A Decision Method for Simulated Confrontation of UAVs Based on Deep Reinforcement Learning

Aiming at the requirements of fast decision-making in entering allowable attack area and dominant area in simulated confrontation games of UAVs, this paper proposed a confrontation strategy solution method based on deep reinforcement learning and game theory which can make the UAV have a higher winning rate when facing highly intelligent opponents. Firstly, the motion model and game model of UAVs are established according to the actual combat scenario. Secondly, the dominant area by extending the allowable attack area is defined which can better guide the UAV to combat. Then a neural network which can process the continuous infinite state space of UAV is constructed and the Minimax algorithm is adopted to solve the optimal value of the action function which implies the optimal confrontation strategy. Finally, the simulation experiment is proposed to verify the effectiveness of the algorithm. The results show that the optimal confrontation strategy solving method proposed in this paper has better intelligence and adaptability than traditional methods.