Standards-Based Wireless Sensor Networking Protocols for

Standards-Based Wireless Sensor NetworkingProtocols for Spaceflight Applications

Raymond S. Wagner, Ph.D.NASA Johnson Space Center

2101 NASA Parkway, Mail Code EV814/ESCGHouston, TX 77058

[email protected]

Abstract—Wireless sensor networks (WSNs) have thecapacity to revolutionize data gathering in bothspaceflight and terrestrial applications. WSNs provide ahuge advantage over traditional, wired instrumentationsince they do not require wiring trunks to connect sensorsto a central hub. This allows for easy sensor installationin hard to reach locations, easy expansion of the numberof sensors or sensing modalities, and reduction in bothsystem cost and weight. While this technology offersunprecedented flexibility and adaptability, implementingit in practice is not without its difficulties. Recentadvances in standards-based WSN protocols for industrialcontrol applications have come a long way to solvingmany of the challenges facing practical WSNdeployments. In this paper, we will overview two of themore promising candidates — WirelessHART from theHART Communication Foundation and ISA100.11a fromthe International Society of Automation — and present thearchitecture for a new standards-based sensor node fornetworking and applications research.

TABLE OF CONTENTS

1. INTRODUCTION .................................................................12. CHALLENGES OF NVIRELESS SENSING .............................23. OVERVIEW OF WSN STANDARDS ....................................24. EXPERIMENTAL METHODOLOGY FOR COMPARINGSTANDARDS ...........................................................................55. CONCLUSIONS .................................... ..............................7REFERENCES .......................................... ..............................7BIOGRAPHY ..........................................................................7

1. INTRODUCTION

Wireless Sensor Networks (WSNs) offer a new paradigmfor acquiring sensor data. Rather than gathering sensordata through wired data buses, WSNs employ a wirelessbackhaul to transmit sensor readings to central locationsfor aggregation and further processing. This provides anumber of potential benefits in spacecraft design, not theleast of which is the potential to substantially reducesystem weight by eliminating wiring harnesses andconnectors. Un-tethering sensors from wires also opensup a new range of possibilities. Sensing infrastructureneed no longer be fixed following spacecraft design andmanufacture: should situational awareness be enhanced

by re-locating a sensor from one panel to another, this iseasily accomplished. Similarly, additional sensors can beeasily added to the existing suite, providing a moredetailed measurement set without requiring more wires tobe strung behind bulkheads and through walls of pressureshells. Finally, wireless sensors can be re-used betweenvehicles once their initial missions have been ended. AWSN node can be relocated from a spent vehicle, such asa lunar larder, to one currently in service, such as a lunarrover or habitat. The node can even be outfitted with anew set of sensors in the process, retaining the commonradio and networking hardware; to give a new functionaltilt built mostly from recycled parts. Re-purposing wiredsystems would be much more difficult,requiring wiring to be stripped from one craft and re-strung in another, necessitating substantial disassembly ofspacecraft in both cases.

While this technology offers unprecedented flexibility andadaptability, implementing it in practice is not without itsdifficulties, particularly y with respect to achievingreliability that is on par with wired sensor approaches.Any practical WSN deployment must contend with anumber of difficulties in its radio frequency (RF)environment including multi-path reflections andinterference from other systems. Techniques must bedesigned to overcome all these factors, while at the sametime operating at a low enough power draw to allowoperational lifetimes on the order of years using small,onboard batteries.

In recent years, a great deal of focus has been given tosolving these cornrnon problems for WSN applications inindustrial automation and control, where the modernfactory, refinery, or offshore drilling platform presents anincredibly challenging RF environment. These efforts bygovernment, academic, and industrial partners haveresulted in standards-based wireless sensor network (S13-WSN)protocols capable of cornmunication reliabilityapproaching that of wired solutions with a very low per-node power consumption and network lifetimesapproaching the decade mark. Given similarities inoperational requirements between mission-criticalindustrial processes and spaceflight applications — namely,the insistence that data transport be both extremelyreliable as well as timely — we maintain that these SB-

WSN protocols hold great promise in the aerospace arenaas well.

In this paper we will overview the two major standards toemerge from the industrial control field — WirelessHARTfrom the HART Communication Foundation andISA100.11a from the International Society of Automation(ISA). Both are rooted in the IEEE 802.15.4 standard andprovide a level of robustness and reliability that shouldnkzke them well suited to spacefli ght applications. Wewill provide a technical review of both protocols startingwith their inspiration as a means to extend the capabilitiesof ZigBee — the first commmercial 802.15.4-derivedprotocol — which has had limited uptake in the mission-critical industrial control market. We will then discuss themethodology of an in-house performance evaluation ofthese protocols in a controlled environment and presentprelinvnary results.

