Reinforcement learning solution for HJB equation arising in constrained optimal control problem

doi:10.1016/j.neunet.2015.08.007

CASIA OpenIR > 多模态人工智能系统全国重点实验室 > 复杂系统智能机理与平行控制团队

	Reinforcement learning solution for HJB equation arising in constrained optimal control problem
	Luo, Biao1 ; Wu, Huai-Ning 2; Huang, Tingwen 3; Liu, Derong 4
发表期刊	NEURAL NETWORKS
	2015-11-01
卷号	71 期号:0 页码:150-158
文章类型	Article
摘要	The constrained optimal control problem depends on the solution of the complicated Hamilton-Jacobi-Bellman equation (HJBE). In this paper, a data-based off-policy reinforcement learning (RL) method is proposed, which learns the solution of the HJBE and the optimal control policy from real system data. One important feature of the off-policy RL is that its policy evaluation can be realized with data generated by other behavior policies, not necessarily the target policy, which solves the insufficient exploration problem. The convergence of the off-policy RL is proved by demonstrating its equivalence to the successive approximation approach. Its implementation procedure is based on the actor-critic neural networks structure, where the function approximation is conducted with linearly independent basis functions. Subsequently, the convergence of the implementation procedure with function approximation is also proved. Finally, its effectiveness is verified through computer simulations. (C) 2015 Elsevier Ltd. All rights reserved.
关键词	Constrained Optimal Control Data-based Off-policy Reinforcement Learning Hamilton-jacobi-bellman Equation The Method Of Weighted Residuals
WOS标题词	Science & Technology ; Technology ; Life Sciences & Biomedicine
DOI	10.1016/j.neunet.2015.08.007
关键词[WOS]	TIME NONLINEAR-SYSTEMS ; ADAPTIVE OPTIMAL-CONTROL ; DYNAMIC-PROGRAMMING ALGORITHM ; POLICY ITERATION ; INPUT CONSTRAINTS ; LINEAR-SYSTEMS ; CONTROL DESIGN ; STABILIZATION
收录类别	SCI
语种	英语
项目资助者	National Natural Science Foundation of China(61233001 ; Beijing Natural Science Foundation(4132078) ; Early Career Development Award of SKLMCCS ; NPRP grant from the Qatar National Research Fund(NPRP 4-1162-1-181) ; 61273140 ; 61304086 ; 61374105)
WOS研究方向	Computer Science ; Neurosciences & Neurology
WOS类目	Computer Science, Artificial Intelligence ; Neurosciences
WOS记录号	WOS:000364160900014
引用统计	被引频次：94[WOS] [WOS记录] [WOS相关记录]
文献类型	期刊论文
条目标识符	http://ir.ia.ac.cn/handle/173211/10729
专题	多模态人工智能系统全国重点实验室_复杂系统智能机理与平行控制团队
通讯作者	Liu, Derong
作者单位	1.Chinese Acad Sci, Inst Automat, State Key Lab Management & Control Complex Syst, Beijing 100190, Peoples R China 2.Beijing Univ Aeronaut & Astronaut, Beihang Univ, Sci & Technol Aircraft Control Lab, Beijing 100191, Peoples R China 3.Texas A&M Univ Qatar, Doha, Qatar 4.Univ Sci & Technol Beijing, Sch Automat & Elect Engn, Beijing 100083, Peoples R China
第一作者单位	中国科学院自动化研究所
推荐引用方式 GB/T 7714	Luo, Biao,Wu, Huai-Ning,Huang, Tingwen,et al. Reinforcement learning solution for HJB equation arising in constrained optimal control problem[J]. NEURAL NETWORKS,2015,71(0):150-158.
APA	Luo, Biao,Wu, Huai-Ning,Huang, Tingwen,&Liu, Derong.(2015).Reinforcement learning solution for HJB equation arising in constrained optimal control problem.NEURAL NETWORKS,71(0),150-158.
MLA	Luo, Biao,et al."Reinforcement learning solution for HJB equation arising in constrained optimal control problem".NEURAL NETWORKS 71.0(2015):150-158.

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
2015NN Reinforcement（530KB）	期刊论文	作者接受稿	开放获取	CC BY-NC-SA	浏览下载