Increased Flexibility and Robustness of Mars Rovers
L. J. Bresina, Keith Golden, Elliot Smith, R. Washington
- 发表年份
- 1999
- 引用次数
- 54
摘要
Increased Flexibility and Robustness of Mars RoversJohn L. BresinaPhone:650.604.3365bresina@ptolemy.arc.nasa.govKeith Golden650.604.3585kgoldenDavid E. SmithPhone:650.604.4383de2smith@ptolemy.arc.nasa.govRichWashington650.604.1140richwptolemy.arc.nasa.govFax:650.604.3594NASA Ames Research CenterMail Stop: 269-2Mo ett Field, CA 94035-100 0 USAIn Proceedings of the5thInternational Syposium on AI, Robotics, and Automation in Space.ABSTRACTOur overall ob jective is to improe the pro ductivityofMars rovers by increasing the exibility and robustnessof their autonomous b ehavior. Toachieve this ob jec-tive, we set out to increase the on-b oard autonomyofrovers and enable commanding at a higher level with amore exible command language. In February, 1999,we demonstrated some of our rover autonomy tech-nologies as part of a Marsokho d rover eld test thatsimulated asp ects of the Mars '01{'05 missions. In thispap er, we present the commanding language employedin this eld test, called theContingent Rover Language(CRL), and describ e the ground to ols and on-b oardexecutive capabilities that were develop ed to generateand execute CRL plans. A key feature of CRL is thatit enables the enco ding of contingent plans sp ecifyingwhat to do if a failure o ccurs, as well as what to do ifa serendipitous science opp ortunity arises.1. INTRODUCTIONTraditionally, spacecraft commanding is accomplishedviarigid time-stamp ed sequences of primitive op era-tions. If anything go es wrong during execution, built-in routines attempt to safe the spacecraft and awaitfurther instructions from Earth. As NASA missionsb ecome more challenging, more sophisticated space-craft are required, as are more advanced means ofcommanding them.As a case in p oint, the MarsPath nder's Microrover Flight Exp eriment made sig-ni cant advances over previous rob otic missions. So-journer had to op erate in an uncertain environmentand resp ond more autonomously to sensor input.With resp ect to the So journer microrover, for the pur-p oses of this pap er, we fo cus on the issues of command-ing and contingency; for more details, see Mishkin,etal., 1998. Like traditional spacecraft, So journer wascommanded with time-stamp ed sequences,NASA contractor with Caelum Research Corp oration.and the commands tended to b e primitive op era-tions. However, there were op erations that were sp ec-i ed at a higher level; the primary example is the\Go to Wayp oint command, which implemented au-tonomous navigation to a sp eci ed co ordinate.A command sequence typically sp eci ed the activitiesfor one sol (Martian day) plus \runout commands incase the next sol's sequence was delayed. These se-quences contained no explicit contingencies; however,contingency resp onses to certain drastic scenarios werepre-loaded on b oth the Path nder lander and rover.The \Backup Mission Load was to b e used in theevent of a communication loss from Earth to the lan-der, and the \Contingency Mission Load was to b eused in the event of a communication loss from thelander to the rover.Our aim is to continue in the technology direction setby the Path nder mission and increase the robustnessof autonomous rovers by enabling a higher level ofcommanding with a more exible and contingent lan-guage. The intended b ene t is to increase rover pro-ductivity without a decrease in safety. Our strategyis to make incremental advancements in this directionso as to maintain relevance to currently planned Marsrover missions and to eventually enable missions b e-yond the current capabilities of ightrovers.With planetary rovers, there is uncertainty ab outmany asp ects of sequence execution: exactly how longop erations will take, howmuchpower will b e con-sumed, and howmuch data storage will b e needed.Furthermore, there is uncertainty ab out environmen-tal factors that inuence such things as rate of bat-tery charging or which scienti c tasks are p ossible. Inorder to allow for this uncertainty, sequences are typ-ically based on worst-case es
关键词
相关论文
Statistical Learning Theory
Yuhai Wu, Vladimir Vapnik
1999
Artificial intelligence: a modern approach
1995
Fractional Differential Equations
Igor Podlubný
2025
Applied Nonlinear Control
Jean-Jacques Slotine, Weiping Li
1991