2. CHALLENGES OF WIRELESS SENSING

Because sensor nodes are designed to operate withoutwire interconnects for data transfer, they will typically nothave wired power connections. This means nodes mustrely on local power sources such as power scavenging oronboard batteries for both data processing andcoon i nication. Sensor nodes must often be small andhave service lives on the order of years, so thisnecessitates very low power operation. With wirelesscoimrnnication representing the largest portion of anode's power utilization, the node must therefore restrictitself to very low-power transnussion with periodicsleeping/waking of radio circuitry. Such low-power radio-frequency (RF) communication is extremely vulnerable toa variety of distortion and interference mechanisms.Multi-path reflections can distort signals, limit data rates,and cause signal fades that prevent nodes from havingclear access to channels, especially in a closedenvironment such as a spacecraft. Other RF signalsources, such as wireless internet, voice, and data systemsmay contend with the sensor nodes for bandwidth.Finally, RF noise from electrical systems and periodicscattering from moving objects such as crew members cancollectively contribute to a highly unpredictable, time-varying communication environment.

Conn i nication reliability is key when replacing wiredsystems with wireless equivalents, so to cope with thisdifficult RF environment, a WSN must rely on severaldifferent mechanisms to deliver reliable performance. Forexample, it must utilize intelligent channel accessmechanisms in order to constantly monitor and exploitchannel quality and availability, which may vary rapidlywith time. In addition, it must employ an intelligentrouting mechanism that can move data reliably from asource node back to the central data manager, possiblyutilizing multiple different paths relayed through multiple

different intermediate nodes. Finally, it must providesimple, reliable mechanisms to expand and contract(scale) the network, allowing nodes to enter and leave asnecessary and changing the routing structure accordingly.Moreover, in addition to providing all these services, the

WSN must be designed so that nodes make minimal use oftheir radios, both for data transfer and networkcoordination, to pen-nit years-long operational lifetimes.

Designing such robustness at the outset is incrediblychallenging, and it requires expertise at all layers of thenetworking stack, from physical radio design to channelaccess schemes, routin g protocols, and distributed data-processing algorithms. Fortunately, a critical mass ofeffort across a variety of fields in wireless sensing hasemerged in the last few years, resulting in thedevelopment of WSN standards that can be leveraged byspaceflight applications. We will now provide a briefoverview of these standardized protocols.

3. OVERVIEW OF WSN STANDARDS

Robust, reliable wireless sensor/actuator networks stand tobenefit a number of industries, and as a result much efforthas been expended in recent years to develop designstandards for WSNs addressing many of theaforementioned problems. The core of these efforts is theIEEE 802.15.4 standard, which targets low-data rateapplications requiring wireless interconnections betweenmeasurement, analysis, and control devices — aggregationswhich can be classified as personal area networks (PANS).This puts 802.15.4 in the same general family as theBluetooth standard (IEEE 802.15.1), though the low-power and low-data rate nature of its intendedapplications differentiates it from the latter. An 802.15.4PAN can be analyzed using a simplified version of theOpen System Interconnection (OSI) protocol stackconsisting of the following layers (bottom to top):physical (PHY), medium access control (MAC),networking (NET), and application (APP). 802.15.4specifies only the PHY and MAC layers. The remaininglayers are provided by subsequent protocols such asZigBee, WirelessHART, ISA100.1 la.

The 802.15.4 PHY specification requires radios to operatein one of three frequency bands: 868-868.8 MHz (Europe,one channel), 902-928 MHz (North America, 30channels), and 2400-2483.5 MHz (worldwide, 16channels). A number of modulation schemes are allowedin the original 2003 standard and the 2006 and 2007updates, but the most common are flavors of directsequence spread spectrum. To regulate access to thechannel, the 802.15.4 MAC describes a carrier sensemultiple access with collision avoidance (CSMA-CA)scheme. That is, a device with a frame to send will firstlisten to the channel, and if there is no activity it can begintransmitting its data. If the sending device finds that the

channel is already in use, it waits for a random period andthen checks the channel for activity again, eithertransmitting or waiting for another random number perioddepending on what it observes. Typically some limit onthe number of attempts to make will be set so that devicesdo not wait ad infinitum to send a single frame [1-3].

Standardized WSN networking stacks build on top of thePHY and MAC provided by 802.15.4, adoptin g those twolayers either outright or with some modification. We willnow discuss the three most prominent alternatives, startingwith ZigBee.

ZigBee

ZigBee is a protocol that is more-or-less designed to ridedirectly above the 802.15.4 PHY and MAC layers,providing NET and APP layers to yield a completeprotocol stack. As is the case with all 802.15.4-basedsystems, ZigBee is designed for low-power, low-data rateapplications. The ZigBee protocol has gone throughseveral iterations to date, the most recent of which is theZigBee 2007 specification. ZigBee 2007 defines twoprotocol stacks. The first is simply called "ZigBee" and isvery similar to the previous single-stack ZigBee releases.It is designed for light-duty use in the home and the office(e.g., home lighting control). The second, called "ZigBeePRO", is a more robust protocol designed for industrialcontrol applications. It provides more reliableperforniance but requires implementation of a larger andmore complicated protocol stack [4].

