Constrained learning-based adaptive neural network trajectory tracking control of USVs with input saturation

This paper explores adaptive neural network control for unmanned surface vessels (USVs) that experience uncertain dynamics and input saturation. The study combines a backstepping control framework with the approximation capabilities of neural networks to develop nonlinear control schemes for both the speed and heading angle subsystems of USVs, ensuring that control signals remain within the specified input limits. To tackle the core difficulty of adaptive law design under input saturation, a convex optimization-based adaptive law framework is established, along with a novel constrained online learning strategy. The proposed strategy can effectively suppress neural network mislearning and avoid the performance deterioration of the closed-loop control system. Stability analysis proves that all tracking errors are bounded, and the effectiveness of the proposed control method is validated through numerical simulations and experimental results.