Path Following Control for Unmanned Surface Vehicles: A Reinforcement Learning-Based Method With Experimental...

Area of Science:

• Robotics and Control Systems
• Artificial Intelligence and Machine Learning

Background:

• Unmanned Surface Vehicles (USVs) require sophisticated control for autonomous navigation.
• Traditional control methods often struggle with the complex, nonlinear dynamics of USVs.

Purpose of the Study:

• To develop a reinforcement learning (RL)-based strategy for robust USV path following control.
• To enhance tracking accuracy and controller robustness by modifying the deep deterministic policy gradient (DDPG) algorithm.

Main Methods:

• Implemented an RL strategy integrating guidance and heading control for direct state-to-command mapping.
• Introduced a twin-critic design and integral compensator to the DDPG algorithm.
• Utilized a pretrained neural network-based USV model to handle unknown nonlinear dynamics.

Main Results:

• The proposed RL-based controller demonstrated superior performance compared to traditional cascade and standard DDPG controllers.
• Validated self-learning and path following capabilities through simulations and real sea experiments.
• Achieved significant improvements in tracking accuracy and robustness.

Conclusions:

• The developed RL strategy offers an effective solution for USV path following control.
• The integration of advanced RL techniques and a predictive USV model enhances autonomous navigation capabilities.
• The method shows promise for real-world applications requiring precise USV maneuverability.