ZigBee defines three classes of devices: ZigBeeCoordinators (ZC), ZigBee Routers (ZR), and ZigBee EndDevices (ZED). Each network has one ZC, which isresponsible for network formation and which can also aidin message routing. ZR's also participate in routing andcan run a sensing/ actuation application as well. ZED'sonly run applications and cannot participate in messagerouting — each ZED must report to either a ZR or the ZC[5,6].

ZigBee uses the 802.1.5.4 PHY and MAC layers directly,though the 2007 version does allow for some limitedfrequency agility in the PHY layer, so that radios can beautomatically switched away from problem channels whenthroughput falls off. Regarding the NET layer, bothZigBee Coordinators and ZigBee Routers participate inmulti-hop routing of messages; ZigBee End Devices onlyaddress messages to their associated parent routingdevice, which is found within their radio transmissionrange. With these device roles, a ZigBee network canhave one of three topologies: (1) star, where all non-coordinator devices report directly to the coordinator; (2)tree, where all ZOD's report to a routing device, androuting devices communicate up and down a tree ofrouting devices in a well-defined hierarchy crowned bythe ZC; and (3) mesh, where both ZOD's and routing

devices are free to communicate with any other routingdevice within radio range. While ZODs are allowed toperiodically cycle into a low-power sleep mode innetworks using 802.15.4's lintited duty cycling feature,ZR's must in general always remain awake. Some linutedprovision for ZR duty cycling does exist using the802.15.4 "beaconing" mechanism, which attempts toestablish a crude synchronization among nodes, thoughthe period of sleep/wake cycles is limited by the inabilityof low-cost ZigBee hardware to maintain precise tinting.Beaconing in general also requires longer wake timesfrom ZEDs [4-6].

While ZigBee has found a market in home and officesettings, the protocol has not been as widely embraced bythe industrial process measurement and control industry;even in its more robust ZigBee PRO form- It has beenfound that the solely contention-based MAC' is not able toreliably provide the message delivery required by criticalindustrial applications. Since 802.15.4-based sensornodes occupy the shared industrial, scientific, and medical(ISM) RF band, they can expect to observe transmissionsfor a variety of protocols with higher radiated power, suchas IEEE 802.11 and Bluetooth. In the presence of suchtraffic, the 802.15.4 MAC will always back off,potentially leaving the nodes unable to get the channelaccess required to send time-critical messages in a timelymanner. This is exacerbated in situations where a greatdeal of multi-path reflection is to be expected [10.11].

As a result, a pair of standards for high reliability wirelessnetworking of sensors in very difficult RF environmentshas emerged.

WirelessHART

The first, WirelessHART, is specified in the HART FieldCommunications Protocol, a long-established devicemeasurement and control protocol in industrialautomation. Revision 7 of the protocol augments theformerly wired-bound HART with a wireless data deliverymechanism based on the 802.15.4 physical layer.WirelessHART is designed from the ground up to enablewireless sensing and actuation in very harsh industrialenvironments, where comnnunication over the wirelessnetwork must have reliability comparable to wirelinecommunication. It also assumes that wireless assets willbe deployed where wired assets are difficult to place, sothat WirelessHART devices must be able to operate for along time on a single set of batteries.

WirelessHART uses the 802.1.5.4 PHY as-is. It operatesin the 2.4 GHz band and employs DSSS channel coding.A significant departure is taken from the 802.15.4 MACin the WirelessHART specification, however. TheWirelessHART MAC, based on the Time-SynchronizedMesh Protocol (TSMP) originally developed by Dust

Networks. Inc., employs time-division multiplexing of thechannel rather than the carrier-sensing and randombacking off of the 802.15.4 MAC. This alternative MACdesign was motivated by the extremely hostile radio-frequency (RF) environments that WSN's are likely toencounter in industrial environments. Industrialdeployments can be expected to be plagued by RFinterference from other wireless systems such as Wi-Finetworks and cordless phones, RF noise from machinery,physical obstruction of radio paths between devices,multipath effects between sources and receivers, and nodelosses due to depleted battery supplies andenvironmentally unfriendly operating conditions. Inaddition, these effects are likely to be highly time-variant,precluding an approach that attempts to compensate bycarefully calibrating a network to account for conditions atthe time of its deployment. A robust, agile system capableof working around changing ambient conditions must bedesigned [7].

The cornerstone of the WirelessHART MAC is timedivision multiple access (TDMA) to the channel, ratherthan the CSMA-CA approach taken by the 802.15.4MAC. This first requires time synchronization amongnodes, which is maintained by embedding time offsetinfonnation in acknowledgement (ACK) packets sent toconfirm successful reception of messages. Piggybackingthis sen-ice on ACK packets allows TSMP to avoidexpensive beaconin g approaches for synchronization suchas in the 802.15.4 MAC. Once a pair of nodes issynchronized, a schedule can be established forconnnunication. Time is divided into slots, and propersynchronization allows for agreement between the pair onslot start time [9].

