Data-driven unmanned surface vessel path following control method based on reinforcement learning

Aiming at many difficult problems of path-following(PF) control of Unmanned surface vessel (USV), such as disturbance of wind, wave and current, nonlinearity of control effect, etc. This paper proposes a data-driven USV PF control method based on SARSA. Using the behavior-reward scoring mechanism of reinforcement learning to learn a set of behavioral rules. These behavioral rules are used for USV to make the best control behavior based on the state of the ternary array consisting of heading deviation, wind and current. The experiment uses the simulation environment to learn the algorithm. Through the analysis of the experimental results and the comparison with the traditional PID algorithm, the feasibility and advantages of the algorithm are verified.