中国科学院机构知识库网格系统: Real-Sim-Real Transfer for Real-World Robot Control Policy Learning with Deep Reinforcement Learning

Real-Sim-Real Transfer for Real-World Robot Control Policy Learning with Deep Reinforcement Learning

文献类型：期刊论文


作者	Liu, Naijun3,4 ; Cai, Yinghao3 ; Lu, Tao3 ; Wang, Rui1,3 ; Wang, Shuo2,3,4
刊名	APPLIED SCIENCES-BASEL
出版日期	2020-03-01
卷号	10 期号:5 页码:16
关键词	robot policy learning reality gap simulated environment deep reinforcement learning
DOI	10.3390/app10051555
通讯作者	Cai, Yinghao(yinghao.cai@ia.ac.cn) ; Lu, Tao(tao.lu@ia.ac.cn) ; Wang, Shuo(shuo.wang@ia.ac.cn)
英文摘要	Compared to traditional data-driven learning methods, recently developed deep reinforcement learning (DRL) approaches can be employed to train robot agents to obtain control policies with appealing performance. However, learning control policies for real-world robots through DRL is costly and cumbersome. A promising alternative is to train policies in simulated environments and transfer the learned policies to real-world scenarios. Unfortunately, due to the reality gap between simulated and real-world environments, the policies learned in simulated environments often cannot be generalized well to the real world. Bridging the reality gap is still a challenging problem. In this paper, we propose a novel real-sim-real (RSR) transfer method that includes a real-to-sim training phase and a sim-to-real inference phase. In the real-to-sim training phase, a task-relevant simulated environment is constructed based on semantic information of the real-world scenario and coordinate transformation, and then a policy is trained with the DRL method in the built simulated environment. In the sim-to-real inference phase, the learned policy is directly applied to control the robot in real-world scenarios without any real-world data. Experimental results in two different robot control tasks show that the proposed RSR method can train skill policies with high generalization performance and significantly low training costs.
WOS关键词	DOMAIN ADAPTATION
资助项目	National Natural Science Foundation of China[61773378] ; National Natural Science Foundation of China[U1713222] ; National Natural Science Foundation of China[U1806204] ; Equipment Pre-Research Field Fund[61403120407] ; Opening Project of Guangdong Provincial Key Lab of Robotics and Intelligent System
WOS研究方向	Chemistry ; Engineering ; Materials Science ; Physics
语种	英语
WOS记录号	WOS:000525298100003
出版者	MDPI
资助机构	National Natural Science Foundation of China ; Equipment Pre-Research Field Fund ; Opening Project of Guangdong Provincial Key Lab of Robotics and Intelligent System
源URL	[http://ir.ia.ac.cn/handle/173211/38868]
专题	智能机器人系统研究
通讯作者	Cai, Yinghao; Lu, Tao; Wang, Shuo
作者单位	1.Chinese Acad Sci, Shenzhen Inst Adv Technol, Guangdong Prov Key Lab Robot & Intelligent Syst, Shenzhen 518055, Peoples R China 2.Chinese Acad Sci, Ctr Excellence Brain Sci & Intelligence Technol, Shanghai 200031, Peoples R China 3.Chinese Acad Sci, Inst Automat, State Key Lab Management & Control Complex Syst, Beijing 100190, Peoples R China 4.Univ Chinese Acad Sci, Beijing 100190, Peoples R China
推荐引用方式 GB/T 7714	Liu, Naijun,Cai, Yinghao,Lu, Tao,et al. Real-Sim-Real Transfer for Real-World Robot Control Policy Learning with Deep Reinforcement Learning[J]. APPLIED SCIENCES-BASEL,2020,10(5):16.
APA	Liu, Naijun,Cai, Yinghao,Lu, Tao,Wang, Rui,&Wang, Shuo.(2020).Real-Sim-Real Transfer for Real-World Robot Control Policy Learning with Deep Reinforcement Learning.APPLIED SCIENCES-BASEL,10(5),16.
MLA	Liu, Naijun,et al."Real-Sim-Real Transfer for Real-World Robot Control Policy Learning with Deep Reinforcement Learning".APPLIED SCIENCES-BASEL 10.5(2020):16.

入库方式： OAI收割

来源：自动化研究所

下载0

Real-Sim-Real Transfer for Real-World Robot Control Policy Learning with Deep Reinforcement Learning

其他版本