星际争霸视角的未来作战自主决策技术  

On future combat autonomous decision technology for starcraft

作　　者：黄彬城 陈思 高放 葛建军[1,2] 吴雪玲 HUANG Bincheng;CHEN Si;GAO Fang;GE Jianjun;WU Xueling(Key Laboratory of Cognition and Intelligence Technology,China Electronics Technology Group Corporation,Beijing 100086,China;Information Science Academy,China Electronics Technology Group Corporation,Beijing 100086,China;CNGC North Automatic Control Technology Institute,Taiyuan 030006,China)

机构地区：[1]中国电子科技集团公司认知与智能技术重点实验室,北京100086 [2]中国电子科技集团公司信息科学研究院,北京100086 [3]北方自动控制技术研究所,太原030006

出　　处：《科技导报》2021年第5期117-125,共9页Science & Technology Review

摘　　要：星际争霸游戏对于研究未来作战自主决策技术有重要参考价值。阐述了星际争霸游戏与自主决策过程的相似性,分析了星际争霸战略、战术决策算法中面临的规划、学习以及不确定性等热点问题,从决策技术复杂度入手,讨论了未来作战自主决策技术面临的瓶颈问题,并提出以打造大型战争游戏为手段,重点从系统顶层架构、游戏AI建模技术、大型战争游戏引擎等关键技术出发,试图指出未来作战自主决策技术发展着力点,为自主决策系统的智能化技术开发和研究提供研究思路和理论基础。StarCraft is an important game for studying the future combat autonomous decision technology.Similarities between StarCraft and the autonomous decision process are described.Planning,learning,and uncertainty in decision-making algorithms for StarCraft are also analyzed.Firstly,the key problem of future combat autonomous decision-making technology is discussed in terms of decision complexity.Then,the article proposes to create a large-scale war game to clarify the development of future battle autonomous decision-making technologies,such as system's top-level architecture,game AI modeling technology,large game engines,etc.in order to provide a useful reference for the development of autonomous decision system intelligent technology.

关 键 词：智能化作战 即时策略性游戏 群体智能 自主策略 仿真博弈 制胜机理 人工智能 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]