MAR Autonomous rovers are a critical element for the success of human exploration of Mars. The robotic tasks required for human presence on Mars are beyond the ability of current rovers; these tasks include emplacement and maintenance of a habitat, fuel production facility, and power generator, landing-site scouting, and mining. These tasks are required before and also during human presence; the ability of rovers to offload work from the human explorers will enable the humans to accomplish their mission. The capacity for these tasks will be realized by significant advancement toward full rover autonomy and, in particular, by overcoming current rover mission limitations in the areas of robust operation, resource utilization, and failure recovery.

The Pathfinder mission demonstrated the potential for robotic Mars exploration, but at the same time indicated clearly the need for more rover autonomy. The highly interactive, ground-intensive control with significant downtime limited the effectiveness of the Sojourner rover. Advances in rover offer increased rover productivity without risk to rover safety.

We are developing an integrated on-board executive architecture that incorporates robust operation, resource utilization, and failure recovery. This work draws from our experience with the architecture for the Deep Space One autonomy experiment, with enhancements in the area of ensuring robust operation in the face of unpredictable, complex environments, such as what a rover encounters on Mars.

Our ultimate goal is to provide a complete agent architecture for rover autonomy. The complete architecture will include long-range mission and path planning, self-diagnosis and fault recovery, and continual monitoring and adjustment of execution resources. The architecture will enable robust operation over long ranges of time and distance, performing complex tasks in a planned and opportunistic manner, and serving as an intelligent, capable tool for human explorers.