Next, the nodes must decide which sub-band of the 16available in the 2.4 GHz PHY they will use. To do so,they agree on a start point in a sequence of channels.Beginning at this agreed-upon channel. at each new slotthe pair switches to the next channel in the sequence. Iftwo nodes find that they are unable to communicate on apoor-quality channel in a given slot, they need only towait one slot-length for a new chance to communicate.Blacklisting of channels with repeatedly poor performanceis supported so that these can be skipped in the slottedsequencing of channels [9].

Each receiving node in TSMP sends an ACK to the senderupon successful reception. Should the sender not receivean ACK, it may switch to an alternate neighboring nodeon its "parent" list and re-try the transmission. This parentlist records valid next-hop neighbors for a givendestination and is formed when the sending node initiallyjoins the network. Each parent's entry in the sendingnode's parent list is mirrored by the node's entry in thatparent's corresponding "child" list. Since alternateparents occupy different points in space, this adds anelement of spatial diversity to the overall MAC scheme.

Furthermore, since the re-try to the alternate parent will beat a later time-slot and on a different frequency, thescheme also features time diversity and frequencydiversity. Finally. the direct sequence spread spectrumchannel coding in the PHY layer adds code diversity, for asystem with a great deal of diversity and, hence, agility incoping with difficult wireless environments [9]. In fact,data transport in a well-formed WirelessHART network istypically greater than 3-sigma (99.7300204%) reliable,and under normal circumstances is greater than 6-sigma(99.9999998%) reliable [8].

All nodes in a WirelessHART network are full-functiondevices capable of routing multi-hop traffic. Thus, allWirelessHART networks form full mesh networktopologies. Note that, while each node is a router, eachcan also be efficiently battery powered due to theaggressive duty-cycling that the TSMP protocol allows (<1% time full-power) [7]. Even routers can periodically beput into a low-power "sleep" state with its radio powereddown, since time synchronization guarantees that whennodes "wake" and power up their radios, they will all bedoing it at the same time. Graph routing tables arecomputed by a central network manager and distributed tonodes on routing paths so that they know which neighborsare their next-hop parents on the path to the requesteddestination node. The number of parent choices availableat each node can be varied depending on the criticality ofthe route [9].

Finally, WirelessHART supports application messageswhich conform to the HART device connnunicationprotocol. In an adaptation of WirelessHART tospaceflight applications, however, this layer could beignored, instead treating the payloads of WirelessHARTpackets as generic buffers for application data.

ISA100.11a

WirelessHART has become the first standard to themarket in industrial automation, but following itsintroduction an effort began in ISA to develop a standardsuitable to a wider class of plant control networks thanjust those made of HART devices. The first standard tocome out of the ISA100 wireless group is ISA100.11a,which provides a wireless backend suitable to use with allmanner of legacy device cormnunication protocols at theapplication layer. ISA100.11a is in the final stages ofratification within ISA100 and should be releasedsometime in late 2009 or early 2010.

ISA100.11a shares many aspects in common withWirelessHART, including a TDMA MAC scheme basedon Dust Networks' TSMP algorithin, and in fact manyfeatures found in WirelessHART are designed intoISAl00.I la. The ISA protocol aims to provide a largerset of options, however, for industrial WSNs.Specifically, ISA 100.1Ia re-introduces a contention-based

MAC along the lines of original 802.15.4/ZigBeespecification in addition to the WirelessHART-likeTDMA-based MAC to enable higher throughput whendesired. While this CSMA-CA option will suffer the sameill effects as ZigBee in difficult RF environments, it willallow sensor nodes to have greater bandwidth in morefriendly environments. Thus, balancing of reliability andthroughput is possible, with the option of tradingreliability for throughput in more harsh environments orachieving higher throughput in environments that aremore friendly to low-power wireless networking.Additionally, ISA100.1 la supports a more simple class ofnon-routing device, whereas WirelessHART requires thatall devices be capable of routing network traffic. Whilethis limits the number of alternative paths available in thenetwork, it does allow designers to trade off device costfor routing redundancy. Several other distinctions ofvarying importance exist between the two protocols, suchas details regarding security, wired plant backbonenetworks, and inclusion of packet fragmentation and re-assembly (included in ISAl00.11a but notWirelessHART). [12]

The co-existence of these two advanced industrial WSNprotocols seems likely. A working group within ISA100has been investi gation a merger of the standards, and thiscontinues to remain an eventual possibility. In the neartenor, dual-boot devices are likely to provide an interimsolution to co-existence: nodes may be able to boot eitherthe ISA100.11 a stack or the WirelessHART stack asdesignated during their connnissioning, with code blocksconnnon to the two protocols shared by theimplementations of each.

