Reinforcement learning-based backstepping sliding mode heading control for unmanned sailboats

Abstract

This paper deals with the problem of heading control for unmanned sailboats using reinforcement learning-based backstepping sliding mode approaches. First, an uncertain dynamic model of unmanned sailboat subject to wind and external disturbances is established. Then, a reinforcement learning-based backstepping sliding mode heading controller (RL-BSMHC) is proposed. The controller consists of three components: backstepping sliding mode heading controller, fixed-threshold dynamic dual mode compensator, and reinforcement learning-based cooperative anti-disturbance component. Simulation results demonstrate that the designed RL-BSMHC is effective to improve the tracking performance of the unmanned sailboat. Moreover, it outperforms some existing heading controllers in tracking accuracy and robustness to unknown disturbances.