Deep-Reinforcement-Learning-Based Active Disturbance Rejection Control for Lateral Path Following of Parafoil System

Abstract

The path-following control of the parafoil system is essential for executing missions, such as accurate homing and delivery. In this paper, the lateral path-following control of the parafoil system is studied. First, considering the relative motion between the parafoil canopy and the payload, an eight-degree-of-freedom (DOF) model of the parafoil system is constructed. Then, a guidance law containing the position deviation and heading angle deviation is proposed. Moreover, a linear active disturbance rejection controller (LADRC) is designed based on the guidance law to allow the parafoil system to track the desired path under internal unmodeled dynamics or external environmental disturbances. For the adaptive tuning of the controller parameters, a deep Q-network (DQN) is applied to the LADRC-based path-following control system, and the controller parameters can be adjusted in real time according to the system’s states. Finally, the effectiveness of the proposed method is applied to a parafoil system following circular and straight paths in an environment with wind disturbances. The simulation results show that the proposed method is an effective means to realize the lateral path-following control of the parafoil system, and it can also promote the development of intelligent controllers.