Analytical contract generation for unmanned aerial vehicle safety : obstacle avoidance

Abstract

The application of contract-based design to UAV guidance, navigation, and control (GNC) systems is proposed as a way to guaranteed performance. Contract-based design has been used to formally verify performance on distributed systems in other applications. The purpose of this research is to develop contracts that encode the limits on operational conditions for safe flight. A reverse-reachability approach with a novel computational implementation is used to compute sufficient conditions for a crash with static obstacles, moving obstacles, and between two UAV's. Contracts that dictate limits on operational parameters are generated based on these results.

As unmanned aerial vehicles (UAVs) are increasingly depended on for high-criticality applications, there is a need for safety-guaranteed operation. We are particularly interested in the scenario of obstacle avoidance under sensor performance degradation. Methods such as gain scheduling and adaptive control have been used to mitigate the effects of changing system performance, but these efforts tend to lack a consistent framework, and do not integrate readily into higher-level analysis. Similarly, many fast path planning methods exist, but they do not come with guarantees of safety if the path needs to be updated in response to an obstacle.