ofmore than one of the others (Bers &Wahl, 2004; Pawlik, 1992; Wahl,1989).

ChemicalChemical antifouling has a long

history of research and has been thesubject of many reviews (Armstronget al., 2000; Clare, 1996; Fusetani,2004; Omae, 2006; Pawlik, 1992;Qian et al., 2010; Raveendran &Mol, 2009; Rittschoff, 2000; andothers). To date, thousands of activenatural products have been identified(Pawlik, 1992). The activity of nat-ural chemistries includes low pH, de-terrents, anesthetics, attachment andmetamorphosis inhibitors, or toxicchemicals. The chemicals may be sur-face bound or water soluble (Omae,2006; Rittschof, 2000; Wahl, 1989).Despite issues with the ecological rele-vance of some of the chemicals, activenatural products have been isolatedfrom an algae, sponges, soft coralsand a limpet that are either avail-able at the surface or released into thewater column when the organismis disturbed (de Nys & Steinberg,2002; Fusetani, 2004; Hay, 1996;Hellio et al., 2002). The mucus of dol-phins, echinoderms, fish and corals

contain antifouling chemicals thathave been found to dissolve glue, pre-vent attachment or act as antimicrobialtoxins (Baum et al., 2002; Bavingtonet al., 2004; Ebran et al., 2000; Ritchie,2006; Shephard, 1994; Videler et al.,1999). Many tunicates have acidicbody pH and low epibiosis, espe-cially in areas where density of acidicvacuoles is high (Hirose et al., 2001;Stoeker, 1980). Additionally, whenlooking at whole animal extracts,some tunicates contain chemicals thatare cytotoxic, antimicrobial and anti-viral (Davis & Wright, 1989). Theeggs of many organisms, includingfish and coral, are well protected withantimicrobial chemistries (Marquiset al., 2005, Ramasamy & Marugan,2007). Terrestrial plants have alsoyielded interesting chemistries suchas tannins, pyrethroids and capsaicin(Feng et al., 2009; Perez et al., 2007;Xu et al., 2005).

More recently, attention has turnedto microorganisms. Antifouling meta-bolites that had been attributed in thepast to algae, sponges, corals, etc., havebeen found on closer study, to be pro-duced by surface associated bacteriaand cyanobacteria (Armstrong et al.,2000; Clare, 1996; Krug, 2006).These surface associated microorgan-isms are distinct from the communitiesin the water column, often found inhigher densities than the water columncommunity and are often highlypigmented (Dobretsov et al., 2005;Faimali et al., 2004; Holmstromet al., 1992). Deterrent biofilms pre-vent fouling by toxic or deterrentchemistries (Dobretsov et al., 2006;Pawlik, 1992).

PhysicalThe two primary physical means

identified to prevent fouling are sur-face energy and surface texture. A sur-

face energy range of 20-30 dynes/cm(Baier, 1972; Dexter, 1979) has beenshown to minimize adhesion andfavor the removal of epibionts. Suchsurface energy values have been mea-sured on the surface of killer whales(Baier & Meyer, 1986), gorgonians(Vrolijk et al., 1990) and healthyteeth (Baier & Meyer, 1986; Glantzet al., 1991). Surface energy also effectssettlement of some organisms, albeitin a species specific manner (Andersonet al., 2003;Meyer et al., 1988;Molino& Weatherbee, 2008; Rittschof &Costlow, 1989). For example, it hasbeen found that barnacles and bryozo-ans prefer to settle on different sur-face energies (Dahlstrom et al., 2004;Rittschof & Costlow, 1989). Diatomsand the green algaeUlva have differentadhesion strengths on surfaces withdifferent wettability. Diatoms are moreeasily removed from hydrophilic sur-faces, whereas Ulva releases more eas-ily from hydrophobic surfaces (Finlayet al., 2002; Kirshnan et al., 2006).Low adhesion surfaces in nature are as-sociated with waxes, oils, surfactants,mucuses or fluorinated or methylatedcompounds (Baum et al., 2003; Krug,2006; Shephard, 1994; Wahl, 1989).

The antifouling properties of sur-face topography have received ex-tensive attention and review (Scardino& deNys, 2011; Scardino et al., 2008).The effectiveness of topography asantifouling appears to be the relation-ship of scale between the texture andthe settling organism. An ultra-smoothsurface offers no refuge from predationor hydrodynamic stresses and is there-fore unattractive (Kohler et al., 1999;Walters & Wethey, 1996). Surfaceswith textures that are smaller thanthe settler reduce settlement and/orattachment strength, the “attachmentpoint theory” (Scardino et al., 2008).Textures that are the same size or


slightly larger than the propagules of-fer the greatest number of attachmentpoints, the strongest attachment andthe best protection to settling organ-isms (Callow et al., 2002; Scardinoet al., 2006). Organisms that haveonly one attachment point (barnacles,arborescent bryozoans, etc.) are morespecific in searching for high qualitypits than colonial organisms with mul-tiple attachment points, likely becausethe colonial organisms outgrow the“refuge” of the pit quickly and cansurvive partial mortality (Walters &Wethey, 1996). The use of hairs inmussel spat (Dixon et al., 1995), spic-ules in gorgonian coral (Scardino &de Nys, 2011; Vrolijk et al., 1990)and spines in some colonial organisms(Dyrynda, 1986; Wahl, 1989) hasbeen shown to prevent fouling andovergrowth of fouling organisms. Ithas been suggested that many other or-ganisms including crabs, brittle stars,molluscs, marine mammals and sharks(adult skin and dogfish egg cases), usetextured surfaces with one or morescales of complexity to prevent micro-and macro-fouling (Baum et al.,2002; Bers & Wahl, 2004; Scardino& de Nys, 2004; Scardino & de Nys,2011).

MechanicalGrooming is a common antifoul-

ing mechanism found in nature.Grooming involves specialized struc-tures that either pick or sweep ananimals surface clean (Wahl, 1989).Decapod crustaceans have highlyevolved brushes used to remove epi-bionts and parasites from specificparts of their bodies like the gills andcarapace (Acosta & Poirrier, 1992;Batang & Suzuki, 2003; Bauer, 1981).Historically, echinoderms and bryozo-ans were thought to use specializedstructures to clean their surfaces

(Campbell&Rainbow, 1977;Dyrynda,1986), but recently this has been calledinto question as the pedicellaria of thecrown of thorns starfish (Acanthasterplanci) were found to be too unrespon-sive and widely placed to be effective inkeeping their surfaces clean (Guentheret al., 2007). Many organisms useciliary cleaning in conjunction withmucus to keep surfaces clean (Wahlet al., 1998). Symbiotic or mutualis-tic relationships such as fish visiting“cleaning stations” (Poulin & Grutter,1996), mutualistic grazing of snailswithin populations (Wahl&Sonnichsen,1992; Wahl et al., 1998) and bran-chiobdellid annelids that feed on epi-bionts in the gill chamber of crayfish(Brown et al., 2002) are examples ofbeneficial relationships that may pre-vent fouling.

The other mechanical method ofantifouling is surface renewal via shed-ding or molting of outer layers. Crus-taceans, stone fish and algae all molt,either the entire surface simulta-neously or in patches, which removesall attached fouling (Bakus et al.,1986; Keats et al., 1997; Wahl, 1989).Additionally, many organisms usemucus as membrane to separate them-selves from their environment. Themucus sloughs off removing foulers,makes adhesion difficult and foulssensory and attachment apparatus ofepibionts (Brown & Bythell, 2005;Davies & Hawkins, 1998; Denny,1989; Dyrynda, 1986; Shephard,1994;Wahl, 1989;Wahl et al., 1998).

BehavioralBehavioral antifouling is the direct

or indirect active avoidance of foulingorganisms (Becker & Wahl, 1996).Burrowing into sediment, movinginto the air, between fresh and saltwater or into areas with very differentoxygen contents and nocturnal activity

or hiding in crevices are all mecha-nisms that remove less tolerant epibio-tic organisms (Becker & Wahl, 1996;Wahl, 1989; Wahl et al., 1998). Or-ganisms with a similar range of toler-ances to their hosts will be unaffectedby these behavioral methods (Brocket al., 1999).

CombinationMost organisms that are well stud-

ied use a combination of methods toprevent surface fouling. This was high-lighted in reviews by Ralston andSwain (2009) and Scardino andde Nys (2011). Crustaceans groomand shed their shells and use behavioralmechanisms like burrowing and mov-ing among habitats to ensure clean sur-faces (Becker & Wahl, 1996; Wahlet al., 1998). Echinoderms groom,slough, excrete anti-adhesive mucus,have chemical antifoulants and mayeven use a strong negatively chargedcuticle to prevent surface colonization(Bakus et al., 1986; Bavington et al.,2004; Bryan et al., 1996; McKenzie& Grigolava, 1996). Corals use anti-bacterial mucus, select specific micro-bial colonists which in turn protectthem from other microbes, sloughmucus and surface layers and secretesecondary metabolites to keep theirsurfaces clean (Brown & Bythell,2005; Ritchie, 2006; Targett et al.,1983). Dolphins and whales are hy-pothesized to use microtopographyin conjunction with an enzymaticallyactive zymogel to prevent attachmentof macrofoulers (Baum et al., 2003;Meyer & Seegers, 2004) and surfacesloughing, skin compliance and a crit-ical surface tension in the preferredrange for minimal adhesion combinedwith breaching may remove fouling atan early stage (Baum et al., 2003; Fish& Rohr, 1999; Scardino & de Nys,2011). Algae have provided many


new chemical metabolites that preventfouling, shed their outer layers andmay remove settled epibionts by flex-ing beyond what their epibionts canwithstand (Nylund & Pavia, 2005;Scardino & de Nys, 2011; Walterset al., 2003;Wikstrom&Pavia, 2004).

