Hybrid Physics-ML Modeling for Marine Vehicle Maneuvering Motions in the Presence of Environmental Disturbances

• DOI: 10.48550/arxiv.2411.13908
• Publication Date: 2024-11-21
• Authors: Zihao Wang; Cheng Jian; Liang Xu; Lizhu Hao; Yan Peng
• Keywords: FOS: Computer and information sciences; Computer Science - Robotics; Robotics (cs.RO)

A hybrid physics-machine learning modeling framework is proposed for the surface vehicles' maneuvering motions to address capability and stability in presence of environmental disturbances. From a deep perspective, based on variant version residual networks with additional feature extraction. Initially, an imperfect physical model derived identified capture fundamental hydrodynamic characteristics marine vehicles. This then integrated feedforward network through block. Additionally, extraction from trigonometric transformations employed machine component account periodic influence currents waves. The method evaluated using real navigational data 'JH7500' unmanned vehicle. results demonstrate robust generalizability accurate long-term prediction capabilities nonlinear dynamic specific conditions. approach has potential be extended applied develop comprehensive high-fidelity simulator.