基于MADDPG算法的有人/无人机智能协同空战方法OA

Manned aerial vehicles(MAVs)and unmanned aerial vehicles(UAVs)cooperative combat is an important development direction for future air warfare.The design of manned/unmanned aerial vehicle coordination intelligent cooperative air warfare methods and the formation of a flexible and efficient combat system will play a huge role in enhancing the effectiveness of future air warfare.This paper utilizes deep reinforcement learning(DRL)algorithms to assist UAV decision-making by sharing real-time global enemy-friendly situational awareness obtained by manned aircraft,thus achieving manned/unmanned cooperative combat.Firstly,a complete air combat simulation environment is established,including UAVs' three-dimensional coordinates,motion states,action space,and reward function construction.Then,based on the characteristics of the combat aircraft,a manned/unmanned cooperative combat system is designed,and the Multi-Agent Deep Deterministic Policy Gradient(MADDPG)algorithm is used for training.Simulation experiments show that the proposed manned/unmanned cooperative combat system exhibits excellent performance in both 2v1 interception tasks and 2v1 dogfighting environments,proving the effectiveness of the algorithm in this field and providing new insights for manned/unmanned cooperative combat.