A Planning and Control Architecture for Resilient Flights Under External Disturbances

Abstract

The autonomous safe flight of unmanned aerial vehicles faces significant challenges in unknown environments with unexpected disturbances. The mainstream is to implement antidisturbance control to mitigate the adverse effects on trajectory tracking. But only compensated in control, the planned trajectory may be risky and untrackable. Other studies primarily focus on the safety of the generated trajectory, ignoring the fact that the disturbances may breach the actuator's capabilities. This article proposes an efficient architecture to address these issues for resilient flights. Specifically, an improved disturbance observer-based safety controller is implemented on the autopilot. Based on the estimated external forces, a high-safety planner is proposed to resiliently generate the disturbance-aware trajectory. The planner is composed of a kinodynamic path searcher that considers actuator constraints and a spatio-temporal joint trajectory optimizer. The uncertainty of external forces on position is incorporated in collision avoidance by computing Hamilton–Jacobi forward reachability sets of error dynamics. Extensive simulations and real-world experiments in windy environments verify the presented system's feasibility and robustness, and show superior performance in terms of flight safety.