基于显式对手建模的一对一超视距空战策略认知  

Opponent strategy cognition of one-on-one BVR air combat based on explicit opponent modeling

作　　者：胡振震 陈少飞 李鹏[1] 陈佳星 张煜 陈璟 HU Zhenzhen;CHEN Shaofei;LI Peng;CHEN Jiaxing;ZHANG Yu;CHEN Jing(College of Intelligence Science and Technology,National University of Defense and Technology,Changsha 410073,China)

机构地区：[1]国防科技大学智能科学学院,长沙410073

出　　处：《航空学报》2025年第4期162-187,共26页Acta Aeronautica et Astronautica Sinica

基　　金：国家自然科学基金(62376280)。

摘　　要：为解释和分析对手的空战策略,针对现有空战策略认知手段欠缺的问题,提出了一种面向一对一超视距(BVR)空战的显式对手建模(EOM)方法。将超视距空战问题视作不完美信息博弈,将时空连续的空战过程离散化,抽象出不同类型的空战行动,引入决策点概念来聚合同分布信息集,定义关键决策变量来考察影响行动的关键因素,利用非参数化机器学习方法构建易于理解的对手策略模型,即决策点上行动概率分布随关键决策变量变化的模型。利用模拟超视距空战开展复盘分析表明,利用该方法构建策略模型相比现有方法能更全面地解释对手的行动和分析对手的弱点,可为策略优化和装备发展提供建议。To explain and analyze the opponent's air combat strategy,an Explicit Opponent Modeling(EOM)method for one-on-one Beyond Visual Range(BVR)air combat is proposed to address the lack of strategy cognition in existing tools.The problem of BVR air combat is regarded as an imperfect information game,and the space-time continuous process of combat is discretized to abstract different types of air combat actions.The concept of decision point is introduced to aggregate the information sets conforming to the same distribution.Key decision variables are defined to examine the key factors affecting the actions.The non-parametric machine learning approach is used to construct an easy-to-understand opponent strategy model.The post-game analysis of the simulated BVR air combat shows that the constructed strategy model by the proposed method can explain the opponent’s actions and analyze the opponent’s weak points more comprehensively than existing methods,and can provide suggestions for strategy optimization and equipment development.

关 键 词：显式对手建模(EOM) 一对一超视距(BVR)空战 博弈 策略模型 决策点 

分 类 号：V219[航空宇航科学与技术—航空宇航推进理论与工程] N945.12[自然科学总论—系统科学]