Research on PID Parameter Tuning and Optimization Based on SAC-Auto for USV Path Following

Abstract

Unmanned surface vessels (USVs) are required to follow a path during a task. This is essential for the USV, especially when following a curvilinear path or considering the interference of waves, and this work has been proven to be complicated. In this paper, a PID parameter tuning and optimizing method based on deep reinforcement learning were proposed to control the USV heading. Firstly, the Abkowite dynamics model with three degrees of freedom (DOF) is established. Secondly, the guidance law on the line-of-sight (LOS) method and the USV heading control system of the PID controller are designed. To satisfy the time-varying demand of PID parameters for guiding control, especially when the USV moves in waves, the soft actor–critic auto (SAC-auto) method is presented to adjust the PID parameters automatically. Thirdly, the state, action, and reward functions of the agent are designed for training and learning. Finally, numerical simulations are performed, and the results validated the feasibility and validity of the feasibility and effectiveness of the proposed method.

Keywords

• unmanned surface vehicle
• deep reinforcement learning
• parameter tuning
• path-following control