联合战役兵棋AI体系框架设计及关键技术分析  

Framework Design and Key Technology Analysis on Joint Campaign Wargame AI System

作　　者：任航 贺筱媛[1] 陶九阳[1] REN Hang;HE Xiaoyuan;TAO Jiuyang(National Defense University of PLA,Beijing 100091,China;Unit 31113 of PLA,Nanjing 210008,China)

机构地区：[1]国防大学,北京100091 [2]解放军31113部队,南京210008

出　　处：《火力与指挥控制》2023年第1期121-129,共9页Fire Control & Command Control

摘　　要：深度强化学习在游戏智能决策领域取得了令人瞩目的突破。多智能体深度强化学习、分层深度强化学习等领域的研究工作正将深度强化学习研究引向深入。由于联合战役兵棋博弈复杂的问题构成难以仅依靠深度强化学习方法解决,因而需要将人的知识经验有机融入强化学习过程中。对深度强化学习算法进行综合分析,并系统总结梳理联合战役兵棋AI的军事运用需求,在此基础上设计联合战役兵棋AI的体系框架并就框架中的相关技术进行探讨。Reinforcement learning(RL)achieves astonishing breakthrough in game intelligence decision-making field.The research work in such fields as Multi-Agent Reinforcement Learning,hierarchical reinforcement learning,etc.broadens the research direction of RL is applied.As joint campaign wargame game is structurally complicated itself,it is hard to find a solver just using RL,which makes the introduction of knowledge and experience from human necessary.Firstly the RL algorithm is analyzed and comprehensively analyzed a summary of the military requirements on AI for joint campaign wargame is carried out.On the basis of that,the system framework for joint campaign wargame is designed and the related technology is expounded.

关 键 词：深度强化学习 联合战役兵棋 军事需求 框架设计 知识经验 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]