Objective This paper investigates the high-precision control challenges associated with the autonomous recovery of an Autonomous Underwater Vehicle (AUV) by a dynamic docking base. The docking process is frequently compromised by complex underwater environments, characterized by time-varying external ocean currents and inherent model uncertainties. To address these issues, this study aims to propose a robust double-loop control strategy capable of achieving rapid, stable, and precise pose alignment between the AUV and the moving mother ship under constrained conditions.

Methods Using the White Dolphin 100 docking system as the primary research platform, a 5-DOF motion model is established to formulate the dynamic docking problem. The proposed control architecture consists of a kinematic outer loop for pose error elimination and a dynamic inner loop for velocity tracking, utilizing an Adaptive Fast Nonsingular Integral Terminal Sliding Mode Control (AFNITSMC) strategy. Specifically, a fast nonsingular integral terminal sliding mode surface is constructed to guarantee the finite-time convergence of system states while effectively eliminating the singularity problems inherent in traditional terminal sliding mode methods. To enhance robustness, an adaptive lumped disturbance estimation law is integrated to online estimate and compensate for uncertainties—such as model parameter mismatch and time-varying currents—without requiring prior knowledge of the disturbance upper bounds. Furthermore, a boundary layer technique is implemented within the switching term of the control law to suppress the chattering phenomenon, thereby protecting the mechanical actuators. The stability and finite-time convergence of the entire closed-loop system are rigorously proven using the Lyapunov stability theory.

Results Extensive simulation experiments were conducted based on the hydrodynamic parameters of the docking 100 system to verify the efficacy of the proposed method. The simulation scenarios incorporated 20% thrust saturation limits, time-varying ocean current disturbances, and 20% model parameter perturbations. The results indicate that the AFNITSMC method achieves rapid pose convergence within 10 seconds, with specific convergence times of 4.6 s, 7.0 s, and 9.39 s for the longitudinal, lateral, and vertical directions, respectively. This performance significantly outperforms the baseline Nonsingular Integral Terminal Sliding Mode Control (NITSMC), which required much longer intervals to stabilize. In terms of steady-state accuracy, the Mean Absolute Errors (MAE) for position were recorded at 0.142 cm, 0.103 cm, and 0.0397 cm, while attitude errors were 0.012° and 0.054°. Compared to the NITSMC method, the proposed strategy reduced position errors by 75.7%, 87.6%, and 95.3%, and attitude errors by 96.5% and 62.2%, demonstrating its superior tracking precision and robustness.