Autonomous Underwater Vehicle Trajectory Tracking Based on Physics Informed Neural Network Modeling

The autonomous underwater vehicle (AUV) serves as a foundational system for ocean exploration and understanding. Nonetheless, the intricate underwater environment and unpredictable dynamic disturbances can lead to decreased tracking accuracy. To address this challenge, this paper introduces a model predictive control (MPC) approach enhanced by Physics-Informed Neural Networks (PINNs). This integration leverages PINNs to refine and predict the model within the traditional MPC framework, thereby enhancing the algorithm's robustness and dynamic responsiveness in the presence of external noise. Simulation results demonstrate the method's exceptional robustness and tracking accuracy in AUV trajectory tracking.