Robust trajectory planning for multi-rotor aerial vehicles subject to saturation faults and wind disturbances

View Video Presentation: https://doi.org/10.2514/6.2023-4041.vid In this paper, we propose a trajectory planning approach to accommodate saturation faults associated with the propulsion system in multi-rotor vehicles under wind conditions. The Urban Air Mobility (UAM) paradigm demands strict safety guarantees in the face of unexpected mid-flight events. We propose a supervisory trajectory updating scheme based on reinforcement learning (RL) that operates on a flight computer by interacting with flight controllers available in most multi-rotor aerial vehicles. The RL-based planner is trained in simulation and can be refined based on real trajectory data to account for the sim-to-real transfer gap. The proposed algorithm is designed to trade off a conservative planning approach required when the vehicle experiences wind conditions and a more aggressive response required when one of the motors suffers a saturation fault. We demonstrate satisfactory results under different fault scenarios considering wind conditions in a simulation environment platform validated with real flight data.