Biomimetic and BioinspiredEngineering SolutionsChemical

Due to our vast experience with in-corporating chemicals into coatings, itis not surprising that natural productsare the most investigated biologicalantifouling. SeaNine 211 is a boosterbiocide added to copper coatings toboost efficacy against fouling plants.It degrades quickly in water and sedi-ment, binds strongly to sediment, haslow environmental toxicity and has anexcellent performance record from laband field tests and ship trials (Thomas& Brooks, 2010; Yebra et al., 2004).SeaNine 211 is based on the naturalproduct isothiozolone originally iso-lated in the 1980s from the soft coralEunicea (Raveendran & Mol, 2009).Econea is a halogenated pyrrol that isthe active ingredient in copper-freeantifouling paints from manufacturerssuch as Petit, Interlux, Sea Hawk andothers. Halogenated pyrrols are com-mon secondary metabolites in bacteriaand sponges (or possibly surface bac-teria associated with sponges) and arepotent settlement and metamorphosisinhibitors for barnacles and otheranimal foulers (Dahms et al., 2006;Omae, 2006). Because of its specificityagainst animal fouling, Econea is oftencombined with a booster like SeaNineto prevent plant fouling as well.

Perhaps the most studied naturalchemistry is the halogenated furanoneoriginally isolated from the red algae

Delisea pulchra. The chemical is pres-ent on the surface of the plant in con-centrations that prevent fouling andthe coverage of epibionts correspondsto concentrations of the furanone(de Nys & Steinberg, 2002). It hasnot yet been successfully incorporatedinto a long-lasting ship hull coating(Chambers et al., 2006); however,some have reported that it is availableproducts called “Netsafe” and “Pearl-safe” marketed in Australia for use incommercial aquaculture (Raveendran& Mol, 2009). We were unable tofind any record of these products forsale at this time so they may no lon-ger be available. Many other naturalchemicals from marine macroorgan-isms show promise for non-toxic orlow-toxicity antifouling paints andare being investigated (see reviews byArmstrong et al., 2000; Fusetani,2004; Omae, 2006; Qian et al.,2010; Raveendran & Mol, 2009).

Terrestrial plants have also yieldedpromising chemistries for antifouling.These include products like tannins,pyrethroids and capsaicin (Feng et al.,2009; Perez et al., 2007; Thomas &Brooks, 2010; Xu et al., 2005). Pyre-throids, synthetic analogs of pyrethrinfrom chrysanthemum flowers, areof particular interest because they arealready approved for use as environ-mentally safe insecticides. These insec-ticides have low toxicity to mammals,do not persist, do not bioaccumulateand are available in industrial quanti-ties (Feng et al., 2009). Tannin is pres-ent in terrestrial plants, mangroves andin some marine algae, primarily as ananti-herbivory chemical. However,some have found it to have antifoulingproperties as well (Brock et al., 2007;Lau & Qian, 1997; Perez et al., 2007;Wikstrom & Pavia, 2004). The long-term efficacy of tannin isolated fromthe quebracho tree was improved by

precipitating it with aluminum form-ing a salt which increased the lifespan of the coating to 1 month in thefield (Perez et al., 2007).

Another avenue of research is iso-lating chemicals from microorganismsand using the microorganisms them-selves. There are many benefits tothis strategy including culturability,abundance and ability to trick or stressthe organisms into producing largequantities of the necessary chemical(Dobretsov et al., 2006; Holmstrom& Kjelleberg, 1994). Holmstrom et al.(2000) were able to keep bacteria alivein a coating for 14 days in the labora-tory. Microencapsulation is anotherstrategy being investigated, not justfor microorganisms but for all naturalproducts, as a way to increase lengthof efficacy. Coatings with micro-encapsulated living bacteria were ableto prevent fouling up to 7 weeks infield trials (Chambers et al., 2006; Yeeet al., 2007).

The use of enzymes and hormonesthat are commercially available is an-other strategy for chemical antifouling.Many patents have been awarded andthere is an enzymatic coating availableon the Danish yacht market, althoughlittle scientific evidence of effective-ness exists (Olsen et al., 2007). En-zymes may act directly by dissolvingglues, lysing cells or decomposingexoskeletons of barnacles (Abarzua &Jakubowski, 1995; Evans & Clarkson,1993; Olsen et al., 2007). They mayalso act indirectly by increasing theeffectiveness of an antifoulant or byacting on the coating to improve re-lease or polishing rates (Olsen et al.,2007). Hormones, such as noradren-aline, may also be used as a non-toxicdeterrent in antifouling coatings(Gohad et al., 2010). However, bothenzymes and hormones have severaldrawbacks to their widespread use


including expense, instability, poten-tially specific response and need forenvironmental approval (Gohad et al.,2010; Olsen et al., 2007; Rittschof,2000).

There are several challenges in get-ting new chemicals approved for use inantifouling paints. The environmentalproblems associated with TBT haveincreased awareness of the potentialrisks associated with introducing newchemistries and attention must begiven to proving environmental com-pliance. This greatly increases thetime and the cost to go from the iden-tification and isolation of a chemicalto commercialization, which can costmillions of dollars and take over10 years to get approval. Furthermore,many natural products are structurallycomplex and only available in smallamounts in the organism so there areissues with obtaining or synthesizingthe active chemicals (Fusetani, 2004;Rittschof, 2000; Rittschof, 2001).Natural products tend to have a shortlife span as they cannot be toxic to theorganism. This leads to issues whenincorporated into a coating as coat-ings need to maintain efficacy for3-12 years of service life (de Nys &Steinberg, 2002; Fusetani, 2004;Ingle, 2007; Marechal & Hellio,2009; Rittschof, 2000; Rittschof,2001). Conversely, the short life spanis beneficial for environmental compli-ance as one of the characteristics of anideal chemical antifoulant is short halflife (Clare, 1996). Other factors for anideal chemical antifoulant includenon-toxicity, activity against a widevariety of fouling organisms, easy in-corporation into a controlled releasecoating and should come from a cul-turable organism or have an activechemistry that can be industrially syn-thesized (Clare, 1996; Hellio et al.,2002; Marechal & Hellio, 2009;

Raveendran & Mol, 2009; Rittschof,2000; Yebra et al., 2004).

PhysicalSlippery coatings are not new

technology. Commercially availablefouling release coatings have beenavailable since the mid-1970s and areproving to be an increasingly success-ful method for fouling control. Thesecoatings combine polydimethylsilox-ane silicone with low surface energy,oils and compliance to reduce theadhesion strength of organisms to asurface. Fouling is removed by hydro-dynamic shear forces or with gentlecleaning (Anderson et al., 2003).There are several lanolin based waxeson the market for use on ship hullsand propellers. While the wax mayprovide short-term antifouling, themain purpose is to lessen attachmentstrength and make cleaning easier.These products are often used over atough epoxy, however the duration ofeffect is unknown and we were unableto find any published data in a peer re-viewed journal to back the claims ofmanufacturers. Results obtained fromthese coatings may vary dependingon the fouling communities. Effectsof surface energy and wettability onsurface colonization are species specificwith some responding favorably to hy-drophobic surfaces and some to hydro-philic or intermediate surfaces (Callowet al., 2002, Dahlstrom et al., 2004).Additionally primary colonizers oftenchange the surface energy of surfaceswhich will change the effect on sub-sequent settlers (Scardino & de Nys,2011).

Mimicking the surface texture ofmarine organisms has been investi-gated as an environmentally friendlyantifoulant. “Sealcoat” is a commer-cially available antifouling coatingthat is flocked with fibers mimicking

the fur of a seal. According to the com-pany’s website, fouling is prevented forup to 5 years. However, no scientificdata exists to back this claim. Otherstudies looking at flocked or furredcoatings found mixed responses withgreen and brown algae, encrustingbryozoans and barnacles deterred, redalgae and hydroids unaffected andsolitary tunicates and tube wormsincreased by these coatings (Phillippiet al., 2001). It must also be remem-bered that seals do not only dependon their fur to keep them fouling freebut also groom and spend large amountsof time out of the water.

Mimics of topographies from otherorganisms such as crustose corallinealgae, molluscs, crabs, brittle stars,soft corals and dogfish egg cases, havebeen investigated for antifouling activ-ity and have shown short-term effi-cacy in laboratory assays. Additionally,topographies from pilot whale andshark skins have been characterizedand had an antifouling activity attrib-uted to the microstructures. These ac-tive topographies range in scale from 1to 300 μm with multiple length scalesoccurring on natural surfaces (Baumet al., 2002; Bers & Wahl, 2004;Scardino & de Nys, 2004; Scardino& de Nys, 2011). The “Sharklet” isan example of a biomimetic textureused as an engineered surface to pre-vent fouling. It has performed well inlaboratory assays against Ulva sporesand Balanus amphitrite cyprids (Carmanet al., 2006; Schumacher et al., 2007).In order to improve the effect of thisand other topographies, a mathe-matical model was created calledthe “Engineered Roughness Index”;this index can also be used to predictsettlement of marine organisms onthe engineered topographies (Longet al., 2010). Engineered surfaceswith hierarchically wrinkled surfaces


have shown promising results in fieldtrials, especially against barnacles(Efimenko et al., 2009; Scardino & deNys, 2011).

