任务级兵棋智能决策技术框架设计与关键问题分析  

Technical Framework Design and Key Issues Analysis in Task-level Wargame Intelligent Decision Making

作　　者：张驭龙 范长俊 冯旸赫 张龙飞 刘忠[1] ZHANG Yulong;FAN Changjun;FENG Yanghe;ZHANG Longfei;LIU Zhong(School of Systems Engineering,National University of Defense Technology,Changsha 410073,China;Unit 31002 of PLA,Beijing 100000,China)

机构地区：[1]国防科技大学系统工程学院,长沙410073 [2]解放军31002部队,北京100000

出　　处：《指挥与控制学报》2024年第1期19-25,共7页Journal of Command and Control

基　　金：国家自然科学基金(71701205,62073333,62206303)资助。

摘　　要：在梳理总结当前兵棋智能决策方法研究现状的基础上,明确指出了开展任务级兵棋智能决策研究的8项难点,从博弈活动中智能体定位、信息流转与知识流转等问题出发,设计给出了任务级兵棋智能博弈框架、智能策略生成优化框架以及多智能体策略协同演进框架,进而分析了高维度下长时序的态势信息数据综合、异步策略分布式学习优化、基于语义模板的任务指令生成、多智能体训练营博弈调度等4项关键问题,并给出了基本解决方案,点明了开展任务级兵棋智能决策研究的技术要点与基本路线,为后续任务级兵棋智能体设计提供了有益参考。The research status of current intelligent decision-making method in wargame is sorted out and summarized,eight difficulties for carrying out the task-level intelligent decision-making method in wargame are specified.The intelligent gaming framework,the intelligent strategy generating and optimizing framework,and the multi-agent intelligent strategy co-evolutionary framework of task-level wargame are designed based on agent positioning,information flow and knowledge transfer and other issues.Such four key issues as situation information data integration,asynchronous strategy distributed learning optimization,task order generation based on semantic template,multi-agent training camp gaming schedule in high-dimension long time sequence are analyzed.The basic resolution scheme is provided,the technical points and basic lines for carrying out the intelligent decision-making research of task-level wargame are pointed out,which provides the beneficial reference for the agent design of the subsequent task-level wargame.

关 键 词：兵棋 博弈 决策 智能 强化学习 

分 类 号：E91[军事]