Decision-Making Method of UAV Maneuvering in Close-Range Confrontation based on Deep Reinforcement Learning

Abstract

In order to solve the problem of obtaining maneuver advantage in unmanned aerial vehicle (UAV) close-range confrontation, a maneuvering decision method based on deep reinforcement learning is presented. Taking one-to-one UAV confrontation as an example, there are two UAVs in the model, one of which uses traversal search algorithm in traditional decision-making method and the other uses deep reinforcement learning algorithm. A reinforcement learning model based on Deep Q Network (DQN) for one-to-one UAV confrontation is established, appropriate reward function is designed. In addition, this paper compares the effects of DQN algorithm, Double DQN algorithm and Dueling DQN algorithm on the above problems. The simulation results show that the deep reinforcement learning method is superior to traversal search algorithm in decision-making for maneuvering and has the ability of real-time decision-making. Double DQN algorithm and Dueling DQN algorithm are better than DQN algorithm in this problem.