Abstract: To address the risk of collisions with unknown obstacles and the challenges of high-precision stable control faced by autonomous underwater vehicles (AUVs) during operations in deep-sea environments, where strong nonlinearities, hydrodynamic disturbances, and external uncertainties interact, we propose an intelligent control method based on a dual-layer collaborative architecture. The upper motion layer employs model predictive control (MPC) to achieve trajectory planning and global optimization under multiple constraints, while the lower execution control layer incorporates adaptive sliding mode control (ASMC) to enhance the system’s robustness against uncertainties and external disturbances, thereby realizing the coordinated optimization of trajectory tracking and dynamic execution. Simulation results demonstrate that in complex underwater environments, the proposed control strategy achieves trajectory tracking accuracy of over 90% while ensuring reliable obstacle avoidance with a minimum safety distance of 0.8 m. Additionally, the tracking errors for both velocity and angular velocity at the lower layer are maintained within 0.002 m/s and 0.002 rad/s, respectively. Comparative experiments show that under complex disturbances, the tracking errors of the proposed method are reduced by approximately 86.5% and 45.8% compared to traditional PID and sliding mode control, respectively, demonstrating strong robustness and dynamic control performance. The study lays a technical foundation for highly reliable autonomous navigation and operations of deep-sea AUVs.