4. EXPERIMENTAL METHODOLOGY FOR

COMPARING STANDARDS

We began our investigation of IEEE 802-15.4-basedprotocols by using off-the-shelf WSN hardware fromCrossbow, Inc. These parts used the 802.15.4 PHY andMAC layers directly and were configured in a startopology with each node reporting directly to the networkbase station. This configuration is analogous to aninstance of the ZigBee protocol stack, configured in a startopology, below the application layer. Moreover, nodeswere operated at their full duty cycle, so that they did notperiodically enter a low-power "sleep" state. Nodes weremounted in the Lunar Habitat Wireless Testbed (LHWT)shown in the foreground of Figure 1. The LHWT residesat Johnson Space Center in Houston, TX and serves as anenvironment for testing the co-existence of multiplewireless systems in a closed, reflective environmentsimilar to that expected in a habitable environment on thelunar surface or on orbit.

For, a test application, we chose wireless micrometeoroidorbital debris (MMOD) impact detection, which providesquite a challenging problem for a wireless data acquisition

system. Nodes monitor single-axis accelerometersmounted perpendicular to the habitat hull for highfrequency, transient signals corresponding to impactevents. Signals may then be analyzed for features ofinterest useful for solving problems such as impactlocalization. In this example, however, nodes merelyattempted to acquire sample points from the transientimpact signals at the maximum sample rate afforded bytheir onboard analog-to-digital (A/D) converters.Measurements were surninarized into packets that werethen sent to the base station for display and logging.

The results were rather disappointing, and stemmed fromtwo main sources: (1) lack of communication reliabilityafforded by the 802.15.4 MAC and (2) the processingarchitecture of the sensor nodes used in the experiment.Regarding the communication reliability, when thenetwork was configured to operate in a 2.4 GHz sub-channel that happened to be shared with another wirelesssystem (e.g., an 802.11 local area network), nodes wererarely able to send packets to the base station. In manycases, nodes dropped connectivity with the base stationentirely. Even when the network was re-configured tobroadcast in a less crowded 2.4 GHz sub-band, nodeswould occasionally suffer excessively longcommunication latencies with the base station. Thisbehavior confirms the criticism of the 802.15.4/ZigBeeMAC with regards to the lack of frequency agility in theavailable sub-bands.

Regarding the processing architecture, the high-bandwidthrequirements of MMOD data acquisition proved to bequite illuminating. We used a calibrated "pinger" togenerate impacts on the hull, and we verified through theuse of accelerometers and a wired data acquisition systemthat the impacts generated highly repeatable accelerometertraces across multiple experiments. However, whenanalyzing the results gathered from the wireless sensors,we found that the recorded si gnals were far from uniform.In several cases, nodes missed the transient event entirely,

reporting only background signals over the course of theexperiment. In others, samples indicated widely varyingpeak magnitudes, although maximum values should benearly consistent when using the calibrated pinger.

Although the accelerometers used with the Crossbownodes do not have a high enough bandwidth to capturetransient impact signals without some under-sampling,aliasing is not sufficient to explain the cases where nodesmissed ping events entirely. Further investigation into thearchitecture of the code naming on the nodes revealed thelikely cause to be the method of task scheduling on anode's embedded microprocessors. Simply put, to capturehi gh-frequency events such as MMOD impacts, nodesmust be able to sample their sensors constantly withoutinterference from other tasks, such as servicing thenetwork stack. When nodes use a single microprocessor

Figure 1: Lunar Habitat Wireless Testbed(foreground) at Johnson Space Center in Houston,TX.

to co-ordinate both data acquisition and networking, high-frequency transient events are likely to be missed either inpart or in total. This motivates a new sensor nodearchitecture more suited to both robust signal acquisitionand robust networking.

Form,ard Plan

In response to these shortcomings, we have designed anew wireless sensor platform which should serve as arobust research and development tool for evaluatingWSNs in a variety of contexts. The platform is highlymodular, allowing components to be changed out toenable investigations concerning competing WSNprotocol stacks, data processing algorithms, and sensormodalities. The design is shown in the diagram in Figure

The core of the new node, referred to in the figure as themain board, manages the data acquisition and processing.It contains a low-power microcontroller, off-chip

expandable memory, and the node's power supply. Themicrocontroller on the main board schedules sampling ofthe board's sensors (using its own AD module) andprocesses the sensed data prior to transmission.

The main board interfaces with a radio module. Typicallya commercial off-the-shelf (COTS) component, the radiomodule consists of a second microcontroller, radiocircuitry, and antenna assembly. The radio module fullyimplements the networking stack for the protocol underinvesti gation and interacts with the main board mainlythrough transmit and receive commands over a hardwareinterface (e.g., serial bus).

With an integrated main board and radio module, the coreof a imilti-purpose sensor node is complete. We can thenadd an application-specific sensor card ; containing all thesensors needed from the application under investigation aswell as auxiliary hardware such as di gital signal

processing chips or more advanced A/D components whenapplication requirements exceed the computation andsignal processing capabilities of the main board'smicrocontroller. In such a scenario, the microcontrollermay be used primarily to schedule the distributed sensingtasks, leaving the bulk of data acquisition and analysis tothese more powerful secondary modules.

The goal of this modular design is twofold. First, itallows us to build a WSN "development kit" that canquickly be custonuzed to meet new distributed sensingneeds for applications research. Second, it allows us tointerchange radio modules implementing different WSNprotocol stacks while rumung the same front-end sensingapplications to meaningfully compare the capabilities ofdifferent protocols and standard implementations.

