Dynamic Weapon Target Assignment Based on Deep Q Network

With the development of intelligent and unmanned warfare, the battlefield environment in the future is becoming increasingly complex, and the field of weapon target allocation has shown the characteristics of multiple styles, fast pace and strong uncertainty. Since the traditional exact algorithm and intelligent optimization algorithm cannot solve the dynamic weapon target allocation problem quickly and efficiently, this paper proposes a dynamic weapon target assignment (DWTA) based on deep Q network (DQN). Based on reinforcement learning, the neural network of deep learning is used for convergence and prediction, and the optimal strategy under the current conditions is obtained through continuous training between the target and the environment, so as to output the optimal interception scheme. In this paper, the simulation environment of weapon target assignment is constructed by Pycharm software and Tkinter model. The results of the assignment show that DQN algorithm is more than 50% faster and more accurate than the Q-learning algorithm. Finally, the proposed algorithm provides a new thinking for DWTA.