强化学习在机器博弈上的应用综述  

Review of Reinforcement Learning Applications in Machine Games

作　　者：杜康豪 宋睿卓 魏庆来[2] DU Kang-hao;SONG Rui-zhuo;WEI Qing-lai(School of Automation and Electrical Engineering,University of Science and Technology Beijing,Bejing 100083,China;Institute of Automation,Chinese Academy of Sciences,Beijing 100190,China)

机构地区：[1]北京科技大学自动化学院,北京100083 [2]中国科学院自动化研究所,北京100190

出　　处：《控制工程》2021年第10期1998-2004,共7页Control Engineering of China

基　　金：国家自然科学基金资助项目(61304079,61673054,61722312,61873300)。

摘　　要：人工智能是未来科技发展的必然趋势,将会对世界产生巨大的影响,而机器博弈更是人工智能研究的热点内容。目前,解决机器博弈问题最先进的算法都来源于强化学习。强化学习是机器学习最重要的方法之一,主要用来解决决策问题。它具有接近人类思维的学习机制,通过试错的方式同环境发生交互,累积最大奖赏并得到最优策略。博弈具有多种多样的形式,内容也十分广泛,根据不同的标准会产生不同的分类,可以将其分为完全信息博弈和非完全信息博弈,但它们都可以通过强化学习进行解决。Artificial intelligence(AI)is the inevitable trend of future scientific and technological development,which will have a tremendous impact on the world.At the same time,machine game is a hot topic in artificial intelligence research.At present,the most advanced algorithms for solving machine game problems are derived from reinforcement learning.Reinforcement learning,with a learning mechanism close to human thinking,is one of the most important methods of machine learning,which is mainly used to solve decision-making problems.It interacts with the environment through trial and error,accumulates the greatest reward and gets the best strategy.Game has a variety of forms and a wide range of contents.According to different standards,different classifications will be produced.Game can be divided into complete information game and incomplete information game,and both of them can be solved by reinforcement learning.

关 键 词：强化学习 机器博弈 非完全信息博弈 围棋 德州扑克 DOTA2 

分 类 号：TP181[自动化与计算机技术—控制理论与控制工程]