Sound has been suggested as anantifouling method. However, thereare no published field test data that sci-entifically prove that it can providelong-term antifouling. This is not atruly biomimetic method as it is not re-ported as a natural antifouling mech-anism. Sound is used by competentlarvae of fish and invertebrates likecrabs to navigate to appropriate set-tlement sites (Radford et al., 2010;Simpson et al., 2008; Stanley et al.,2010). Specific habitats have differentauditory signatures and larvae can usethese to differentiate and pilot to theiradult habitats (Radford et al., 2010).Both high- and low-frequency soundwaves have been shown to be effectiveat inhibiting settlement of barnaclesand mussels (Branscomb & Rittschof,1984; Donskoy & Ludyanskiy, 1995;Guo et al., 2011). Additionally, ultra-sound waves have been used to destroybarnacle larvae via cavitation for ballastwater treatment (Seth et al., 2010).The use of low-frequency sound is lim-ited because it is audible to humansand other organisms and thereforenoise pollution is an issue. Severalcompanies worldwide (Ultrasonic An-tifouling, ASM, Sonihull and others)offer ultrasonic units that can be in-stalled that are purported to preventfouling or conversely to kill settlingfouling organisms thereby makingthem easy to remove. However, thismethod is variable in effect, with set-tlement rates ranging from 1% up to55% for low and high frequency, re-spectively (Branscomb & Rittschof,1984; Guo et al., 2011). Guo and col-leagues (2011) reported a settlementrate for barnacle cyprids in the labora-tory of about 20% for their best ultra-

sonic treatment compared to a rate ofaround 70% for the control so themethod is not perfect. Additionally,Sonihull reports changes in fish be-havior when their ultrasonic units arein use so there are noise pollution con-cerns with high-frequency sound aswell.

Physical methods of antifouling areoften inferred but seldom proved dueto challenges associated with testingliving materials. Surface topographyeffects are scale dependent (Scardinoet al., 2006; Schumacher et al., 2007)and effectiveness in fouling preventionmay vary geographically (Bers et al.,2010). Finding a universal physicalantifoulant may be difficult and theresults are often short lived, lasting amonth or less in field testing (Holmet al., 1997).

MechanicalMechanical cleaning is performed

on ships, aquaculture nets, instru-ments and other marine structureswhen they become fouled, either be-cause an antifouling coating was notused or if that coating becomes fouled.The U.S. Navy cleans their vesselswhen a set level of fouling is reachedas set out in the Naval Ships’ Techni-cal Manual (Cologer, 1984; NSTM,2006). Cleaning is reactive and hasbeen shown to speed the rate of re-colonization and increase the risk oftransport of nonindigenous species(Floerl et al., 2005). Additionally,commercially available brush cleaningdevices (i.e., SCAMP, Mini-Pamper,etc.) are harsh and may damage theantifouling coating. A new directionfor mechanical antifouling is to mimicnatural grooming. This is the idea be-hind the HullBUG (Hull BioinspiredUnderwater Grooming), an autono-mous robot that will proactively passover a hull while a ship is in port. Its

mode of action is a gentle wiping orbrushing of the surface on a frequentschedule sufficient to remove foulingat its earliest stages before it can be-come established (Borchardt, 2010;Tribou & Swain, 2010). Resultsfrom field testing of fouling releaseand copper-coated panels subjectedto grooming are so promising that fur-ther experiments and scale up on thismethod are being investigated.

Ecospeed is a commercially avail-able hull coating system. It is a toughglass flake reinforced vinyl estercoating. When combined with hullcleaning, this non-toxic coating ispurported to maintain a fouling freesurface with no repainting for up to25 years. Additionally, the coatingsmooths during cleaning, decreasingdrag. Again, no scientifically publisheddata exists to back the claims madeby the manufacturer.

Surface renewal has been at-tempted as an antifoulant for shiphull coatings. Polymers were devel-oped that hydrolyze in seawater leav-ing a clean surface as they dissolve(Candries et al., 2000). To date, how-ever, these coatings have only beensuccessful when combined with bio-cides as the rate of dissolution andthickness of the coating requiredwould be too great without the helpof toxic chemicals.

Mechanical antifouling has provento be an effective but imperfect methodof keeping submerged surfaces clean.Cleaning requires the deployment ofequipment and usually divers, whichincreases both the expense and hu-man risk factor of this antifoulingmechanism. When applied to toxiccoatings, cleaning may increase the re-lease of biocides, at least in the shortterm (Schiff et al., 2004). Cleaningmay cause damage to coatings whichincreases the rate of re-colonization


and may increase the risk of transportof invasive species (Floerl et al., 2005;Piola & Johnston, 2008). Mechanicalantifouling works better when com-bined with another antifouling methodsuch as a biocidal coating or a foulingrelease surface. Grooming, however,is proactive and more closely matchesmany of the behavioral activitiesfound in nature. Many organismsbenefit from self or mutual groomingto maintain their surfaces free offouling.

BehavioralBehavioral methods include re-

moving a vessel from the water whennot in use or moving between freshand salt water. The former is com-monly practiced by recreational boatowners; however, removing a largevessel from the water is impractical, es-pecially if that ship is frequently used.Moving vessels between fresh andsalt water, by traversing through thePanama Canal, for instance, is per-formed occasionally and has beencredited with preventing the unob-structed movement of Caribbean andPacific species between the two bodiesof water. Brock and colleagues (1999)found that moving a ship into freshwater for 9 days was sufficient to re-move 90% of fouling from the hull.However, tolerant fouling organismswill not be affected by this antifoulingmethod as shown by the survival andsubsequent introduction of the mus-sel Mytilus galloprovincialis to Oahu,Hawaii, fromWashington. The musselwas one of the 10% of fouling organ-isms remaining on the USS Missouriafter its Pacific transit and was seenspawning shortly after arrival in PearlHarbor and later found colonizingthe ballast tanks of a submarine (Apteet al., 2000).

CombinedIt is unlikely that any one antifoul-

ing mechanism will be sufficient toprevent all fouling in all situationsthat may be encountered by sub-merged structures. Indeed, everyorganism that is well studied withregards to natural antifouling uses acombination of strategies to maintaina clean surface. The most effectivecoatings in use today also use morethan one antifouling mechanism; anti-fouling coatings use a biocide com-bined with self-polishing or ablativemechanism to keep an active layer atthe surface. Fouling release coatingscombine low surface energy, oils andcompliance to maximize self-cleaning.To date, most researchers investigatinga biomimetic solution to antifoulinghave focused on only one method.However, that is beginning to changewith the recent publication of reviewsfocusing on combined antifoulingmechanisms (Ralston & Swain, 2009;Scardino & de Nys, 2011).

Bioinspired ApproachesThe examples highlighted above

represent biomimetic solutions forbiofouling control. Very little researchexists that takes lessons from natureand adapts or alters them for a truebioinspired solution. It has only beenrecently that novel uses have been pro-posed from biological models. For ex-ample, the dopamine based adhesivesystem in mussels, a common foulingorganism, has been investigated as a wayto obtain better adhesion of non-stickcoatings to a substrate. The dopamineadhesive allows testing on polyethyl-ene glycol (PEG) and other slipperypolymers where before it was not pos-sible because the polymers would notstick to anything. Those coatingsusing the bioinspired PEG-DOPA sys-tem outperformed traditional silicone

fouling release coatings in laboratoryassays, comparing both the settlementand adhesion of a common foulingdiatom and alga (Statz et al., 2006).

ConclusionsMarine organisms can achieve

long-term protection from foulingusing short-lived renewable mecha-nisms. This is attributed to using acombination of chemical, physical,mechanical and behavioral mecha-nisms. Much research has been pub-lished investigating the specific waysthat organisms maintain a clean sur-face but frequently focus on only onemechanism without considering the ef-ficacy of a holistic combined method.The challenge, for us, is to identify andselect the best natural systems to solvethe problem of biofouling. Throughimproved knowledge of natural sys-tems, we will be better able to bothmimic and innovate using biologicalmodels to find engineering solutions.Results so far have been promisingbut better interactions between biolo-gists, ecologists, engineers, chemistsand materials scientists are neededand publishing of results is vitally im-portant. Despite some issues, biomi-metics and bioinspiration hold greatpromise for new antifouling solutions.For example, the HullBUG groomingmethod now being developed by theOffice of Naval Research demonstrateshow a proactive grooming method willenhance the long-term effectivenessof the presently available commercialantifouling or fouling release surfaces.


the Office of Naval Research (grantsN000140210217, N000140810034and N000140910843), who has


funded much of this work and con-tinues to support research directedtowards discovering improved anti-fouling technology and the partici-pants in the ONR Coatings ResearchGroup.

Authors:Emily Ralston and Geoffrey SwainFlorida Institute of Technology150 W. University Blvd.,Melbourne, FL 32901Emails: [email protected];[email protected]

ReferencesAbarzua, S., & Jakubowski, S. 1995. Bio-

technological investigation for the prevention

of biofouling. I. Biological and biochemical

principles for the prevention of biofouling.

Mar Ecol Prog Ser. 123:301-12. doi: 10.3354/

meps123301.

Acosta, C., & Poirrier, M. 1992. Grooming

behavior and associated structures of the

mysid Mysidopsis bahia. J Crustacean Biol.

12:383-91. doi: 10.2307/1549032.

Anderson, C., Atlar, M., Callow, M., Candries,

M., Milne, A., & Townsin, R. 2003. The

development of foul-release coatings for

seagoing vessels. J Mar Des Operations.

B4:11-23.

