Constrained Policy Optimization with Explicit Behavior Density for Offline Reinforcement Learning

Jing Zhang1; Chi Zhang2; Wenjia Wang1,3; Bing-Yi Jing4

2023

37th Conference on Neural Information Processing Systems (NeurIPS)

1049-5258

ADVANCES IN NEURAL INFORMATION PROCESSING SYSTEMS 36 (NEURIPS 2023)

DEC 10-16, 2023

null,New Orleans,LA

10010 NORTH TORREY PINES RD, LA JOLLA, CALIFORNIA 92037 USA

NEURAL INFORMATION PROCESSING SYSTEMS (NIPS)

Due to the inability to interact with the environment, offline reinforcement learning (RL) methods face the challenge of estimating the Out-of-Distribution (OOD) points. Existing methods for addressing this issue either control policy to exclude the OOD action or make the Q function pessimistic. However, these methods can be overly conservative or fail to identify OOD areas accurately. To overcome this problem, we propose a Constrained Policy optimization with Explicit Behavior density (CPED) method that utilizes a flow-GAN model to explicitly estimate the density of behavior policy. By estimating the explicit density, CPED can accurately identify the safe region and enable optimization within the region, resulting in less conservative learning policies. We further provide theoretical results for both the flow-GAN estimator and performance guarantee for CPED by showing that CPED can find the optimal Q-function value. Empirically, CPED outperforms existing alternatives on various standard offline reinforcement learning tasks, yielding higher expected returns.

通讯

英语

CPCI-S

Guangzhou Municipal Science and Technology Project[2023A03J0019] ; null[2023A03J0003]

Computer Science

Computer Science, Artificial Intelligence ; Computer Science, Information Systems

WOS:001220600005004

Web of Science

1.The Hong Kong University of Science and Technology
2.Kuaishou Technologies
3.The Hong Kong University of Science and Technology(Guangzhou)
4.Southern University of Science and Technology

Jing Zhang,Chi Zhang,Wenjia Wang,et al. Constrained Policy Optimization with Explicit Behavior Density for Offline Reinforcement Learning[C]. 10010 NORTH TORREY PINES RD, LA JOLLA, CALIFORNIA 92037 USA:NEURAL INFORMATION PROCESSING SYSTEMS (NIPS),2023.