We plan to interface this design with radio modulesimplementing both the WirelessHART and ISA100.11astacks, for which COTS components are now coming tomarket. Using this platform, we intend to pursue thefollowing lines of investigation:

RF issues: How reliable is the data delivery ofthese advanced industrial WSN protocols inpractice? How resistant are transmissions tomulti-path and other RF interference? What areachievable throughput rates? How well does thissystem co-exist with other 2.4 GHz devices suchas wireless LAN?

• Power issues: How power-hungry are theseprotocols in practice — do they achieveoperational lifetimes in excess of five years asadvertised? How does the sensing task affectbattery lifetimes? How do scheduled and event-driven sensing differ in their powerrequirements?

Application issues: How feasible is it toaccurately sense high-frequency transient events?Can timing information derived from the CDMAMAC be used to accurately synchronize time-stamping of measurements across nodes?

Protocol issues: How will future protocolsimprove upon WirelessHART and ISA100.11a(e. g_ next-generation ZigBee)? When themarket matures, which standard or standards willbe best suited to spaceflight applications? Willmodifications of standards be necessary?

radlo 11111111 S e11SOT

1110(fille board carol

Figure 2: architecture of sensor node underdevelopment; radio and sensing modules interfacewith a main controller board.

5. CONCLUSIONS

Standards-based WSN protocols developed for

demanding industrial environments hold much promise foraerospace applications. They have been designed to copewith extremely difficult RF environments to providehighly reliable data delivery and operational liftimes onthe order of years using only onboard batteries.Preliminary research verifies the shortconungs ofprotocols relying solely on the CSMA-CA MAC of802.15.4, such as ZigBee, and encourages furtherinvestigation into newer protocols with TDMA MACoptions, such as WirelessHART and ISA100.1 la. Using amodular platform dividing the processing load of sensingand networking between a pair of microcontrollers, weintend to fully investigate these protocols from bothnetwork and applications research angles.

REFERENCES

[1] IEEE standard 802.1.5.4-2003, http://standards.ieee.org/getieee802/download/802.15.4-2003.pdf.

[2] IEEE standard 802.15.4-2006, http://standards.ieee.org/getieee802/download/802.15.4-2006.pdf.

[3] IEEE standard 802.15.4a-2007, http://standards.ieee.org/getieee802/download/802.15.4a-2007.pdf.

[4] B.M. Blum; "ZigBee and ZigBee PRO: Which FeatureSet is Right for You?", Embedded.com, Oct. 2008,http://www.einbedded.com/design/210700316.

[5] J.K. Young, "Clearing Up the Mesh About WirelessNetworking Topologies", Embedded.com, Aug. 2008,http : //www. embedded. c om/design/networking/210200649 (partl), http://www.enibedded.com

/design/networking/210200676 (part 2).

[6] "ZigBee", Wikipedia, http://en.v^TMpedia.org/wiki/Zigbee.

[7] DUST Networks, "Technical Overview of TimeSynchronized Mesh Protocol (TSMP)",

[8] "WirelessHART Technical Data Sheet", HARTComrnunication Foundation, http://www.hartconun.org/protocoUtraining/resources/wiHART_resources/wirelesshart—datasheet.pdf.

[9] "IEC 62591: Industrial communication networks —WirelessHart Cornn'linucation Network andCorrnnunication Profile", International ElectrotechnicalCormnission (IEC) draft document 65C/532/CD, 2009.

[10] J. Song, S. Han, A.M. Mok, D. Chen, M. Lucas, M.Nixon, W. Pratt. "WirelessHART: Applying WirelessTechnology in Real-Time Industrial Process Control",Proc. IEEE Real-Time and Embedded Technology andApplications Symposium, 2008. pp. 377-386.

[11] T. Lennvall, S. Svensson, F. Hekland, "A Comparisonof WirelessHART and ZigBee for IndustrialApplications", Proc. IEEE Intl. Workshop on FactoryCommurucation Systems, 2008, pp. 85-88.

[12] "Wireless Systems for Industrial Automation: ProcessControl and Related Applications", InternationalSociety of Automation (ISA) Standard 100.11a, Draft2a, 2009.

BIOGRAPHY

Raymond Wagner leads the wireless sensor networkresearch and development program at NASA's JohnsonSpace Center; and he is involved in related programs fordevelopment of wireless commrmications systems forhabitat and surface operations. He has been withJohnson Space Center since the Fall of 2008, prior towhich he earned his Ph.D. in electrical engineering atRice University in Houston, Texas, with a thesisconcerning distributed data processing algorithms forwireless sensor networks. His research interests includewireless sensor networks, digital signal processing,computer networking, and wireless communications.

Avionic Systems DivisionNASA Johnson Space Center, Houston, Texas

Standards-Based Wireless Sensor NetworkingProtocols for Spaceflight Applications

Raymond S. Wagner, Ph.D.