Apte, S., Holland, B., Godwin, L., & Gardner,

J. 2000. Jumping ship: A stepping stone

event mediated transfer of a non-indigenous

species via a potentially unsuitable environ-

ment. Biol Invasions. 2:75-79. doi: 10.1023/

A:1010024818644.

Armstrong, E., Boyd, K., & Burgess, J. 2000.

Prevention of marine biofouling using natural

compounds from marine organisms. Biotech

Ann Rev. 6:221-41. doi: 10.1016/S1387-

2656(00)06024-5.

Baier, R. 1972. Influence of the initial sur-

face condition of materials on bioadhesion.

In: Proc. 3rd Int. Congr. Mar. Corr. Foul.

Maryland: Northwestern University Press.

Baier, R., & Meyer, A. 1986. Biosurface

chemistry for fun and profit. Chemtech.

16:178-85.

Bakus, G., Targett, N., & Schulte, B. 1986.

Chemical ecology of marine organisms.

J Chem Ecol. 12:951-87. doi: 10.1007/

BF01638991.

Batang, Z., & Suzuki, H. 2003. Gill-cleaning

mechanisms of the burrowing thalassinidean

shrimps Nihonotrypaea japonica and Upogebia

major (Crustacea: Decapoda). J Zool. 261:

69-77. doi: 10.1017/S0952836903003959.

Bauer, R. 1981. Grooming behavior and

morphology in the decapod crustacea.

J Crustacean Biol. 1:153-73. doi: 10.2307/

1548154.

Baum, C., Meyer, W., Stelzer, R., &

Fleischer, L-G. 2002. Average nanorough

skin surface of the pilot whale (Globicephala

melas, Delphinidae): Considerations on the

self-cleaning abilities based on nanoroughness.

Mar Biol. 140:653-7. doi: 10.1007/s00227-

001-0710-8.

Baum, C., Simon, F., Meyer, W., Fleischer,

L-G., Siebers, D., Kacza, J., & Seeger, J. 2003.

Surface properties of the skin of the pilot whale

Globicephala melas. Biofouling. 19:181-6.

doi: 10.1080/0892701031000061769.

Bavington, C., Lever, R., Mulloy, B.,

Grundy, M., Page, C., Richardson, N., &

McKenzie, J. 2004. Anti-adhesive glyco-

proteins in echinoderm mucus secretions.

Comp Biochem Physiol B. 139:607-17.

doi: 10.1016/j.cbpc.2004.07.008.

Becker, K., & Wahl, M. 1996. Behaviour

patterns as natural antifouling mechanisms

of tropical marine crabs. J Exp Mar Biol

Ecol. 203:245-58. doi: 10.1016/0022-

0981(96)02575-0.

Bers, A., Diaz, E., Da Gama, B., Vieira-Silva,

F., Dobretsov, S., Valdivia, N., … Wahl, M.

2010. Relevance of mytilid shell micro-

topographies for fouling defence—A global

comparison. Biofouling. 26:367-77.

doi: 10.1080/08927011003605888.

Bers, A., & Wahl, M. 2004. The influence of

natural surface microtopographies on fouling.

Biofouling. 20:43-51. doi: 10.1080/

08927010410001655533.

Borchardt, J. 2010. Grooming the fleet.

Mech Eng. 132:33-35.

Branscomb, E., & Rittschof, D. 1984. An

investigation of low frequency sound waves

as a means of inhibiting barnacle settlement.

J ExpMar Biol Ecol. 79:149-54. doi: 10.1016/

0022-0981(84)90215-6.

Brock, E., Nylund, G., & Pavia, H. 2007.

Chemical inhibition of barnacle larval settle-

ment by the brown alga Fucus vesiculosus.

Mar Ecol Prog Ser. 337:165-74. doi: 10.3354/

meps337165.

Brock, R., Bailey-Brock, J., & Goody, J.

1999. A case study of the efficacy of freshwater

immersion in controlling introduction of

alien marine fouling communities: The USS

Missouri. Pac Sci. 53:223-31.

Brown, B., & Bythell, J. 2005. Perspectives on

mucus secretion in reef corals. Mar Ecol Prog

Ser. 296:291-309. doi: 10.3354/meps296291.

Brown, B., Creed, R., & Dobson, W. 2002.

Branchiobdellid annelids and their crayfish

hosts: Are they engaged in a cleaning symbiosis?

Oecologia. 132:250-5. doi: 10.1007/s00442-

002-0961-1.

Bryan, P., Rittschof, D., & McClintock, J.

1996. Bioactivity of echinoderm ethanolic

body-wall extracts: An assessment of marine

bacterial attachment and macroinvertebrate

larval settlement. J Exp Mar Biol Ecol. 196:

79-96. doi: 10.1016/0022-0981(95)00124-7.

Callow, M., Jennings, A., Brennan, A.,

Seegert, C., Gibson, A., Wilson, L., …

Callow, J. 2002. Microtopographic cues for

settlement of zoospores of the green fouling

alga Enteromorpha. Biofouling. 18:237-45.

doi: 10.1080/08927010290014908.

Campbell, A., & Rainbow, P. 1977. The

role of pedicellariae in preventing barnacle

settlement on the sea-urchin test. Mar Behav

Physiol. 4:253-60. doi: 10.1080/

10236247709386957.

Candries, M., Atlar, M., & Anderson, C.

2000. Considering the use of alternative


antifoulings: The advantages of foul-release

systems. In: Conference Proceedings, ENSUS

2000. pp. 88-95. Newcastle, UK: University

of Newcastle-upon-Tyne.

Carbery, K., Owen, R., Frickers, T., Otero,

E., & Readman, J. 2006. Contamination of

Caribbean coastal waters by the antifouling

herbicide Irgarol 1051. Mar Pollut Bull.

52:635-44. doi: 10.1016/j.marpolbul.2005.

10.013.

Carman, M., Estes, T., Feinberg, A.,

Schumacher, J., Wilkerson, W., Wilson, L.,…

Brennan, A. 2006. Engineered antifouling

microtopographies—correlating wettability

with cell attachment. Biofouling. 22:11-21.

doi: 10.1080/08927010500484854.

Carson, R., Damon, M., Johnson, L., &

Gonzalez, J. 2009. Conceptual issues in

designing a policy to phase out metal-based

antifouling paints on recreational boats in

San Diego Bay. J Environ Manag. 90:2460-8.

doi: 10.1016/j.jenvman.2008.12.016.

Chambers, L., Stokes, K., Walsh, F., &

Wood, R. 2006. Modern approaches to

marine antifouling coatings. Surf Coat Tech.

201:3642-52. doi: 10.1016/j.surfcoat.2006.

08.129.

Clare, A. 1996. Marine natural product

antifoulants: Status and potential. Biofouling.

9:211-29. doi: 10.1080/08927019609378304.

Cologer, C. 1984. Six year interaction of

underwater cleaning with copper based anti-

fouling paints on Navy surface ships. Nav Eng J.

96:200-8. doi: 10.1111/j.1559-3584.1984.

tb01829.x.

Dafforn, K., Glasby, T., & Johnston, E.

2008. Differential effects of tributyltin and

copper antifoulants on recruitment of non-

indigenous species. Biofouling. 24:23-33.

doi: 10.1080/08927010701730329.

Dahlstrom, M., Jonsson, H., Jonsson, P.,

& Elwing, H. 2004. Surface wettability as a

determinant in the settlement of the barnacle

Balanus improvisus (Darwin). J Exp Mar Biol

Ecol. 305: 223-32. doi: 10.1016/j.jembe.

2003.12.013.

Dahms, H., Ying, X., & Pfeiffer, C. 2006.

Antifouling potential of cyanobacteria:

A mini-review. Biofouling. 22:317-27.

doi: 10.1080/08927010600967261.

Davis, A., & Wright, A. 1989. Interspecific

differences in fouling of two congeneric ascidians

(Eudistoma olivaceum and E. capsulatum): Is

surface acidity an effective defense? Mar Biol.

102:491-7. doi: 10.1007/BF00438350.

Davies, M., & Hawkins, S. 1998. Mucus

from marine molluscs. In: Advances in Marine

Biology, eds. Blaxter, J., Southward, A., &

Tyler, P., 1-71. San Diego: Academic Press.

Denny, M. 1989. Invertebrate mucous secre-

tions: Functional alternatives to vertebrate

paradigms. Symp Soc Exp Biol. 43:337-66.

de Nys, R., & Steinberg, P. 2002. Linking

marine biology and biotechnology. Curr Opin

Biotech. 13:244-8. doi: 10.1016/S0958-

1669(02)00311-7.

Dexter, S. 1979. Influence of substratum

critical surface tension on bacterial adhesion—

in situ studies. J Colloid Interf Sci. 70:

346-54. doi: 10.1016/0021-9797(79)90038-9.

Dixon, D., Sole-Cava, A., Pascoe, P., &

Holland, P. 1995. Periostracal adventitious

hairs on spat of the mussel Mytilus edulis.

J Mar Biol Assoc UK. 75:363-72.

doi: 10.1017/S0025315400018233.

Dobretsov, S., Dahms, H., & Qian, P. 2006.

Inhibition of biofouling by marine micro-

organisms and their metabolites. Biofouling.

22:43-54. doi: 10.1080/08927010500504784.

Dobretsov, S., Dahms, H-U., Tsoi, M., &

Qian, P-Y. 2005. Chemical control of epibiosis

by Hong Kong sponges: The effect of sponge

extracts on micro- and macrofouling com-

munities. Mar Ecol Prog Ser. 297:119-29.

doi: 10.3354/meps297119.

