综合信息系统研究中心知识库

In order to improve the intelligent level of UCAV in one-to-one air combat, an autonomous maneuvering decision algorithm based on deep reinforcement learning is proposed. UCAV learns strategies by sensing the environment, performing maneuvering actions, and getting feedback. In this way, we can avoid the limitations of existing theories and human operations. Firstly an environment is modeled to simulate the real-time situation of air combat. Then a situation assessment method based on Energy-Maneuverability theory is utilized to design the reward functions. Finally model based on deep reinforcement learning is created for UCAV to learn strategies to gain the advantage for the opponent.