A Reinforcement Learning Framework to Adaptively Schedule Controllers for UAVs Operating Under Harsh Environmental Conditions

In this article, we present a hierarchical supervisory reinforcement learning (RL) framework to achieve precise trajectory tracking for UAV(s) operating in dynamic and complex environments. The UAV is equipped with multiple controllers, and each controller is tuned to provide a desired response under specific environmental conditions. Our objective is to dynamically schedule these controllers in response to abrupt environmental changes. To this end, we develop an RLbased framework for adaptive controller scheduling. We derive sufficient conditions for switching stability and validate our approach through extensive numerical simulations.