Donskoy, D., & Ludyanskiy, M. 1995. Low

frequency sound as a control measure for zebra

mussel fouling. In: Proceedings of the 5th

International Zebra Mussel and Other Aquatic

Nuisance Organisms Conference. pp. 103-112.

Pembroke, Ontario: Professional Edge.

Dyrynda, P. 1986. Defensive strategies of

modular organisms. Philos T Roy Soc Lon B.

313:227-43. doi: 10.1098/rstb.1986.0035.

Ebran, N., Julien, S., Orange, N., Auperin,

B., & Molle, G. 2000. Isolation and charac-

terization of novel glycoproteins from fish

epidermal mucus: Correlation between their

pore-forming properties and their antibacterial

activities. Biochim Biophys Acta. 1467:271-80.

doi: 10.1016/S0005-2736(00)00225-X.

Efimenko, K., Finlay, J., Callow, M., Callow,

J., & Genzer, J. 2009. Development and

testing of hierarchically wrinkled coatings

for marine antifouling. ACS Appl Mat Int.

1:1031-40. doi: 10.1021/am9000562.

Evans, L., & Clarkson, N. 1993. Antifouling

strategies in the marine environment. J Appl

Bacteriol Symp Suppl. 74:119S-24S.

Faimali, M., Garaventa, F., Terlizzi, A.,

Chiantore, M., & Cattaneo-Vietti, R. 2004.

The interplay of substrate nature and biofilm

formation in regulating Balanus amphitrite

Darwin, 1854 larval settlement. J Exp Mar

Biol Ecol. 306:37-50. doi: 10.1016/j.jembe.

2003.12.019.

Feng, D., Ke, C., Li, S., Lu, C., & Guo, F.

2009. Pyrethroids as promising marine anti-

foulants: Laboratory and field studies. Mar

Biotechnol. 11:153-60. doi: 10.1007/s10126-

008-9130-9.

Finlay, J., Callow, M., Ista, L., Lopez, G.,

& Callow, J. 2002. The influence of surface

wettability on the adhesion strength and

settled spores of the green algae Enteromorpha

and the diatom Amphora. Integr Comp Biol.

42:1116-22. doi: 10.1093/icb/42.6.1116.

Fish, F., & Rohr, J. 1999. Review of

dolphin hydrodynamics and swimming

performance. Technical Report no. 1801. SSC.

136 pp.

Floerl, O., Inglis, G., & Marsh, H. 2005.

Selectivity in vector management and investi-

gation of the effectiveness of measures used to

prevent transport of non-indigenous species.

Biol Invasions. 7:459-75. doi: 10.1007/

s10530-004-4863-5.


Fusetani, N. 2004. Biofouling and antifoul-

ing. Nat Prod Res. 21:94-104. doi: 10.1039/

b302231p.

Glantz, P., Baier, R., Attstrom, R., Meyer, A.,

& Gucinski, H. 1991. Comparative clinical

wettability of teeth and intraoral mucosa.

J Adhes Sci Technol. 5:401-8. doi: 10.1163/

156856191X00459.

Gohad, N., Shah, N., Metters, A., & Mount,

A. 2010. Noradrenaline deters marine inver-

tebrate biofouling when covalently bound

in polymeric coatings. J Exp Mar Biol Ecol.

394:63-73. doi: 10.1016/j.jembe.2010.07.014.

Guenther, J., Heimann, K., & de Nys, R.

2007. Pedicellariae of the crown-of-thorns sea

star Acanthaster planci are not an effective

defence against fouling. Mar Ecol Prog Ser.

340:101-108. doi: 10.3354/meps340101.

Guo, S., Lee, H., Chaw, K., Miklas, J., Teo,

S., Dickinson, G., … Khoo, B. 2011. Effect

of ultrasound on cyprids and juvenile barna-

cles. Biofouling. 27:185-92. doi: 10.1080/

08927014.2010.551535.

Hall, J., Giddings, J., Solomon, K., & Balcomb,

R. 1994. An ecological risk assessment for the

use of Irgarol 1051 as an algaecide for anti-

fouling paints. Crit Rev Toxicol. 29:367-437.

Hay, M. 1996. Marine chemical ecology:

What’s known and what’s next? J Exp Mar

Biol Ecol. 200:103-34. doi: 10.1016/S0022-

0981(96)02659-7.

Hellio, C., Berge, J., Beaupoil, C., Le Gal, Y.,

& Bourgougnon, N. 2002. Screening of

marine algal extracts for anti-settlement

activities against microalgae and macroalgae.

Biofouling. 18:205-15. doi: 10.1080/

08927010290010137.

Hirose, E., Yamashiro, H., & Mori, Y. 2001.

Properties of tunic acid in the ascidian Phallusia

nigra (Ascidiidae, Phlebobranchia). Zool Sci.

18:309-14. doi: 10.2108/zsj.18.309.

Holm, E., Cannon, G., Roberts, D., Schmidt,

A., Sutherland, J., & Rittschof, D. 1997.

The influence of initial surface chemistry on

development of the fouling community at

Beaufort, North Carolina. J Exp Mar Biol

Ecol. 215:189-203. doi: 10.1016/S0022-

0981(97)00040-3.

Holmstrom, C., & Kjelleberg, S. 1994. The

effect of external biological factors on settle-

ment of marine invertebrate and new anti-

fouling technology. Biofouling. 8:147-60.

doi: 10.1080/08927019409378269.

Holmstrom, C., Rittschof, D., & Kjelleberg,

S. 1992. Inhibition of settlement by larvae

of Balanus amphitrite and Ciona intestinalis

by a surface-colonizing marine bacterium.

Appl Environ Microb. 58:2111-5.

Holmstrom, C., Steinberg, P., Christov, V.,

Christie, G., & Kjelleberg, S. 2000. Bacteria

immobilised in gels: Improved methodologies

for antifouling and biocontrol applications.

Biofouling. 15:109-17. doi: 10.1080/

08927010009386302.

Ingle, M. 2007. Naval Sea Systems Command

Antifouling Program. Presented at the National

Paint & Coatings Association International

Marine and Offshore Coatings Expo. Virginia

Beach, VA.

Keats, D., Knight, M., & Pueschel, C. 1997.

Antifouling effects of epithallial shedding in

three crustose coralline algae (Rhodophyta,

Coralinales) on a coral reef. J Exp Mar Biol

Ecol. 213:281-93. doi: 10.1016/S0022-

0981(96)02771-2.

Kohler, J., Hansen, P., & Wahl, M. 1999.

Colonization patterns at the substratum-water

interface: How does surface microtopography

influence recruitment patterns of sessile

organisms? Biofouling. 14:237-48. doi: 10.1080/

08927019909378415.

Krishnan, S., Wang, N., Ober, C., Finlay, J.,

Callow, M., Callow, J., … Fischer, D. 2006.

Comparison of the fouling release properties

of hydrophobic fluorinated and hydrophilic

PEGylated block copolymer surfaces: Attach-

ment strength of the diatom Navicula and

the green algae Ulva. Biomacromolecules.

7:1449-62. doi: 10.1021/bm0509826.

Krug, P. 2006. Defense of benthic invertebrates

against surface colonization by larvae: A

chemical arms race. In: Progress in Molecular

and Subcellular Biology Subseries Marine

Molecular Biotechnology, eds. Fusetani, N.,

& Clare, A., 1-53. Berlin Heidelberg: Springer-

Verlag.

Lau, S., & Qian, P. 1997. Phlorotannins

and related compounds as larval settlement

inhibitors of the tube-building polychaete

Hydroides elegans. Mar Ecol Prog Ser.

159:219-27. doi: 10.3354/meps159219.

Long, C., Schumacher, J., Robinson, P., III,

Finlay, J., Callow, M., Callow, J., & Brennan,

A. 2010. A model that predicts the attachment

behavior of Ulva linza zoospores on surface

topography. Biofouling. 26:411-9. doi: 10.1080/

08927011003628849.

Marechal, J-P., & Hellio, C. 2009. Challenges

for the development of new non-toxic anti-

fouling solutions. Int J Mol Sci. 10:4623-37.

doi: 10.3390/ijms10114623.

Marquis, C., Baird, A., deNys, R., Holmstrom,

C., & Koziumi, N. 2005. An evaluation

of the antimicrobial properties of the eggs

of 11 species of scleractinian corals. Coral

Reefs. 24:248-53. doi: 10.1007/s00338-005-

0473-7.

McKenzie, J., & Grigolava, I. 1996. The

echinoderm surface and its role in prevent-

ing microfouling. Biofouling. 10:261-72.

doi: 10.1080/08927019609386285.

Meyer, A., Baier, R., & King, R. 1988. Initial

fouling of nontoxic coatings in fresh, brackish,

and sea water. Can J Chem Eng. 66:55-62.

doi: 10.1002/cjce.5450660108.

Meyer, W., & Seegers, U. 2004. A pre-

liminary approach to epidermal antimicrobial

defense in the Delphinidae. Mar Biol.

144:841-4. doi: 10.1007/s00227-003-

1256-8.

Molino, P., & Weatherbee, R. 2008. The

biology of biofouling diatoms and their role

in the development of microbial slimes.

Biofouling. 24:365-79. doi: 10.1080/

08927010802254583.

Naval Ships’ Technical Manual. 2006.

Chapter 081—Waterborne underwater hull

cleaning of Navy ships. Publication # S9086-

CQ-STM-010/CH-081Revision 5.Washington,

DC: Naval Sea Systems Command.


