Moving ahead in unstructured terrain - A versatile, low-cost andlight-weight snake robot prototype for motion optimisation

Kai Henning Koch1 Moritz Noltner2 Gero Plettenberg2 Manuel Rossler2 Katja Mombaur3

I. GENERAL MOTIVATION AND PROBLEM DEFINITION

Research on snakes and snake robots, are among the mostactive fields in the domain of biological inspired robotics.Despite the fact that snake locomotion is generally energyinefficient and features poor payload capacities comparedto other locomotion types (bipedal), researchers were eversince fascinated and inspired by the outstanding terrainability(creeping, swimming, climbing) of snake locomotion.In the last decades the world has seen a large variety ofrobotic prototypes based on different approaches to maxi-mize versatility and terrainability. The goal was to enablethose robots to operate safely in remote or hazardous en-vironment (e.g. incident sites) to support exploration andrecovery missions.However various aspects are still unexplored such as dif-ferent snake locomotion types and capacities, related bodyshapes and finally also their potential exploitation to improveterrainability. Therefore we propose a combined approach toparallely build a snake robot prototype as well as its virtualcounterpart from the same design resources. Simulation andoptimisation can be used to explore these aspects closelyrelated to real experiments.

II. DISCUSSION OF RELATED WORK

To the best of our knowledge, the first publication investi-gating the four different types of snake locomotion is from J.Gray [1]. The first operational robotic device was the ACMIII in 1972 presented to the public by S.Hirose [2]. Sincethen a lot of research has been conducted on characterisation[3], [4], mathematical abstraction of snake locomotion andcontrol [5], [6], [7] and its exploitation for building roboticdevices, applying either 2D (planar) using casters [8], [9],[10] or no wheels [11] as well as 3D locomotion [12],[13], [14] modes or even climbing [15]. A detailed andcomprehensive discussion of the recent state of art and openproblems is given in [16]. Despite alternative approaches tomultilink locomotion e.g. described in [17], [18] the authorsrarely found optimisation approaches to investigate differentaspects of snake-like locomotion in literature.

Financial Support of HGS and University of Heidelberg is gratefullyacknowledged

13rd Year PhD-Student - Optimisation in Robotics & andBiomechanics, IWR - University of Heidelberg, GermanyHenning.Koch@iwr.uni-heidelberg.de

2Bachelor-Student - RoboticsLab - Optimisation in Robotics & andBiomechanics, IWR - University of Heidelberg, Germany

3PhD Supervisor - Optimisation in Robotis & Biome-chanics, IWR - University of Heidelberg, GermanyKatja.Mombaur@iwr.uni-heidelberg.de

Fig. 1. Assembled robot snake specifically posed to have a view of thelower flat and upper rounded contact-surfaces. In the front of the headsegment one can observe the outlet for further sensor systems (e.g. camera).In contrast to the tail element, the head is not intended to contributesubstantially to locomotion.

III. OWN APPROACH AND CONTRIBUTION

We propose to build a virtual and a real robot snake proto-type from the same data resource: a parametric CAD design(see fig 2). Combined with rapid prototyping this approachsimplifies the transfer of the results from simulation andmotion optimisation into real world experiments. As a proofof concept a first prototype was designed and transferred intoa virtual and a real robot snake. Main principles governingdesign decisions of the first prototype were: simplicity, light-weight and versatility (reconfiguration). Therefore mainlystandard components were used to build the kinematic struc-ture, the control and the actuation system. Furthermore theresulting robot snake should be able to perform 3D trajecto-ries, but at least all 4 natural locomotion modes (serpentine,

Fig. 2. Above figure shows a rendered snapshot of our visualization toolshowing the robot’s simulation model including outer shell and the innerkinematic structure.

Fig. 3. Left and right-side figure show the CAD part of the shell componentA and B respectively. On the middle view one may observe the general shellprofile as an upper rounded and a lower flat surfaces. The shell parts areequipped with front and rear interface to allow for relative motion arroundthe joint axis. The front interface of each component is a revolution of theshell profile around the actual joint axis to maintain the smooth glidingsurface, with its subsequent shell’s rear interface, and prevent emergence ofobstructive features for a large range of joint angles during locomotion. Theouter shell profile is surrounded with blades [11] parallel to the direction ofmotion. At the front interface theses blades are rounded smoothly whilston the rear interface they are edged to prevent backdraft and improveanisotropic friction conditions.

concertina, rectilinear, sidewinding) found among snakeswith a large stability margin and maximum terrainability.

A. Hardware design

The internal kinematic structure was solely based on off-the-shelf standard components. As proposed in [13] thekinematic structure is built of a chain of Dynamixel AX-12packs (commonly used e.g. in DARwIn-OP[19]). To achieve3D motion capabilities, they are placed in alternating orderwith their joint axes rotated by 90deg.The outer shell is then placed on top. It should feature asmooth gliding surface [20], a large support polygon [13]and the possibility to feature anisotropic friction conditionswith its environment [7], [11]. Casters are not used toimprove terrainability. The design should also allow forchange of shape (quick replacement) whilst maintaining theinternal kinematic and control structure (see fig 3). Theshell components have been described by parametric CADand printed on a FDM (fused deposition modeling) rapidprototyping device, with an internal hollow structure formaximum weight efficiency. As this prototype is a proof ofconcept the structure is kept open for quick access to all

Fig. 4. Image sequence of a rendered motion, showing the simulationmodel performing the rectilinear gait. One may observe clearly a verticaloscillations traveling from the head to the tail segment.

internal components. A later version could be equipped witha water/dust-proof skin as suggested in [13].

