群体追逃微分博弈  

Group pursuit-evasion differential games

作　　者：高红伟 孟斌斌 刘剑[3] 戴照鹏 GAO Hongwei;MENG Binbin;LIU Jian;DAI Zhaopeng(School of Mathematics and Statistics,Qingdao Uni-versity,Qingdao 266071,Shandong,China;Intelligent Game andDeci-sion Laboratory,National Innovation Institute of Defense Technology,Academy of Military Science,Beijing 100071,China;PLANaval Submarine Academy,Qingdao 266199,Shandong,China)

机构地区：[1]青岛大学数学与统计学院,山东青岛266071 [2]军事科学院国防科技创新研究院智能博弈与决策实验室,北京100071 [3]海军潜艇学院,山东青岛266199

出　　处：《运筹学学报（中英文）》2024年第3期46-62,共17页Operations Research Transactions

基　　金：国家自然科学基金(No.72171126)。

摘　　要：本文以微分博弈和经典的追逃问题为主线,对群体追逃微分博弈的历史发展脉络进行梳理。针对大规模群体追逃问题,从平均场博弈视角出发,阐释了强化学习技术的应用前景。提出探索解决逆向追逃微分博弈的观点,可适用于水下无人舰艇、陆地机器人以及空中无人机集群等同类场景。区别于其他综述性文章,作者对于俄罗斯以及苏联在本领域发展历史中代表性的学术流派给予了较多关注。With differential games and classical pursuit-evasion problems as the main focus,this article aims to trace the historical development of group pursuit-evasion differential games.By addressing large-scale group pursuit-evasion issues from the point of mean-field games,the prospects of applying reinforcement learning techniques are elucidated.It proposes exploring solutions to inverse pursuit-evasion differential games,suitable for scenarios such as underwater autonomous vessels,terrestrial robots,and swarms of unmanned aerial vehicles.Diverging from other review papers,it devotes significant attention to the distinctive academic schools of thought in Russia and the former Soviet Union,highlighting their infuence in the evolution of this field.

关 键 词：追逃微分博弈 群体智能博弈 平均场博弈 逆向博弈 强化学习 

分 类 号：O225[理学—运筹学与控制论]