Rover Autonomy Study #1

We conducted this study to determine the relative benefits of developing various autonomy software technologies for a surface rover in the proposed Mars Science Laboratory (MSL) mission scheduled for 2009. Since the rover prototypes had been extensively field-tested in Mars-like terrain on Earth, we had access to an extensive body of real-world information.

We decomposed the mission into functional steps (acquire panorama, develop range map, plan path, etc.) covering long-range traverse, short-range approach to target, and sample acquisition and handling. For each of these steps in each mission element, we noted the kinds and frequencies of failure, and the time that was lost while the controllers developed a strategy to mitigate the failure.

For each of the science operations (moving samples to the rover's onboard analytic lab, conducting contact experiments, moving to a new site, etc.), we developed a utility function, based on interviewing an expert, which captured the relative importance of each activity. For example, the first sample collected in a bag may be worth 40% of the total mission value. In general, intrusive experiments, such as grinding up a rock sample and analyzing it with a mass spectrometer, merited the highest values.

Click here to see set of graphs.

We calculated the abilities of the autonomy software technologies under study to mitigate potential failures, as well as the difficulty in developing each of them. Subsequent work transformed the difficulty estimation into dollars. Since the cost of each technology cannot be predicted with certainty, we established uncertainty estimates in return-on-investment with regard to performance and, through modeling, to science return.

Each autonomy software technology was judged by two attributes: ability to save time (measured in Martian days, or "sols"), and cost.

The relative contributions of the autonomy technologies appear in the graphs at right.