B. Control structure & Power supply

The control structure is composed of a low-level controlunit (ATmega Micro-Controller) connected to the actuationmodules. It receives high-level motion commands asyn-chronously while processing joint angle trajectories for eachmotion module in fixed time intervals. The controller is basedon the motion controller of [6]. Currently, high-level motioncommands are sent on a RS232 serial line from a standardcomputer. This task is going to be shifted to a Rasperry Piboard, to be mounted into the head soon.The actuator modules communicate on a daisy-chain half-duplex UART bus with a standard asynchronous serial com-munication protocol with the low-level control unit [21].

C. Simulation

Figure 2 depicts the virtually built snake prototype. Ina first verification the proposed motion controller [6] wassimulated kinematically as shown in figure 4. Next we willproceed to compute dynamic characteristics from the CADdesign to compile the complete kinematic and dynamicrepresentation of the snake prototype.

IV. CONCLUSION & OUTLOOK

In this project it was possible to build a light-weight, low-cost and versatile prototype in simulation and manufacturedbased on a common CAD design and rapid prototyping. Theresulting robot prooved good maneuverability on flat terrainduring different locomotion modes - thus we conclude thatthe first proof of concept has been successfully completed.In the future efforts will be focused on computing dynamiccharacteristics and contact modeling to investigate differentaspects of snake-like locomotion based on optimisation ap-proaches.

REFERENCES

[1] J. Gray, “The mechanism of locomotion in snakes,” Journal ofexperimental biology, vol. 23, pp. 101–120, 1946.

[2] S. Hirose, Biologically inspired robots: Snake-Like Locomotors andManipulators. Oxford University Press, 1993.

[3] S.Ma, “Analysis of snake movement forms for realization of snake-like robots,” in International Conference on Robotics & Automation,1999.

[4] B.C.Jayne, “Kinematics of terrestial snake locomotion,” Copeia, vol. 4,pp. 915–927, 1986.

[5] J.Ostrowski and J.Burdick, “Gait kinematics for a serpentine robot,”in International Conference on Robotics and Automation, 1996.

[6] K.Lipkin, I.Brown, A.Peck, H.Choset, J.Rembisz, P.Gianfortoni, andA.Naaktgeboren, “Differentiable and piecewise differentiable gaits forsnake robots,” in International Conference on Intelligent Robots andSystems, 2007.

[7] P.Liljeback, K.Y.Pettersen, O.Stavdahl, and J.T.Gravdahl, “Controlla-bility and stability analysis of planar snake robot locomotion,” IEEETransactions on Automatic Control, vol. 56, pp. 1356–1378, 2011.

[8] S.Ma, H.Araya, and L.Li, “Development of a creeping snake-robot,”in Computational Intelligence in Robotics and Automation, 2001.

[9] S.Ma, N.Tadokoro, B.Li, and K.Inoue, “Analysis of creeping loco-motion of a snake robot on a slope,” in International Conferenc eonRobotics & Automation, 2003.

[10] Y.Shan and Y.Koren, “Design and motion planning of a mechanicalsnake,” Transactions on systems, man and cybernetics, vol. 23, pp.1091–1100, 1993.

[11] M.Saito, M.Fukaya, and T.Iwasaki, “Serpentine locomotion withrobotic snakes,” in IEEE Control Systems Magazine. IEEE, 2002.

[12] H.Ohno and S.Hirose, “Design of slim slime robot and its gait oflocomotion,” in International Conference on Intelligent Robots andSystems, 2001.

[13] C. Wright, A. Jonson, A. Peck, Z. McCord, A. Naaktgeboren, P. Gi-antfortoni, M. Gonzalez-Rivero, R. Hatton, and H. Choset, “Designof a modular snake robot,” in IEEE/RSJ International Conference onIntelligent Robots and System, 2007.

[14] R.Worst and R.Linnemann, “Construction and operation of a snake-like robot,” GMD - German National Research Center for Information,Tech. Rep., 1996.

[15] M.Yim, S.Homans, and K.Roufas, “Climbing with snake-like robots,”Xerox Palo Alto Research Center, Tech. Rep.

[16] A. Transeth, K. Y. Pettersen, and P. Liljeback, “A survey on snakerobot modeling and locomotion,” Robotica, vol. 27, pp. 999–1015,2009.

[17] F.L.Chernousko, “Snake-like locomotions of multilink mechanisms,”Journal of Vibriation and Control, vol. 9, pp. 235–256, 2003.

[18] F. Chernousko, “Modelling of snake-like locomotion,” Applied Math-ematics and Computation, vol. 164, pp. 415–435, 2005.

[19] I.Ha, Y.Tamura, H.Asama, J.Han, and D.W.Hong, “Development ofopen humanoid platform darwin-op,” in SICE Annual Conference,2011.

[20] P.Liljeback, O.Stavdahl, K.Y.Pettersen, and J.T.Gravdahl, “Two newdesign concepts for snake robot locomotion in unstructured environ-ments,” PALADYN Journal of Behavioral Robotics, 2011.

[21] User’s Manual - Dynamixel AX-12, Robotis Std., 06 2006.