Nehring, S. 2001. After the TBT era: Alter-

native anti-fouling paints and their ecological

risks. Senck Marit. 31:341-51.

Nylund, G., & Pavia, H. 2005. Chemical

versus mechanical inhibition of fouling in the

red alga Dilsea carnosa. Mar Ecol Prog Ser.

299:111-21. doi: 10.3354/meps299111.

Olsen, S., Pedersen, L., Laursen, M., Kiil, S.,

& Dam-Johansen, K. 2007. Enzyme-based

antifouling coatings: A review. Biofouling.

23:369-83. doi: 10.1080/08927010701566384.

Omae, I. 2003. Organotin antifouling paints

and their alternatives. Appl Organometal

Chem. 17:81-105. doi: 10.1002/aoc.396.

Omae, I. 2006. General aspects of natural

products antifoulants in the environment.

Handbook Env Chem. 5:227-62. doi: 10.1007/

698_5_057.

Owen, R., Knap, A., Toaspern,M.,&Carbery,

K. 2002. Inhibition of coral photosynthesis

by the antifouling herbicide Irgarol 1051. Mar

Pollut Bull. 44:623-32. doi: 10.1016/S0025-

326X(01)00303-4.

Pawlik, J. 1992. Chemical ecology of the

settlement of benthic marine invertebrates.

Oceanogr Mar Biol Annu Rev. 30:273-335.

Perez, M., Garcia, M., Blustein, G., & Stupak,

M. 2007. Tannin and tannate from the

quebracho tree: An eco-friendly alternative

for controlling marine biofouling. Biofouling.

23:151-9. doi: 10.1080/08927010701189484.

Phillippi, A., O’Connor, N., Lewis, A., &

Kim, Y. 2001. Surface flocking as a possible

anti-biofoulant. Aquaculture. 195:225-38.

doi: 10.1016/S0044-8486(00)00556-1.

Piola, R., & Johnston, E. 2008. The potential

for translocation of marine species via small

scale disruptions to antifouling surfaces.

Biofouling. 24:145-55. doi: 10.1080/

08927010801930480.

Poulin, R., & Grutter, A. 1996. Cleaning

symbioses: Proximate and adaptive explanations.

Bioscience. 46:512-7. doi: 10.2307/1312929.

Qian, P., Xu, Y., & Fusetani, N. 2010.

Natural products as antifouling compounds:

Recent progress and future perspectives.

Biofouling. 26:223-34. doi: 10.1080/

08927010903470815.

Radford, C., Stanley, J., Tindle, C.,

Montgomery, J., & Jeffs, A. 2010. Localised

coastal habitats have distinct underwater

sound signatures. Mar Ecol Prog Ser.

401:21-9. doi: 10.3354/meps08451.

Ralston, E., & Swain, G. 2009. Bioinspira-

tion—the solution for biofouling control?

Bioinspir Biomim. 4:015007. doi: 10.1088/

1748-3182/4/1/015007.

Ramasamy, M., & Murugan, A. 2007. Foul-

ing deterrent chemical defense in three muricid

gastropod egg masses from the Southeast

coast of India. Biofouling. 23:259-65.

doi: 10.1080/08927010701278857.

Raveendran, T., & Mol, V. 2009. Natural

products antifoulants. Curr Sci. 97:508-20.

Ritchie, K. 2006. Regulation of microbial

populations by coral surface mucus and

mucus-associated bacteria. Mar Ecol Prog Ser.

322:1-14. doi: 10.3354/meps322001.

Rittschof, D. 2000. Natural product anti-

foulants: One perspective on the challenges

related to coating development. Biofouling.

15:119-27. doi: 10.1080/08927010009386303.

Rittschof, D. 2001. Natural product anti-

foulants and coatings development. In:

Marine Chemical Ecology, eds. McClintock,

J., & Baker, B., 543-66. Boca Raton, Florida:

CRC Press.

Rittschof, D., & Costlow, J., Jr. 1989.

Bryozoan and barnacle settlement in relation

to initial surface wettability: A comparison

of laboratory and field studies. Topics Mar

Biol. 53:411-6.

Scardino, A., & de Nys, R. 2004. Fouling

deterrence on the bivalve shell Mytilus gallo-

provincialis: A physical phenomenon. Biofouling.

20:249-57. doi: 10.1080/08927010400016608.

Scardino, A., & de Nys, R. 2011. Mini-

review: Biomimetic models and bioinspired

surfaces for fouling control. Biofouling. 27:

73-86. doi: 10.1080/08927014.2010.536837.

Scardino, A., Guenther, J., & de Nys, R.

2008. Attachment point theory revisited: The

fouling response to a microtextured matrix.

Biofouling. 24:45-53. doi: 10.1080/

08927010701784391.

Scardino, A., Harvey, E., & de Nys, R. 2006.

Testing attachment point theory: Diatom

attachment on microtextured polyamide bio-

mimetics. Biofouling. 22:55-60. doi: 10.1080/

08927010500506094.

Scarlett, A., Donkin, P., Fileman, T., &

Morris, R. 1999. Occurrence of the antifoul-

ing herbicide, Irgarol 1051, within coastal-

water seagrasses from Queensland, Australia.

Mar Pollut Bull. 38:687-91. doi: 10.1016/

S0025-326X(99)00003-X.

Schiff, K., Diehl, D., & Valkirs, A. 2004.

Copper emissions from antifouling paint

on recreational vessels. Mar Pollut Bull.

48:371-7. doi: 10.1016/j.marpolbul.2003.

08.016.

Schumacher, J., Aldred, N., Callow, M.,

Finlay, J., Callow, J., Clare, A., & Brennan, A.

2007. Species-specific engineered antifouling

topographies: Correlations between the settle-

ment of algal zoospores and barnacle cyprids.

Biofouling. 23:307-17. doi: 10.1080/

08927010701393276.

Seth, N., Chakravarty, P., Khandeparker, L.,

Anil, A., & Pandit, A. 2010. Quantifi-

cation of the energy required for the

destruction of Balanus Amphitrite larva

by ultrasonic treatment. J Mar Biol Assoc

UK. 90:1475-82. doi: 10.1017/

S0025315409991548.

Shephard, K. 1994. Functions for fish

mucus. Rev Fish Biol Fisher. 4:401-29.

doi: 10.1007/BF00042888.

Simpson, S., Meekan, M., Jeffs, A.,

Montgomery, J., & McCauley, R. 2008.

Settlement-stage coral reef fish prefer the

higher-frequency invertebrate-generated

audible component of reef noise. Anim

Behav. 75:1861-8. doi: 10.1016/j.anbehav.

2007.11.004.

Stanley, J., Radford, C., & Jeffs, A. 2010.

Induction of settlement in crab megalopae by

ambient underwater reef sound. Behav Ecol.

21:113-20. doi: 10.1093/beheco/arp159.


Statz, A., Finlay, J., Dalsin, J., Callow, M.,

Callow, J., & Messersmith, P. 2006. Algal

antifouling and fouling release properties of

metal surfaces coated with a polymer inspired

by marine mussels. Biofouling. 22:391-9.

doi: 10.1080/08927010601004890.

Stoeker, D. 1980. Relationships between

chemical defense and ecology in benthic

ascidians. Mar Ecol Prog Ser. 3:257-65.

doi: 10.3354/meps003257.

Swain, G. 1999. Redefining antifouling

coatings. J Prot Coat Linings. 16:26-35.

Swain, G. 2010. The importance of ship

hull coatings and maintenance as drivers for

environmental sustainability. In: Ship Design

and Operation for Environmental Sustain-

ability. pp 8. London: Royal Institute of

Naval Architects.

Swain, G., Kovach, B., Touzot, A., Casse, F.,

& Kavanagh, C. 2007. Measuring the

performance of today’s antifouling coatings.

J Ship Prot. 23:164-170.

Targett, N., Bishop, S., O’Connell, O., &

Yoder, J. 1983. Antifouling agents against the

benthic marine diatom, Navicula salinicola

Homarine from the gorgonians Leptogorgia

virgulata and L. setacea and analogs. J Chem

Ecol. 9:817-29. doi: 10.1007/BF00987807.

Thomas, K., & Brooks, S. 2010. The envi-

ronmental fate and effects of antifouling paint

biocides. Biofouling. 26:73-88. doi: 10.1080/

08927010903216564.

Thomas, K., Fileman, T., Readman, J., &

Waldock, M. 2001. Antifouling paint booster

biocides in the UK coastal environment and

potential risks of biological effects. Mar

Pollut Bull. 42:677-88. doi: 10.1016/S0025-

326X(00)00216-2.

Tribou, M., & Swain, G. 2010. The use of

proactive in-water grooming to improve the

performance of ship hull antifouling coatings.

Biofouling. 26:47-56. doi: 10.1080/

08927010903290973.

Videler, H., Geertjes, G., & Videler, J. 1999.

Biochemical characteristics and antibiotic

properties of the mucus envelope of the

queen parrotfish. J Fish Biol. 54:1124-7.

doi: 10.1111/j.1095-8649.1999.tb00864.x.

Vrolijk, N., Targett, N., Baier, R., & Meyer,

A. 1990. Surface characterisation of two

gorgonian coral species: Implications for a

natural antifouling defence. Biofouling.

2:39-54. doi: 10.1080/08927019009378128.

Wahl, M. 1989. Marine epibiosis. I. Fouling

and antifouling: Some basic aspects. Mar

Ecol Prog Ser. 58:175-89. doi: 10.3354/

meps058175.

Wahl, M., Kroger, K., & Lenz, M. 1998.

