CASIA OpenIR  > 智能系统与工程
Learning Deep Decentralized Policy Network by Collective Rewards for Real-Time Combat Game
Peixi Peng1; Junliang Xing1; Lili Cao1; Lisen Mu2; Chang Huang2
2019
会议名称International Joint Conference on Artificial Intelligence
会议日期August 10-16, 2019
会议地点Macao, China
摘要

The task of real-time combat game is to coordinate multiple units to defeat their enemies controlled by the given opponent in a real-time combat scenario. It is difficult to design a high-level Artificial Intelligence (AI) program for such a task due to its extremely large state-action space and real-time requirements. This paper formulates this task as a collective decentralized partially observable Markov decision process, and designs a Deep Decentralized Policy Network (DDPN) to model the polices. To train DDPN effectively, a novel two-stage learning algorithm is proposed which
combines imitation learning from opponent and reinforcement learning by no-regret dynamics.  Extensive experimental results on various combat
scenarios indicate that proposed method can defeat different opponent models and significantly outperforms many state-of-the-art approaches.

关键词Multi-agent Learning Deep Decentralized Policy Network Real-time Combat Game
收录类别SCI
七大方向——子方向分类决策智能理论与方法
文献类型会议论文
条目标识符http://ir.ia.ac.cn/handle/173211/26156
专题智能系统与工程
通讯作者Junliang Xing
作者单位1.Institute of Automation, Chinese Academy of Sciences
2.Horizon Robotics
第一作者单位中国科学院自动化研究所
通讯作者单位中国科学院自动化研究所
推荐引用方式
GB/T 7714
Peixi Peng,Junliang Xing,Lili Cao,et al. Learning Deep Decentralized Policy Network by Collective Rewards for Real-Time Combat Game[C],2019.
条目包含的文件 下载所有文件
文件名称/大小 文献类型 版本类型 开放类型 使用许可
IJCAI19StarCraftFina(762KB)会议论文 开放获取CC BY-NC-SA浏览 下载
个性服务
推荐该条目
保存到收藏夹
查看访问统计
导出为Endnote文件
谷歌学术
谷歌学术中相似的文章
[Peixi Peng]的文章
[Junliang Xing]的文章
[Lili Cao]的文章
百度学术
百度学术中相似的文章
[Peixi Peng]的文章
[Junliang Xing]的文章
[Lili Cao]的文章
必应学术
必应学术中相似的文章
[Peixi Peng]的文章
[Junliang Xing]的文章
[Lili Cao]的文章
相关权益政策
暂无数据
收藏/分享
文件名: IJCAI19StarCraftFinal.pdf
格式: Adobe PDF
此文件暂不支持浏览
所有评论 (0)
暂无评论
 

除非特别说明,本系统中所有内容都受版权保护,并保留所有权利。