Autonomous Robots 14, 103–126, 2003c 2003 Kluwer Academic Publishers. Manufactured in The Netherlands.Planetary Rover Developments Supporting Mars Exploration,Sample Return and Future Human-Robotic ColonizationPAUL S. SCHENKER, TERRY L. HUNTSBERGER, PAOLO PIRJANIAN, ERIC T. BAUMGARTNERAND EDDIE TUNSTELJet Propulsion Laboratory, 4800 Oak Grove Drive, Pasadena, CA 91109, [email protected]. We overview our recent research on planetary mobility. Products of this effort include the Field Inte-grated Design & Operations rover (FIDO), Sample Return Rover (SRR), reconfigurable rover units that function asan All Terrain Explorer (ATE), and a multi-Robot Work Crew of closely cooperating rovers (RWC). FIDO roveris an advanced technology prototype; its design and field testing support NASA’s development of long range, insitu Mars surface science missions. Complementing this, SRR implements autonomous visual recognition, nav-igation, rendezvous, and manipulation functions enabling small object pick-up, handling, and precision terminaldocking to a Mars ascent vehicle for future Mars Sample Return. ATE implements on-board reconfiguration ofrover geometry and control for adaptive response to adverse and changing terrain, e.g., traversal of steep, sandyslopes. RWC implements coordinated control of two rovers under closed loop kinematics and force constraints,e.g., transport of large payloads, as would occur in robotic colonies at future Mars outposts. RWC is based ina new extensible architecture for decentralized control of, and collective state estimation by multiple heteroge-neous robotic platforms—CAMPOUT; we overview the key architectural features. We have conducted experimentswith all these new rover system concepts over variable natural terrain. For each of the above developments, wesummarize our approach, some of our key experimental results to date, and our future directions of planneddevelopment.Keywords: mobile robots, cooperating robots, all terrain mobility, robotic colonies, robot architecture, reconfig-urable robots1. IntroductionThere is growing international interest in wide-rangingexploration of the Martian surface. A better understand-ing of Mars’ surface geology, morphology, geochem-istry, and atmospheric science will yield importantinsights about comparative planetary origins, poten-tial for past/present life, and capabilities of the Marssurface environment to sustain a permanent human-robotic colonized presence.Thus, institutions worldwide are pursuing devel-opment of Mars mission platforms/payloads, bothfixed and mobile, toward these science objectives.There are many options for such Mars surface ex-ploration: stationary landers with affixed instruments/samplers, gravity-impact penetrators, shallow anddeep drills, subsurface/tethered “moles”, light air-planes, touch-and-go atmospheric balloons, and semi-autonomous surface mobility. The word “semi”connotes earth-based planning, command-sequencingand analysis of rover activity sequences and dataproducts—as done by a science-engineering teamthrough periodic data down-link and command up-links. We have done past related work on dexterouslanded manipulators (Schenker et al., 1995, 1999a) re-sulting in a concept for NASA’s Mars Polar Landermission of 1998. More recently we have focused ondeveloping mobile science platforms—science rovers,such as the FIDO technology prototype shown inFig. 1.104 Schenker et al.Figure 1. FIDO Rover in desert field test on a cobbled lake bed,mast/science arm extended (inset, rover continuously traverses a sandwash, rear view, the mast/arm stowed (Weisbin et al., 1999)).1.1. Evolution of Mars Robotic Surface MobilityNear-term mission objectives include long-rangemobility and highly instrumented in situ sci-ence operations. Such remotely-commanded, over-the-horizon/OTH, semi-autonomous mobile scienceplatforms will enable remote field geology. As onespecific example, NASA’s upcoming Mars’03 (MarsExploration Rovers) mission seeks to dramatically ex-tend physical and observational scope of the 1997Mars Pathfinder/Sojourner exploration (Shirley andMatijevic, 1997; Shirley et al., 1997)—from 10’sofmeters about a nearby lander, with the rover carry-ing a single rear mounted instrument (AXPS/AlphaX-ray Proton Spectrometer)—to 1000’s of metersover open terrain with an on-board integrated sci-ence package (Mast instruments include a high res-olution multi-spectral panoramic camera, bore-sightedNIR point spectrometer, and integrated thermal emis-sion spectroscopy; rover arm instrumentation includesa color micro-imager, M¨ossbauer spectrometer, androck abrasion tool).Beyond this near-term vision there are major tech-nical challenges and diverse opportunities confrontingdevelopment of later Mars surface systems: Challengesinclude extending the spatial range and duration ofautonomous science operations (including on-boardscience analysis); enabling Mars sample return to earth,providing mobile access to increasingly high risk,scientifically rich areas; and broadening robotic op-erations to teams of cooperating agents, e.g., robotwork crews that support one another’s objectives (co-ordinated assembly, inspection, maintenance of bothscience and habitat) and extended robotic presence(health maintenance & self-repair).In the sections ahead we report on our approach tosome of these problems. We begin with a summaryof FIDO rover, Section 2, whose computing and elec-tronics architecture is shared by a number of our otherrobotics research vehicles, and whose technical con-cept and terrestrial field experimentation support theNASA MER’03 payload. We progress from FIDO workto Section 3, next describing a smaller class of light,agile, highly autonomous research rovers that we con-currently developed. These vehicles, the LightweightSurvivable Rover (LSR) and Sample Return Rover(SRR) have novel mechanical design, materials struc-ture and on-board sensory guidance. In particular,SRR provides a very rich technical infrastructure forvisually-guided navigation, manipulation and the in-tegration of these two functions in precision fieldrendezvous and payload transfers.As an outgrowth of these efforts on rover science/ au-tonomy, we recently began work on terrain-adaptivemobility. The objective, detailed in Section 4, is tohave a rover that adapts, in a physically optimal, be-haviorally intuitive way, to variable terrain—reactingautonomously, quickly and definitively