Non-toxic protection against epibiosis.

Biofouling. 12:205-26. doi: 10.1080/

08927019809378355.

Wahl, M., & Sonnichsen, H. 1992. Marine

epibiosis. IV. The periwinkle Littorina littorea

lacks typical antifouling defences—why are

some populations so little fouled? Mar Ecol

Prog Ser. 88:225-35. doi: 10.3354/meps088225.

Walters, L., Smith, C., & Hadfield, M. 2003.

Recruitment of sessile marine invertebrates

on Hawaiian macrophytes: Do pre-settlement

or post-settlement processes keep plants free

from fouling? Bull Mar Sci. 72:813-39.

Walters, L., & Wethey, D. 1996. Settlement

and early post-settlement survival of sessile

marine invertebrates on topographically

complex surfaces: The importance of refuge

dimensions and adult morphology. Mar

Ecol Prog Ser. 137:161-71. doi: 10.3354/

meps137161.

Wikstrom, S., & Pavia, H. 2004. Chemical

settlement inhibition versus post-settlement

mortality as an explanation for differential

fouling of two congeneric seaweeds. Oecologia.

138:223-30. doi: 10.1007/s00442-003-1427-9.

Xu, Q., Barrios, C., Cutright, T., & Newby,

B. 2005. Assessment of antifouling effec-

tiveness of two natural product antifoulants

by attachment study with freshwater bacteria.

Environ Sci Pollut Res. 12:278-84.

doi: 10.1065/espr2005.04.244.

Yebra, D., Kiil, S., &Dam-Johansen, K. 2004.

Antifouling technology—past, present and

future steps towards efficient and environmen-

tally friendly antifouling coatings. Prog Org

Coat. 50: 75-104. doi: 10.1016/j.porgcoat.

2003.06.001.

Yee, L., Holmstrom, C., Fuary, E., Lewin, N.,

Kjelleberg, S., & Steinberg, P. 2007. Inhibi-

tion of fouling by marine bacteria immobilised

in K-carrageenan beads. Biofouling. 23:

287-94. doi: 10.1080/08927010701366280.


B O O K R E V I E W

Sex, Drugs, and Sea Slime: The Oceans’Oddest Creatures and Why They MatterBy Ellen PragerUniversity of Chicago Press, April 15, 2011 (International Publication: May 15, 2011)184 pp., $26.00 (Hardcover)

Reviewed by Jason GoldbergU.S. Fish and Wildlife Service

When it comes to reviewing this book,please let me be candid about a personalbias: as a volunteer at the Smithsonian’sNational Museum of Natural History,I’ve been known to give tours focusedon the Sant Ocean Hall’s oddities, suchas the two-horned narwhal skull, giantsquid, and the seadevil. There’s a methodto such madness: if you can capture a layaudience’s attention with the exciting andunusual, they may be more receptive todiscussing more serious messages aboutthe importance of our oceans. With Sex,Drugs, and Sea Slime, Dr. Ellen Prager hasunquestionably written one of the morebizarre and fascinating books in ocean lit-erature. Combine Tina Fey with CarlSagan or perhaps Mel Brooks and RachelCarson and you have this book. Prager’swriting is reminiscent of the works ofMary Roach, Anthony Aveni, andMichael Shermer, all of whom also havea talent for writing about science’s odderside. To the best of my knowledge,however, none has written as eloquentlyabout, as Prager calls it, the lobster’s“Super Soaker Pee Blaster.” As she writes,the purpose of her book is to be “a briefand entertaining look at some of theoceans’ most fascinating creatures, theirunusual tactics for survival, and their in-valuable links to humankind. The endgoal is to showcase the importance ofthe great diversity of life in the sea, whyit is at risk, and why we should all care.”With this book, she has succeeded excep-tionally well in achieving her goals.

Everyone in the Marine TechnologySociety likely has some favorite story of abizarre creature or other factoid aboutthe oceans they learned while in schoolor on the job. Prager has thoroughly re-searched and captured the best of thesetales. More importantly, what she hasreally done is write entertainingly aboutwhy the ocean is relevant regardless ofwhere you might live. She has an abilityto wax poetic about the ocean’s strangestdenizens, whether it is the unassumingdinoflagellate or the majestic humpbackwhale. The range of material she coversin such a slender volume is really quite as-tonishing. I often found myself wonderingas I read the book whether she would coverthis species or that, and inevitably she did.Plankton, hagfishes, corals, eels, parrotfish,conesnails, cephalopods, kelp, andmore—they’re all in here, along with all the(copious) mucus and other excretionsthey produce. As the book’s title indicates,there’s also plenty of sex, as well as a few sexchanges. For those of you who might bewondering, yes, she included the pearlfish,a species that provides clear proof thatevolution has a sense of humor.

Some chapters focus on specific spe-cies, such as those on plankton or thedenizens of a coral reef, while others targetspecialized functions, such as species thatmight compete in the “X-Games” or livein extreme environments. The descriptionsof the species and their unusual habits arealways entertaining and sometimes laugh-out-loud funny. She then turns serious at

the end of each chapter when she coverswhy these species matter. I was verypleased to see that Prager doesn’t justcover the usual reasons, such as food,drugs, and recreation, but that she high-lights things many people don’t realize,such as the value of corals in mitigatingstorm damages. She concludes the bookon a more serious but optimistic note,highlighting the dangers that still lurk inocean conservation and suggesting actionswe can all take so we can continue to enjoythe ocean for generations to come.

Overall, Prager’s work makes for fasci-nating reading for anyone interested inmarine science. While it does sometimesget a little technical, it is certainly ap-propriate for lay audiences. It’s also an in-valuable resource for anyone who talksabout the oceans and needs some good ref-erences to spice up their talk, although theinclusion of an index would have beenbeneficial for such purposes. If you wantto get another sense of her book, you candownload a free National Public Radiopodcas t f rom ht tp : / /www.npr .o rg /2011/04/07/135043954/under-the-sea-sex-is-slimy-business. The same combi-nation of titillating humor and practicaldiscussion of the ocean’s value is dem-onstrated in Prager’s interview.

Our livelihoods depend on the ocean,and the talent that has helped develop andeffectively use technology is extraordinary.It is therefore incumbent upon each of usto be able to talk in some way with thepublic and decision-makers about how


the well-being of everyone living on TerraFirma relies on blue, brown, or whitewater. Books such as Prager’s offer an out-line that many of us can use to foster a dis-cussion about our own respective fields.Yes, the book is about weird science. Thegiggle factor is unmistakable. Even so, eachtime it gets weird, it casts light on the won-ders of the ocean andmakes that weirdnessimportant and meaningful. Perhaps forthat reason, more than any other, thebook is highly recommended.


The MTS Journal

includes technical

papers, notes and

commentaries

of general interest in

most issues.

Contact Managing EditorAmy Morgante

([email protected])

if you have material you

would like considered.

Specifications for

submitting a manuscript

are on the MTS website

under the Publications menu.

www.mtsociety.org

E-mail:

[email protected]

Or visit our homepage at

www.mtsociety.org

UPCOMING MTS JOURNAL ISSUES CALL FOR PAPERS

September/October 2011General Issue

November/December 2011Legacy Underwater Munitions: Assessment, Evaluation of Impacts, and Potential Response Technologies Guest Editors: Geoffrey Carton and Terrance Long

Check the Society website for future Journal topics.

www.mtsociety.org

Open Positions?

The MTS Career Center is an online resource to help employers and job seekers make career connections in the marine technology industry. Career professionals bring fresh ideas and new energy, and can keep your company current with modern trends and tactics.

The MTS Career Center Provides:• Unmatched, Targeted Exposure for Job Postings MTS represents the largest audience of quali ed marine technology

industry professionals.• Easy Online Job & Applicant Management You can enter job descriptions, check the status of postings, renew or

discontinue postings and even make payments online. Plus, you can easily manage your applicants through automatic email noti cations and applicant history access.

• Resume Database Access With a paid job posting, you can search the resume database and

use an automatic noti cation system to receive email noti cations when new resumes match your criteria.

• Employment Branding Opportunities Along with each job posting, you can include information about your

company and a link to your website.

MTS Career CenterVisit www.mtsociety.org/careers

International Marine Forensics SymposiumApril 3 – 5, 2012 • National Harbor, MD

Keynote speaker James Cameron (con rmed).

His undersea documentaries include Expedition Bismarck, Ghosts of the Abyss, Volcanoes of the Deep Sea, Aliens of the Deep andLast Mysteries of the Titanic.

The Symposium is sponsored by the Marine Technology Society,

American Society of Naval Engineers,Society of Naval Architects & Marine Engineers.

This symposium will bring together marine professionals and historians to exchange information on historic marine

losses, on marine forensic investigation processes and tools, and on case studies where causes of failures and

losses have been determined or are under continued study. The Symposium will report on the latest research and understanding of the Titanic, Lusitania, Edmund Fitzgerald, the Monitor and Passaic, HMS Prince of Wales, Bismarck, HMS Hood, and the Andrea Doria.

The International Marine Forensics Symposium will be held at the Gaylord Hotel, National Harbor, MD. Exhibition space is available at $650 for a 10 x 10-foot space. Freeman Exhibit Services will handle the show decorating. A number of Sponsorship Opportunities are also available to maximize your impact with Symposium attendees. Contact Mary Beth Loutinsky at [email protected] for details.

