490 日本ロボット学会誌 Vol. 27 No.5, pp.490～493, 2009

解説 Recent Developments at NASA/JPL

∗ ∗ RichardVolpe and Richard Doyle ∗Jet Propulsion Laboratory, California Institute of Technology

this allowed for a significant expansion of their original 1. Introduction science objectives, but has also provided the chance to While progress in AI, Robotics and Automation has thoroughly exercise the robotics systems, both in their been achieved in the United States along multiple original form, and as enhanced by subsequent software fronts, this report will focus on JPL contributions to upgrades. NASA space exploration objectives. Complementary As originally flown, the spacecraft functionality contributions from NASA and university partners are included: lander descent image motion estimation also noted. As an overview, this article will only briefly (DIMES), image feature tracking for visual odometry, describe these activities, and detailed descriptions can stereo vision for obstacle detection, local path planning be obtained both from the i-SAIRAS ’08 proceedings, for obstacle avoidance (GESTALT), vehicle kinematics as well as other publications. and control for driving, and manipulator kinematics and This overview will first present relevant technologies control for instrument placement. This onboard soft- from recent and upcoming Missions, including the ware suite was originally limited due to project devel- ongoing Mars Exploration Rovers (MER) mission, the opment schedule and flight system limitations, includ- recent Mars Lander mission, and the upcoming ing the 20 [MHz] flight processor. Additionally, ground 2011 (MSL) rover mission. control software with robotics heritage included the two Second, this overview will discuss major robotic primary operations interfaces: RSVPforengineering, software systems’ technology development, including and Maestro for science. the Couple Layer Architecture for Robotic Autonomy However, with the longevity of the flight system, and (CLARAty), the Dynamics and Real-Time Systems technology development funding from the Mars Tech- simulation (DARTS), the Maestro Operations Interface, nology Program, several additional capabilities were and initial efforts to combine these. matured, validated, and transferred to the flight system. Third, the overview will describe recent efforts in ex- These include: visual target tracking for autonomous ploration technology development as targeted to three instrument placement (VTT), model and vision based planetary mission scenarios: Mars, Lunar, and atmo- manipulator collision prevention, global path planning spheric ( and ). (Carnegie Mellon’s Field D∗,seeFig. 1), and image pro- Fourth, major recent technology re- cessing for autonomous dust-devil detection. All have search is described, and a summary is provided. been exercised on the flight systems on Mars. Usage of any algorithm, original or added, is at the discretion of 2. Mars Flight Robotics Progress mission planners according to the requirements of the Launched in 2003, the MER rovers, and Op- activities for any specific day. portunity, have continued to operate on the In 2008, the rovers were joined on Mars for four surface well beyond their target lifetimes.Notonly has months by the Phoenix lander, which touched down north of the arctic circle on Mars, and survived through 原稿受付 2009 年 5 月 1 日 the Martian summer. While Phoenix was a static lan- キーワード： Mars Exploration, Software Technology, Aerial der, it employed some of the same technology used by Robotics, Astrobiology Technology ∗4800 Oak Grove Drive, Pasadena, California 91109 MER. First and foremost, the manipulator on Phoenix

