基于DRL与微分对策的无人机空战决策研究

Research on UAV Air Combat Decision MakingBased on DRL and Differential Games

随着无人机战场环境越来越复杂,空战对抗将逐渐成为主要的一种无人机作战方式。为了能够确保我方无人机在快速演变的战场态势下抓住先机、精确决策、快速致胜,需要根据实际作战环境、作战样式,建立无人机和环境进行交互的规则、无人机空战对抗中采用的战术使用规则,并结合规则,通过智能决策算法,达到提升无人机空战对抗胜率的目的。提出一种结合微分对策(Differential Games,DG)的深度强化学习方法(Deep Reinforcement Learning,DRL)解决此问题,利用深度强化学习的智能决策性以及微分对策的准确机动性,实现战术决策到机动决策。最后以空战对抗1V1为例,对提出的方法进行验证,结果证明方法可行有效。

As the UAV battlefield environment becomes more and more complex,air combat confrontation will gradually become a major UAV combat mode.In order to ensure that our UAVs seize the lead,make accurate decisions,and win quickly under the rapidly evolving battlefield situation,we need to set interactive rules between UAVs and the environment and combat operation rules used in UAV air combat confrontation based on the actual combat environment and combat style.The rules are combined with through intelligent decision-making algorithms,the purpose of improving the win rate of the UAV air combat confrontation can be realized.A deep reinforcement learning method combined with differential countermeasures is proposed to solve this problem.The intelligent decision-making of deep reinforcement learning and the accurate mobility of differential countermeasures are used to realize form tactical decision making to maneuver decision making.Finally,the air combat confrontation 1V1 is taken as an example,the proposed method is verified,and the results prove that the method is feasible and effective.