LOS Guidance‐Based Path Following Control for Unmanned Sailboat via Neural Reinforcement Learning

ABSTRACT This article proposes a line‐of‐sight (LOS) guidance‐based adaptive path following control method for unmanned sailboats using a neural reinforcement learning (RL) approach. To cope with different sailing conditions, three adaptive strategies are designed for downwind, upwind, and crosswind scenarios, ensuring reliable path adherence under a wide range of wind directions. A neural actor‐critic RL architecture is developed, in which the actor network generates control inputs while the critic network approximates a value function to evaluate long‐term performance, thereby enabling the RL mechanism to compensate for model uncertainties and environmental disturbances and to adaptively optimize the control policy online. In addition, an integral sliding mode control law is incorporated to further enhance robustness and improve transient response. Via rigorous Lyapunov analysis, the closed‐loop system was demonstrated to be stable. Simulation results validate the efficacy and adaptability of the proposed control strategy, demonstrating its superior tracking performance under different sailing conditions.