Formal Verification of Autopilots in Unmanned Aircraft Systems

The verification and validation (V&V) and certification problems for avionics systems in Unmanned Aircraft Systems (UAS) have been regarded as great challenges in realizing routine UAS into the National Air Space (NAS). Whereas current-day emerging avionics systems, such as embedded autopilots in UAS, are geared towards more autonomous operations, the currently employed V&V techniques in the industry largely rely on exhaustive testing, and this poses a technology gap between the software capability and the certification techniques in that airworthiness of autonomous systems is limited by the only certifiable operations, i.e., fixed, pre-planned operations considered in test scenarios. The research effort in this paper attempts to reduce the technology gap by developing a V&V technique that formally verifies a flight system instead of accumulating confidence by test cases. The approach relies on the promise demonstrated by the model checking technique in formally verifying the safety and correctness of finite-state machine systems and tries to extend its validity into the verification of UAS autopilot systems. The paper includes the description of applied model checking methods, underlying algorithms that abstract the hybrid system into a finite-state machine, and simulation results.