基于MADDPG的无人艇集群任务调度与协同对抗方法
By combining the improved MADDPG algorithm with LSTM and MWNN, the real-time and accuracy issues of mission scheduling in complex sea areas of unmanned surface vessel (USV) swarms were resolved, achieving optimal scheduling and cooperative combat in multi-USV battles and improving the stability and scalability of the swarm system.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- SHANGHAI UNIV
- Filing Date
- 2023-02-13
- Publication Date
- 2026-07-17
AI Technical Summary
Existing unmanned surface vessel (USV) swarm task scheduling algorithms have poor real-time performance in complex sea areas, making it difficult to achieve accurate task allocation and threat assessment, resulting in resource waste and local optima problems. Traditional reinforcement learning methods are unstable and inefficient in complex environments.
An improved MADDPG algorithm is adopted, which combines a Long Short-Term Memory Network (LSTM) and a Membership Function-Wavelet Neural Network (MWNN). By predicting the state and threat level of enemy vessels, the reward value is corrected and the mission scheduling strategy is optimized. The algorithm is trained using a priority experience replay method.
It improves the prediction accuracy and training speed of unmanned surface vessel (USV) swarm combat, realizes optimal task scheduling and stable collaborative combat in multi-USV combat, and solves the problems of poor stability and scalability of swarm systems.
Smart Images
Figure CN116050795B_ABST