Improved robust trajectory tracking of underactuated AUVs via unified Tube-NMPC with proactive finite-time disturbance compensation in uncertain ocean environments

This paper presents a unified tube-based nonlinear model predictive control (TNMPC) framework for 3D trajectory tracking of underactuated autonomous underwater vehicles (AUVs) subject to parametric uncertainty and deep-ocean current disturbances. The proposed architecture integrates a nominal nonlinear model predictive planning scheme, tube-based ancillary feedback and a finite-time extended state observer (FTESO) within a single predictive control structure. The FTESO jointly estimates the full system state and lumped disturbances. The state estimate initializes the nominal nonlinear model predictive control (NMPC) predictor, while the current disturbance estimate generates a bounded same-cycle feedforward input and the prediction horizon remains nominal. Under the stated sufficient conditions, tube tightening and composite Lyapunov analysis establish an RPI tube, post-entry recursive feasibility and input-to-state practical stability (ISpS). The proposed approach is evaluated on a high-fidelity 6-DOF AUV model through a structured scenario sweep and a continuous Monte Carlo campaign under up to 50% parametric uncertainty and persistent ocean-current disturbances, yielding lower aggregate cross-track, depth and heading errors than the DMPC and TMPC benchmarks on the reported 3D helical and sinusoidal trajectories.