结合机器学习的水下作业路径自适应规划方法及系统
By combining multi-source data fusion from satellite remote sensing, underwater in-situ perception, and edge computing, and constructing a deep collaboration between hybrid modeling and intelligent algorithms, the problem of adapting to time-varying environments in underwater operation path planning is solved, achieving efficient and safe path planning and improving the intelligence level of underwater operation equipment.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- CHINA WATERBORNE TRANSPORT RES INST
- Filing Date
- 2025-12-30
- Publication Date
- 2026-07-17
AI Technical Summary
Traditional path planning methods struggle to effectively integrate multi-source data and cannot respond promptly to dynamic scenarios such as sudden changes in ocean currents and complex terrain, resulting in insufficient efficiency and safety in underwater operations. They also lack a two-way iterative optimization system that combines virtual and real-world data, and the optimization of key parameters relies on offline debugging, making it impossible to achieve an optimal balance among multiple indicators.
Employing a satellite remote sensing, underwater in-situ sensing, and edge computing architecture, high-precision real-time sensing is achieved by acquiring data from multi-source sensing devices and combining it with an intelligent weight allocation algorithm. A hybrid modeling framework based on physical constraints and data-driven approaches is constructed, and an online adaptive update mechanism for the forgetting factor is designed. The optimal correction path is generated by combining reinforcement learning and transfer learning. A closed-loop algorithm for simulation pre-training and field fine-tuning is constructed, and a sensor fault tolerance mechanism is added to achieve linkage between planning, control, and feedback. A two-way iterative system between virtual and reality is constructed, and the parameters of the entire process are optimized through digital twin simulation and field experiments.
It achieves dynamic weight allocation and hierarchical fusion of multi-source data, ensuring the accuracy and real-time nature of flow field and terrain perception. The model can dynamically adapt to the time-varying characteristics of ocean currents, improve the efficiency and adaptability of path planning, and enhance the intelligence level and safety of underwater operation equipment.
Smart Images
Figure CN121722144B_ABST