Self-planning methodology for planetary robotic explorers

Planetary robotic explorers must plan and re-plan their actions as new mission and environmental information becomes available. Here an on-line action plan generation procedure is proposed. Action plans are scripts that include navigation, sensing, and task instructions. The plans are constructed from physically realizable actions, called action modules, that are assembled to produce a successful action plan. The approach is based on a hierarchical selection process, which includes a genetic algorithm, to select a feasible action plan. The proposed methodology attempts to fully utilize the physical capability in rugged terrain, while minimizing the risk of mission failure. The method is demonstrated in the context of an example task and some guidelines are presented.