首页 | 本学科首页   官方微博 | 高级检索  
     

实时格斗游戏的智能决策方法
引用本文:唐振韬,梁荣钦,朱圆恒,赵冬斌.实时格斗游戏的智能决策方法[J].控制理论与应用,2022,39(6):969-985.
作者姓名:唐振韬  梁荣钦  朱圆恒  赵冬斌
作者单位:中国科学院自动化研究所,中国科学院大学,中国科学院自动化研究所,中国科学院自动化研究所
基金项目:科技部科技创新2030“新一代人工智能”重大项目(2018AAA0101005), 中国科学院战略性先导研究项目(XDA27030400), 中国科学院青年创新促 进会项目(2021132)资助.
摘    要:格斗游戏作为实时双人零和对抗博弈的代表性问题,具有实时对抗和快速响应的重要研究特性.相应针对性方法的提出有效反映了游戏人工智能领域的重要研究进展及发展方向.本文以格斗游戏人工智能竞赛作为研究背景,将智能决策方法分为启发式规则型、统计前向规划型与深度强化学习型三大类型,介绍相应的智能决策方法在实时格斗游戏中的研究进展.为分析格斗游戏智能决策方法的表现性能,本文提出了胜率、剩余血量、执行速率、优势性和伤害性的5个性能因子,系统分析智能决策方法的性能优势及不足.最后,对未来的在格斗游戏中研究发展趋势进行展望.

关 键 词:实时格斗游戏  统计前向规划  深度强化学习  性能因子  智能决策
收稿时间:2021/10/19 0:00:00
修稿时间:2022/4/26 0:00:00

Intelligent decision making approaches for real time fighting game
TANG Zhen-tao,LIANG Rong-qin,ZHU Yuan-heng and ZHAO Dong-bin.Intelligent decision making approaches for real time fighting game[J].Control Theory & Applications,2022,39(6):969-985.
Authors:TANG Zhen-tao  LIANG Rong-qin  ZHU Yuan-heng and ZHAO Dong-bin
Affiliation:Institute of Automation, Chinese Academy of Sciences,University of Chinese Academy of Sciences,Institute of Automation, Chinese Academy of Sciences,Institute of Automation, Chinese Academy of Sciences
Abstract:Fighting game is a classical real-time two player zero sum game, which has the obvious characteristics of real-time confrontation and extremely rapid decision response. Research and study conducted on this platform reflects the important research progress and development direction in the field of game artificial intelligence. In this paper, we focus on the application and development of intelligent decision-making in real-time fighting games. Methods that are applied in fighting games are categorized into three approaches, including heuristic rules, deep reinforcement learning, and statistical forward planning. Their corresponding process and development in the field of real-time fighting games is deeply studied. In order to systematically show the superiority and inferiority of different methods, five key metrics are proposed to analyze their performance, including win rate, remaining hit points, speed of wining, advantages of wining, and damage to the enemy. After the systematic analysis, the potential directions of intelligent decision-making in real-time fighting games are concluded for future research.
Keywords:real-time fighting game  statistical forward planning  deep reinforcement learning  key metric  intelligent decision
点击此处可从《控制理论与应用》浏览原始摘要信息
点击此处可从《控制理论与应用》下载全文
设为首页 | 免责声明 | 关于勤云 | 加入收藏

Copyright©北京勤云科技发展有限公司  京ICP备09084417号