摘要
针对陆战对抗中智能体状态动作空间复杂和行为模式固定的问题,提出任务分层架构下的博弈策略生成方法.使用策略式博弈模型对陆战对抗问题进行分析建模,给出智能体任务执行中的收益矩阵构建方法,并通过求解混合策略,使智能体行为同时具有合理性和多样性.以陆军战术对抗兵棋推演为平台进行测试,实验证明智能体策略可解释性强,行为模式多样,在与AI和人类选手对抗时都具有较高的胜率.
In view of complex state-action space and fixed behavior pattern problems of agent in land warfare confrontation,a game strategy generation method under task hierarchical architecture is proposed.The strategic-form game model is used to analyze and model the land warfare confrontation problem,and the construction method of the payoff matrix in the task execution of the agent is given.The hybrid strategy is solved to make the behavior of the agent reasonable and diversified.The platform is tested by the war game of army tactical confrontation.The agent strategy has strong explicability,diverse behavior patterns and higher winning rate when confronting AI and human players.
作者
王玉宾
孙怡峰
吴疆
李智
张玉臣
WANG Yubin;SUN Yifeng;WU Jiang;LI Zhi;ZHANG Yuchen(PLA SSF Information Engineering University,Zhengzhou Henan 450001,China)
出处
《指挥与控制学报》
CSCD
2022年第4期441-450,共10页
Journal of Command and Control
基金
国家自然科学基金(61902427)资助。
关键词
陆战对抗
智能体
博弈策略
生成方法
land warfare confrontation
agent
game strategy
generation method