Adaptive USV trajectory tracking: Preview-optimized rudder controller with self-tuning dynamics model

• A self-tuning dynamics model is proposed that autonomously calibrates with minimal data to enhance accuracy and sim-to-real transfer without manual intervention. • A preview-optimized rudder controller is developed without online numerical solvers by analytically embedding rate constraints, enabling faster and deterministic computation. • The controller is grounded in a tailored optimal control framework that theoretically guarantees tracking performance by incorporating future trajectory information with the guidance law. • Extensive simulations and sea trials validate the method’s effectiveness and robustness under stringent rudder constraints and complex paths. This paper addresses the critical challenge of trajectory tracking control for unmanned surface vehicles, focusing on two practical obstacles: performance degradation under rudder rate constraints and the difficulty in adapting dynamic models across varying operational conditions and vessel parameters. To overcome these issues, we propose an adaptive control framework integrating a preview-optimized rudder controller with a self-tuning dynamics model. The proposed approach employs a preview mechanism that anticipates future trajectory changes and proactively optimizes rudder commands within rate limits, substantially mitigating overshoot and maintaining high tracking accuracy. This controller is enhanced by an adaptive dynamics model that automatically calibrates itself using limited real-world data, ensuring reliable prediction performance and facilitating seamless transfer from simulation to real-world applications. Extensive simulations and sea trials under various challenging scenarios, including high-inertia loads and complex paths, consistently demonstrate the proposed method’s effectiveness in resolving the dual challenges of rate-limited actuation and model generalization, ultimately achieving superior tracking accuracy and robust performance.