Unmanned surface vehicles(USVs)have significant strategic importance in maritime military confrontation and maritime border defense.To address the challenges in denial tasks,including cooperative operations,target allocation,and game-theoretic decision-making,a cooperative denial strategy for multiple USVs against suspicious target vessels in a combat environment is proposed.Firstly,the combat environment is modeled based on the context of multi-USV cooperative denial tasks.Secondly,a denial framework in many-to-many scenarios is established,where the denial targets are assigned based on the combat situation assessment and the Hungarian algorithm.Additionally,a multi-agent reinforcement learning method coupling target assignment with prioritized experience replay is proposed,integrating the Multi-Agent Deep Deterministic Policy Gradient(MADDPG),Twin Delayed Deep Deterministic Policy Gradient(TD3),and prioritized experience replay mechanism.Finally,a simulation environment is constructed,and the denial strategy is trained and tested using a centralized training and decentralized execution architecture.The experimental results demonstrate that the proposed denial strategy effectively completes the denial tasks against suspicious targets under various numbers of USVs.Specifically,it achieves a 94%task success rate and 584 step consumption in two-on-two scenarios,while outperforming other methods in terms of convergence and learning efficiency.This provides a theoretical reference for cooperative decision-making involving multiple USVs.
维普
万方数据