2010 IEEE Aerospace ConferenceApplications and Architectures for Wireless Sensor Networks Session

Thursday, March 11, 2010


Overview

• Standards-based wireless sensor network (WSN) protocols show promisefor spaceflight applications

— mush R&D for reliable wireless sensor data transport can be leveraged

— standards-based WSN protocols already being used for mission-critical industrialprocess control in difficult RF environments

• Three main standards of interest derived from IEEE 802.15.4:— ZigBee (first to market but limited uptake in industrial control)

— WirelessHART (more robust, recently come on to market)

— ISA 100.11 a (next-generation, combines benefits of WirelessHART and ZigBee)

• NASA-JSC evaluation of protocols:— common hardware platform needed to meaningfully compare protocols

— R&D sensor node designed modularly to allow different standards-based radiomodules and application-specific sensor packages to interface through commonmicrocontroller motherboard

3/10/2010

2


Benefits of Wireless Sensor Networks(WSNs)

• Freeing sensors from wires offers many advantages:— removing wires/connectors reduces launch weight

— sensors can be added, relocated without expensive re-design and during missions

— sensor nodes can be re-cycled from spent vehicles (e.g., Altair lander) to in-service vehicles (e.g., lunar habitat, LER)

— sensors can be placed where running wires prohibitive

• Potential applications:— MMOD, leak location systems

— structural monitoring (e.g., stress/strain)

— radiation, gas, fire, airborne contaminant (e.g., lunar dust) detection

— temperature, light, etc. monitoring/control

— flexible prototyping of next-gen EVA suit sensor systems

• Potential problems:— nodes must be very low power for years-long service lifetimes

— reliable RF comm. difficult with low-power radios (channel access, multi-pathreflections, other RF sources )

3/10/2010

3


Benefits of Standards-basedWSNs

A standardized wireless sensor system offers significant benefits:

• Increased reliability through mesh network transport — many possiblepaths for data to reach control systems

• Scalability/Expandability — mesh network routing automatically discoversnew sensor nodes

• Reusability — many sensors/sensor applications can use the same networkto route data to command systems: applications co-operate rather thancompete

• Vendor selection — designing to open standard allows sourcing frommultiple vendors, prevents vendor lock-in

3/10/2010

4


General WSN Model

F-I° sensor 1

client •'° --application { radio sensor n

middleware module ; interface n

^•' gateway ••'•'•. .•• sensor N

clientapplication

3/10/2010


Standards Overview: IEEE 802.15.4

• 802.15.4 (2003) specifies the following:

Physical (PHY) layer — typically direct-sequence spread spectrum in 900MHz or 2.4 GHzMedium Access Control (MAC) layer — contention-based carrier sensemedium access with collision avoidance (CSMA-CA)

• Subsequent standards required to define network (NWK) and application(APP) layers

3/10/2010

6


Standards Overview: ZigBee

• Developed by ZigBee Alliance; first major 802.15.4-based, low-power, low-datarate standard

Uses both 802.15.4 PHY and MAC layers— Supports star, tree, and mesh topologies at network layer

Simple ZigBee End Devices run applications: can frequently sleep.— ZigBee Routers run apps, route traffic: can sleep rarely or never (depends on

MAC settings)— Industrial adoption has been slow, partly due to criticism of end-to-end

reliability of MAC (carrier sense multiple access with collision avoidance:CSMA-CA)ZigBee PRO stack in ZigBee-2007 release attempts to provide greaterreliability, though MAC still CSMA-CA (with some frequency agility)

3/10/2010

7

BIpNPI MelYla'. MMOp.pc9. , 2 X BIaNN mlm- MMw.p{ i-0 _

,S 1]

S ]Q

I^ + pp 149 indt 981a:17w Ap 181839: i •nd%8km.TWAp,1I1323;1I

u ^,•[anll o•. Mant,tl.p•i Q • Wo11r1 Morello•: MMOD.pay ...

+i 151

10

0 T

]

0

em •.om I.loo.nda Slnw:Tha Apr /e 18'3l11

oep Imovda Slnce Th. Apr I6 F8'11:18

Bl— M.A.. MMBp.Pa%1 -,JB Y a

15

a

o

mda Si ..: Th. Apr 16 18:23:19:

AP. 1818^39^10


Experimental Setup: ZigBee

ao

A

l ooa ,.mo 1.1m—mYe SInw:Thu A pt-18182:18.. _IebNM nennni: Ivnno.p<•c _A

e9

,p

6

a,w w_a ,mo +.mn e.rm

weds Sis[a:ThU Apr 181838:19e^on...lom ^ M^u..p.•w„ as=i

ro

I]

S

a

I olp = I om I: plnasanln:Ynu np11e 1e 1: a:1e

plwltal Mw171iR MMIP^ .,

e r r*nlae,.. ..+e^le,: a•rwanra^^1..1 — nI—,k epi*

m - x ro^

+a a

,0 10 e

e

OJ

I^ IUIB 1® I^ LInds 9NG:1TIl

3/10/2010

8

• 12 Crossbow MicaZ nodes with standard sensor boards— measuring accel. responses for MMOD