JRSJ Vol. 27 No. 5 —10— June, 2009 Recent Robotics Developments at NASA/JPL 491

Fig. 1 MER tracks made while driving with GESTALT and D∗ was crucial forsampleacquisition and delivery to the gram has funded several forms of software infrastruc- science instruments on the top of the spacecraft. JPL ture for onboard control, physics-based simulation, and robotics led the development, implementation and op- ground control operations. erations for manipulator control for constrained mo- CLARAty has been developed to provide a common tion while digging, grinding, and scraping, and pre- software infrastructure for control. It consists of cision placement during sample delivery. Addition- two layers. TheFunctional Layer is an object-oriented ally, robotics technologies were employed for operations framework for defining the relationship between all com- (adapted versions of MER’s operator interfaces, RSVP ponents of robot control, while providing support for and Maestro); physics-based simulation of entry, de- heterogeneous , and complementary or compet- scent and landing (EDL) for mission planning; and im- itive algorithms. The Decision Layer contains delib- age processing of orbital imagery for landing site selec- erative planning and scheduling as well as execution, tion. and has a defined interface for accessing the Func- Planned for 2011 is the next rover mission, the Mars tional Layer capabilities at different levels of abstrac- Science Laboratory (MSL). MSL is roughly twice as big tion. CLARAty development has been a cooperative as MER in all dimensions, and approximately five times effort with NASA JPL, NASA Ames, Carnegie Mel- more massive. This allows for a larger science payload, lon University (CMU), and the University of Minnesota. and carrying of a nuclear power source (RTG). Without Algorithm development from these institutions, as well solar panels to become dust-covered, and with the con- as others such as MIT, have been captured in the frame- stant heatfromtheRTG,MSL will not suffer the same work. thermal and power cycling problems that contributed to DARTS is a physics-based simulation system which the conservative lifetime predictions for MER—instead has been developed for emulation, controls, mission its primary mission is one full Martian year. Its robotic planning, and operations. While original versions of capabilities will include all those available for MER, the system were used for free-flying spacecraft, recent plus manipulator-based drilling to acquire rock samples. missions have required simulation of in-situ situations— As was enabled by the longevity of MER, it is antici- these included entry, descent, and landing; surface rov- pated that future software upgrades will add function- ing; or aerial operations. Models developed for EDL ality to the rover. have been developed and used in a software deployment named DSENDS. Additionally, DSENDS can be used 3. Robotics Software Technology for aerial robot (aerobot) simulations. Models for sur- To support technology development such as that suc- face wheel-terrain interactionhavebeen coupled with cessfully employed for MER, the Mars Technology Pro- terrain models and data in another software deployment

日本ロボット学会誌 27 巻 5 号 —11— 2009 年 6 月 492 Richard Volpe Richard Doyle named ROAMS. Both DSENDS and ROAMS are used for technology development for Mars, , and Titan, and missions such as Phoenix and MSL. Maestro is a Java software system for user interfacing during mission operations or technology development. It has been used for rover and aerobot field testing; MER and Phoenix missions, and is planned for use in MSL. Its component architecture, allows for contribu- tions from several organizations, including NASA Ames. Recent efforts have combined CLARAty, ROAMS, and Maestro into an end-to-end simulation of a full mis- Fig. 2 NASA JPL ATHLETE sion. Such a combined system has value as a “flight sim- ulator” for operators and scientists to train for future missions, and as a mechanism by which to test tech- nology components in a full system. For instance, new autonomy technology can be more easily integrated, and feedback on its utility can be obtained.

4. Planetary Robotics Applications

In addition to development ofsoftware systems, com- ponent technologies are under development for Mars, Moon, and extraterrestrial atmospheric exploration. For Mars, autonomy needs can be divided in four cat- egories: long traverse, instrument placement, onboard Fig. 3 NASA JPL Titan Aerobot Testbed science data processing, and sampling. For long tra- verse, mobility on steep terrain may require tethered ging, and drilling; and drive or walk over rough terrain systems, therefore JPL is investigating design and con- while actively controlling its attitude. Walking algo- trol of such systems (e.g. Cliffbot and Axel). For instru- rithms for ATHLETE are under development by NASA ment placement and onboard science, JPL is develop- Ames, and operations of the system are performed using ing autonomous visual rock detection and instrument modified versions of the RSVP and Maestro interfaces pointing (OASIS). For sampling, JPL is investigating from the Mars program. Several field tests of ATH- autonomous rock coring and caching from a MER class LETE have been performed, both alone and in conjuc- rover. Some of these may mature quickly enough for tion with other systems, such as JSC’s Chariot. eventual use on MSL. Others may only be suitable for Finally, for planned future missions to planetary at- future missions such as a possible sample caching mis- mospheres, JPL is developing several different aerobot sion in 2018. designs. Pressurized are under development The NASA Exploration Systems Mission Director for Venus applications, where the upper atmosphere is (ESMD) has a strong program to develop all the needed cool enough to permit conventional spacecraft design systems for a return to the Moon. Two major com- for the payload. Challenges for this mission scenario ponents of the program architecture are mobile assets are primarily materials technology for the caustic at- for the lunar surface: smallpressurized rovers being mosphere. Similar, unpressurized Montgolfier designs developed by NASA JSC (Chariot), and larger mobile arealso under development for Titan. Using solar or habitats based on JPL’s ATHLETE system (shown in RTG heating, and changing altitude to catch chang- Fig. 2). ATHLETE is a six-limbedsystem,whereeach ing wind directions, some level of navigation is possible. limb has 6 degrees-of-freedom (DoFs), plus a wheel. Additionally, JPL is studying fully controlled blimps for With a total of 42 DoFs; ATHLETE can carry and autonomous exploration of Titan (as shown in Fig. 3). dock payloads, carry and employ tools for grasping, dig- Earthtests of these systems have shown the ability to

