Intrinsic evolution of safe control strategies for autonomous spacecraft

Autonomous space vehicles need adaptive control strategies that can accommodate unanticipated environmental conditions. Although it is not difficult to construct alternative control strategies, a proper evaluation frequently can only be done by actually trying them out in the real physical environment. It therefore becomes imperative that any candidate control strategy be deemed safe - i.e., it does not damage any systems - prior to being tested online. How to do this has been a challenging problem. We propose a solution to this problem. Our approach uses an evolutionary algorithm (EA) to intrinsically evolve new control strategies. All candidate strategies are checked for safety using formal methods. More specifically, an EA evolves a series of finite state machines (FSMs), each of which encodes a unique control strategy. Model checking guarantees whether all safety properties are satisfied in the strategy. A numerical example is included to illustrate our approach.