Oceans of Opportunity:International Cooperation and Partnerships Across the Paci c

September 19–22, 2011Kona, Hawaii

Register Today for Outstanding Topics in Oceanology, the Latest Technology in the Exhibition Hall, and

Excellent Networking Opportunities!

go to: http://www.oceans11mtsieeekona.org/main.cfm/CID/16/Registration/

Of cial Notice of Marine Technology Society Annual Meeting2011 Of cer Elections and Notice of Annual Membership Meeting

The terms of the individuals holding the following Marine Technology Society (“MTS”) of ces end on December 31, 2011: Vice President of Publications, Vice President of Industry and Technology, Vice President of Education and Research, and Vice President of Government and Public Affairs. Elections for these of ces will be conducted at the MTS Annual Membership Meeting and Awards Luncheon held in conjunction with the OCEANS’11 MTS/IEEE Kona Conference, September 21, 2011, at the Hilton Waikoloa Village, Kona, Hawaii. The meeting will begin at noon PST.

Article IV, Section 1.6 of the MTS Bylaws de nes the nominations and election procedures for MTS of cer elections. On June 20, 2011, the MTS Nom-inating Committee presented MTS’s President-Elect with at least two (2) nominees for each of the of ces to be elected at this year’s Annual Meeting. After reviewing these nominations, the President-Elect directed MTS staff to prepare the ballot for this year’s election and Member Proxy, both of which contain the names of the nominees for each

of ce to be elected at this year’s election. The MTS Bylaws establish August 1, 2011 as the record date for determining the MTS members entitled to notice of the Annual Meeting and eligible to vote in this year’s election. If you are an active member of MTS as of August 1, 2011 (membership dues paid), you are a member in good standing and therefore entitled to vote in this year’s of cer elections.

Pursuant to MTS’s Bylaws, a member entitled to vote may do so in person at the Annual Meeting or by appointing a proxy to vote on the member’s behalf. If you wish to appoint a proxy to vote on your behalf at the MTS Annual Membership Meeting, please complete the online Member Proxy (http://www.votenet.com).

This Proxy form must be completed no later than 5:00 p.m. (Eastern Standard Time) on September 11, 2011 (“Proxy Deadline”). Proxies that are incomplete or received after the Proxy Deadline will not be accepted or counted. A hardcopy proxy form may be requested by calling the MTS of ce at 410.884.5330.

CORP OR AT E MEMBERSABCO Subsea

Houston, TexasAMETEK Sea Connect Products, Inc.

Westerly, Rhode IslandC & C Technologies, Inc.

Lafayette, LouisianaC-MAR Group

Houston, TexasCompass Publications, Inc.

Arlington, VirginiaConverteam

Houston, TexasDelcor USA

Houston, TexasDOF Subsea USA

Houston, TexasDynacon, Inc.

Bryan, TexasE.H. Wachs Company

Houston, TexasFluor Offshore Solutions

Sugar Land, TexasFugro Chance, Inc.

Lafayette, LouisianaFugro Geoservices, Inc.

Houston, TexasFugro-McClelland Marine Geosciences

Houston, TexasFugro Pelagos, Inc.

San Diego, CaliforniaGeospace Offshore Cables

Houston, TexasGS-Hydro US

Houston, TexasHARRIS CapRock Communications

Melbourne, FloridaHydroid, LLC

Pocasset, MassachusettsInnerspace Corporation

Covina, CaliforniaINTECSEA

Houston, TexasInterMoor, Inc.

Houston, TexasiRobot Corporation

Durham, North CarolinaJ P Kenny, Inc.

Houston, TexasKongsberg Maritime, Inc.

Houston, TexasL-3 Communications

Houston, TexasL-3 MariPro

Goleta, CaliforniaLockheed Martin Sippican

Marion, MassachusettsMarine Cybernetics AS

Trondheim, NorwayMitsui Engineering and Shipbuilding Co. Ltd.

Tokyo, JapanMohr Engineering & Testing

Houston, TexasOceaneering Advanced Technologies

Hanover, MarylandOceaneering International, Inc.

Houston, TexasOdyssey Marine Exploration

Tampa, FloridaOil States Industries, Inc.

Arlington, TexasPerry Slingsby Systems, Inc.

Houston, TexasPhoenix International Holdings, Inc.

Largo, MarylandQinetiQ North America – Technology Solutions Group

Slidel, LouisianaRaytheon Technical Service Company, LLCSaipem America, Inc.

Houston, TexasSchilling Robotics, LLC

Davis, CaliforniaSEA CON Brantner and Associates, Inc.

El Cajon, CaliforniaSonTek/YSI, Inc.

San Diego, California

South Bay Cable Corp.Idyllwild, California

Subconn, Inc.Burwell, Nebraska

Subsea 7 (US), LLCHouston, Texas

TechnipHouston, Texas

Teledyne Geophysical InstrumentsHouston, Texas

Teledyne Oil & GasDaytona Beach, Florida

Teledyne RD Instruments, Inc.Poway, California

Tyco Electronics Subsea Communications Morristown, New Jersey

UniversalPegasus International, Inc.Houston, Texas

BUSINE S S MEMBERSAanderaa Data Instruments, Inc.

Attleboro, MassachusettsAshtead Technology, Inc.

Houston, TexasBastion Technologies, Inc.

Houston, TexasBennex Subsea, Houston

Houston, TexasBioSonics, Inc.

Seattle, WashingtonBIRNS, Inc.

Oxnard, CaliforniaC.A. Richards and Associates, Inc.

Houston, TexasCochrane Technologies, Inc.

Lafayette, LouisianaCompass Personnel Services, Inc.

Katy, TexasContros Systems & Solutions GmbH Kiel, GermanyDeepSea Power and Light

San Diego, CaliforniaDeepwater Rental and Sypply

New Iberia, LouisianaDOER Marine

Alameda, CaliforniaDPS Offshore Inc.

Houston, TexasDTC International

Houston, TexasEnergy Sales, Inc.

Redmond, WashingtonEnviron-Tech Diving

Stanwood, WashingtonFalmat, Inc.

San Marcos, CaliforniaFugro Atlantic

Norfolk, VirginiaFugro GeoSurveys, Inc.

St. John’s, Newfoundland and Labrador, Canada

Global Industries Offshore, LLCHouston, Texas

Horizon Marine, Inc.Marion, Massachusetts

ICANMt. Pearl, Newfoundland and Labrador, Canada

Intrepid Global, Inc.Houston, Texas

IPOZ Systems, LLCKaty, Texas

IVS 3DPortsmouth, New Hampshire

iXBlue, Inc.Cambridge, Massachusetts

KDU Worldwide Technical ServicesSarjah, United Arab Emirates

KnightHawk EngineeringHouston, Texas

Liquid Robotics, Inc.Palo Alto, California

Makai Ocean Engineering, Inc.Kailua, Hawaii

Matthews-Daniel CompanyHouston, Texas

Oceanic Imaging Consultants, Inc.Honolulu, Hawaii

OceanWorks InternationalHouston, Texas

Poseidon Offshore MiningOslo, Norway

Quest Offshore ResourcesSugar Land, Texas

Remote Ocean Systems, Inc.San Diego, California

RRC Robotica SubmarinaMacaé, Brazil

SeaBotix San Diego, CaliforniaSeaLandAire Technologies, Inc.

Jackson, MississippiSeaView Systems, Inc.

Dexter, MichiganSonardyne, Inc.

Houston, TexasSIMCorp Marine Environmental, Inc.

St. Stephens, CanadaSound Ocean Systems, Inc.

Redmond, WashingtonStress Subsea, Inc.

Houston, TexasSubsea Riser Products, Inc.

Houston, TexasSURF Subsea, Inc.

Magnolia, TexasTeam Trident

Cypress, TexasTechnology Systems Corporation

Palm City, FloridaTeledyne Impulse

San Diego, CaliforniaTension Member Technology

Huntington Beach, CaliforniaVideoRay, LLC

Phoenixville, PennsylvaniaWET Labs, Inc.

Philomath, OregonXodus Group

Houston, Texas

INS T I T U T ION A L MEMBERSAssociacio Institut Ictineu Centre, Catala De Recerca Submarina

Barcelona, SpainCanadian Coast Guard

St. John’s, Newfoundland and Labrador, CanadaCity of St. John’s

Newfoundland and Labrador, CanadaCLS America, Inc.

Largo, MarylandConsortium for Ocean Leadership

Washington, DCDepartment of Innovation, Trade and Rural Development

St. John’s, Newfoundland and Labrador, CanadaFundação Homem do Mar

Rio de Janeiro, BrazilHarbor Branch Oceanographic Institute

Fort Pierce, FloridaInternational Seabed Authority Kingston, JamaicaMarine Applied Research & Exploration

Richmond, CaliforniaMarine Institute

Newfoundland and Labrador, Canada Monterey Bay Aquarium Research Institute

Moss Landing, CaliforniaNational Research Council Institute for Ocean Technology

St. John’s, Newfoundland and Labrador, Canada

Naval Facilities Engineering Service CenterPort Hueneme, California

NOAA/PMELSeattle, Washington

NoblisFalls Church, Virginia

OceanGate, LLCEverett, Washington

Oregon State University College of Oceanic and Atmospheric Sciences

Corvalis, OregonSociety of Ieodo Research

Jeju-City, South Korea

The Marine Technology Society gratefully acknowledges the critical support of the Corporate, Business, and Institutional members listed.Member organizations have aided the Society substantially in attaining its objectives since its inception in 1963.

Marine Technology Society Member Organizations

5565 Sterrett Place, Suite 108Columbia, Maryland 21044

Postage for periodicalsis paid at Columbia, MD,and additional mailing offices.

Documents

Biomimetics and Marine Technology