• Configured in "star" topology with full duty cycle (nosleeping of end devices) using 2.4 GHz 802.15.4 PHYand CSMA-CA 802.15.4 MAC


Experimental Findings: ZigBee

• high-frequency applications (e.g., MMOD) require separateapplication and networking processors

calibrated "ping" used, but sensors don't capture same accel. traces acrossexperimentscritical signal features un-sampled when one processor performing both A/Dconversion and networking duties (e.g., Crossbow)

• CSMA-CA MAC can lead to poor performance in crowded RFenvironments— packet delivery unreliable with 12-node network contending with RFID,

Bluetooth, 802.11 g/n devices, multipath reflections, and interference fromother unidentified sources in 2.4 GHz ISM band

3/10/2010

9


Standards Overview: WirelessHART

• Developed by HART Communication Foundation for harshindustrial environments

— Uses 802.15.4 PHY— Uses time-division multiple access (TDMA) as alternative to 802.15.4 MAC

MAC based on network-wide clock synchronization:— allows aggressive duty-cycling of all nodes— allows application timestamping/synchronization

MAC diversity through channel hopping (frequency) and multiple next-hoproute choices (spatial)

— Blacklisting of bad channels supported— NWK topology is full mesh with all nodes acting as routers

Focused on reliable transport: normally > 99.9999998% reliable deliveryRatified in 2007 1 compliant products shipped in 2008

3/10/2010

10


Experimental Setup: WirelessHART (Dust)

• new node in development:— Dust Networks/RF Monolithics

radio module

— TI MSP430 as applicationprocessor

• Development kit-based setuppictured

— final hardware (Fall 2009) hasfootprint of two AA batterieslaid side-by-side

3/10/2010 11


Experimental Findings: WirelessHART (Dust)

• Latency data — to be inserted (experiments in progress)

• Reliability data — to be inserted (experiments in progress)

3/10/2010

12


Standards Overview: ISA100.11 a

• Currently in development by International Society ofAutomation (ISA) -based results of U.S. Department ofEnergy study

Extends WirelessHART capabilities to provide single wireless backhaulfor multiple processing monitoring/control applicationsUses 802.15.4 PHYUses either TDMA or CSMA-CA MAC based on quality of servicerequested by applicationWorking with WirelessHART; targeting inter-operability of standards(dual-boot option)Likely ratified 2009; draft standard compliant parts in development

3/10/2010

13


NASA-JSC Work on WSN Standards

Investigating feasibility of IEEE 802.1 _5.4 wirelessinfrastructure for multiple applications (leak location,MMOD impact detection, env. monitoring, etc.):

Work to date highlights some problems with 802.15.4/ZigBee MAC(Crossbow)

• can be susceptible to RF interference

Currently developing flexible COTS testbed for evaluatingWlrelessHART/ISA 100.11 a :

• WSN protocol stack, transceiver in radio module• Separate microcontroller in sensor interface board handles

sensing/processing; sends/receives packets to/from modem as needed• Additional hardware (A/D for faster sampling, DSP for more

processing) added to sensor board as needed

3/10/2010

14


NASA-JSC Sensor Node Architecture

Sensor nodes composed of three basic components...

• main board:

— contains application processor (TI MSP430 micro controller), memory, power supply;responsible for sensor data acquisition, pre-processing, and task scheduling,, re-used in everyapplication with growing library of embedded C code

• radio module:

— COTS radio module implementing standardized WSN protocol (e.g, WirelessHART,ISA100.11a); treated as WSN "modem" by main board.

• sensor card:

— contains application-specific sensors, data conditioning hardware, and any advanced hardwarenot built into main board (DSPs, faster A/D, etc.); requires (re-) development for each application

Joe1

radio main sensormodule board card

3/10/2010 15


WSN Standards Forward Work

JSC lunar habitat mockup provides representative environment for WSNtesting. Issues to investigate include:

— U issues• Data delivery reliability — resistance to

multi-path, interference, noise• Data throughput rate• Interoperability — assess impacts on

2.4 GHz 802.11 WLAN— Power issues

• Radio/networking component— Low power, full mesh networking

• Sensing/processing component— Scheduled sensing— Event-driven sensing

— Application issues• Feasibility of sensing transient events• Usefulness of MAC-derived application

time synchronizatoin— Protocol issues:

• extending past WirelessHARVISA100.1 la to future protocols

3/10/2010

16


Conclusions

Emerging WSN standards show significant promise for spaceflightapplications

Work remains to compare standards and validate performance in relevantenvironments:

• interoperability with other wireless device must be demonstrated• extended lifetimes with battery operation must be shown

Modular hardware platform is necessary for WSN R&D:• use different networking stack modules with same application

processor/sensors for meaningful protocol comparisons• allow new sensor suites to be paired with common application

processor, networking stack for applications research

3/10/2010

17

Documents

Standards-Based Wireless Sensor Networking Protocols for