一种基于碰撞风险驱动的多无人船协同避碰方法
By constructing a kinematic model of multiple unmanned vessels and screening key neighboring vessels through risk assessment, generating risk-enhanced temporal states, and inputting them into a multi-agent reinforcement learning network, the problem of accuracy and efficiency of cooperative collision avoidance of multiple unmanned vessel systems in partially observable environments is solved, and more efficient cooperative collision avoidance decision-making is achieved.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- OCEAN UNIV OF CHINA
- Filing Date
- 2026-04-29
- Publication Date
- 2026-07-17
AI Technical Summary
Existing multi-unmanned vessel cooperative collision avoidance methods struggle to accurately identify critical hazardous targets in some observable environments, resulting in high complexity of state representation and insufficient utilization of neighboring vessel motion trends and dynamic risk changes, thus affecting the accuracy and effectiveness of decision-making.
By establishing a kinematic model of multiple unmanned vessels, constructing a partially observable Markov game process, assessing collision risks and screening key risk neighboring vessels, generating risk-enhanced temporal states, inputting them into a multi-agent reinforcement learning network for training, and outputting a cooperative collision avoidance strategy.
It improves the collaborative collision avoidance decision-making capability of multiple unmanned surface vessels in partially observable environments, reduces low-risk information redundancy, lowers the state space dimension, and enhances the efficiency of identification and decision-making for key hazardous targets.
Smart Images
Figure CN122111030B_ABST
Abstract
Citation Information
Patent Citations
A multi-agent reinforcement learning unmanned ship formation collision avoidance method based on policy sequential update
CN122151957A