Waypoint Tracking Control for a Quadrotor based on PID and Reinforcement Learning

Waypoint tracking is a common task for quadrotors, which is usually achieved by the Inner-Outer loop PID control method. The outer loop is implemented by establishing PID control for the position and velocity errors of the quadrotor. This method requires the quadrotor's desired flight rate as an input. However, setting a fixed desired flight rate cannot match to changes in waypoints and the quadrotor's own state, which can lead to a large time consumption or even mission failure. To solve this problem, this paper uses a neural network trained by the DDPG algorithm to control the desired flight rate of the quadrotor during flight, so that the quadrotor can adjust the desired flight rate flexibly to achieve a better waypoint tracking performance. Simulation results show that the method proposed in this paper can improve the performance of the Inner-Outer loop PID control system through adjusting the desired flight rate.