Design of Quadrotor UAV Active Disturbance Rejection Controller Based on DDPG

In this paper, a linear Active Disturbance Rejection Controller (LADRC) based on deep reinforcement learning algorithm is proposed for attitude control of quadrotor UAV. The Deep Deterministic Policy Gradient (DDPG) algorithm was used to optimize the parameters of LADRC. The linear active disturbance rejection control has fewer adjustable parameters, less dependent on the accurate mathematical model of the controlled system, and has strong anti-interference and tracking capabilities, and then combines DDPG with LADRC to adapt the control parameters online to adapt to environmental changes. In order to effectively balance the needs of exploration and utilization, and at the same time improve the learning efficiency, a priority experience pool was introduced on the basis of the original experience pool. The simulation results show that compared with the traditional LADRC algorithm, the DDPG-LADRC algorithm has fast tracking speed and stronger anti-interference performance.