A Decomposed Symbolic Approach to Reactive Planning

Autonomous systems in uncertain dynamic environments must reconfigure themselves in response to unanticipated events and goals in real-time. We present an approach to reactive configuration planning based on the principle of decomposition. Reactive plans are susceptible to exponential state space explosion. We address this problem through transition-based decomposition by generating compact decomposed goal-directed plans. We further minimize state explosion by adopting a symbolic representation based on Ordered Binary Decision Diagrams. We demonstrate our reactive planner on representative spacecraft subsystem models.