Data-driven nonlinear model predictive control for AUV trajectory tracking under oceanic disturbances

In this paper, trajectory tracking of fully actuated autonomous underwater vehicles (AUVs) under modeling uncertainties, unknown ocean-wave disturbances, and actuator saturation is investigated. A unified data-driven nonlinear model predictive control (DD-NMPC) framework is developed, in which online model updating, disturbance estimation, and Lyapunov-constrained predictive control are coordinated within a single closed-loop architecture. An RLS/VRF-based online parameter estimation and model reconstruction scheme is introduced to continuously update the prediction model used by NMPC, thereby improving model fidelity under time-varying hydrodynamic effects while reducing reliance on pre-identified hydrodynamic coefficients. In parallel, a dual-layer nested adaptive sliding-mode disturbance observer (DLN-ASMDO) is designed for time-varying ocean-wave disturbances without requiring a priori disturbance-bound information, and the resulting disturbance estimates are incorporated into the NMPC loop as feedforward compensation. A Lyapunov-decrease condition is embedded together with the input constraints in the NMPC formulation, providing a basis for stability and recursive-feasibility analysis under actuator limitations. Lyapunov analysis and comparative simulations indicate improved trajectory-tracking accuracy, disturbance attenuation, and control smoothness.