JRSJ Vol. 27 No. 5 —12— June, 2009 Recent Robotics Developments at NASA/JPL 493 determine position with ground feature tracking, exe- Wettergreen received the i-SAIRAS ’08 award for Most cuting desired altitude and course corrections to loiter Outstanding Contribution to the Technical Program. near science targets, and drop sample capturing tubes 6. Summary with tethered retrieval. This article has provided a brief overview of NASA re- 5. Other Developments lated space robotics efforts in the US. Much of the tech- In addition to the MarsandESMD programs de- nology and mission implementation described is from scribed above, NASA also provides funding to scientist- NASA JPL, given its prominent role in Mars explo- led teams through the ASTEPastrobiology program. ration, and long history in space system development. Within this program and related funding, CMU has Also, key collaborations with other institutions such as researched topics such as performing spectroscopy on JSC, Ames, and CMU are also described. In addition roverplatforms, plowing as a mobility concept for steep to Mars exploration, relevant progress in lunar, aerial, crater walls, field testing to simulate search for life in and astrobiology technology has been discussed. the Atacama Desert, autonomous mapping at kilome- Acknowledgement The research described in this ter scales, and designing a rover concept for the lunar paper was carried out at the Jet Propulsion Labora- polar cold traps. The most compelling of these recent tory, California Institute of Technology, under a con- CMU efforts is the DEPTHX underwater autonomous tract with the National Aeronautics and Space Admin- vehicle, which successfully explored the world’s deepest istration. () in Zacaton, Mexico. For the excel- Copyright 2009 California Institute of Technology. lence and range of his research and results, Dr. David Government sponsorship acknowledged.

Richard Volpe Richard Doyle Dr. Richard Volpe is Manager of the Mobil- Mr. Richard Doyle is Manager of the Mis- ity and RoboticSystemsSection of the Jet sion Software, Computing, and Network- Propulsion Laboratory, NASA/Caltech. ing Program Office at the Jet Propulsion The section is a team of over 80 robotics en- Laboratory, California Institute of Tech- gineers doing research and spaceflight im- nology in Pasadena, California. He is an plementation of robotic systems for roving, advisory board member at IEEE Intelli- digging, ballooning, drilling, and other modes of in-situ plan- gent Systems, where he contributes the department “AI in etary exploration. Richard is also a member of JPL’s Science Space.” He served as a member of the Executive Council and Technology Management Committee (STMC), and has of the Association for the Advancement of Artificial Intel- been a member of the Phoenix Mars Lander ligence (AAAI) during the period 2000–2003. He is an As- team. From 2001 through 2004, Richard served as the man- sociate Fellow of the American Institute of Aeronautics and agerofMarsRegional Mobility and Subsurface Access in Astronautics (AIAA). Dr. Doyle is a recipient of the NASA JPL’s Space Exploration Technology Program Office. In ad- Exceptional Service Medal. He holds the Ph.D. in Computer dition to guiding technology development for future robotic Science from the MIT Artificial Intelligence Laboratory. He and the Moon, he has been actively in- gave the invited talk entitled “The Emergence of Spacecraft volved in 2003 & 2011 rover mission development. This has Autonomy” at AAAI-97 in Providence, RI. He has been the included managing internal JPLrovertechnology develop- US Program Chair for the International Symposium on Ar- ment, as well as external university research funded by the tificial Intelligence, Robotics and Automation in Space (i- Mars Technology Program. Richard holds a PhD in Applied SAIRAS), held at Tokyo in 1997, Noordvijk, The Nether- Physics from Carnegie Mellon University. lands in 1999, Montreal, Canada in 2001, Nara, Japan in 2003, and Munich, Germany in 2005. He was General Chair for the i-SAIRAS held at University City, Los Angeles in 2007. He is Local Arrangements Chair for the International Joint Conference on Artificial Intelligence (IJCAI-09) to be held in Pasadena, California in July 2009.

日本ロボット学会誌 27 巻 5 号 —13— 2009 年 6 月