Timezone: »

Cross-Domain Policy Adaptation via Value-Guided Data Filtering
Kang Xu · Chenjia Bai · Xiaoteng Ma · Dong Wang · Bin Zhao · Zhen Wang · Xuelong Li · Wei Li

Thu Dec 14 08:45 AM -- 10:45 AM (PST) @ Great Hall & Hall B1+B2 #1313
Event URL: https://github.com/Kavka1/VGDF »

Generalizing policies across different domains with dynamics mismatch poses a significant challenge in reinforcement learning. For example, a robot learns the policy in a simulator, but when it is deployed in the real world, the dynamics of the environment may be different. Given the source and target domain with dynamics mismatch, we consider the online dynamics adaptation problem, in which case the agent can access sufficient source domain data while online interactions with the target domain are limited. Existing research has attempted to solve the problem from the dynamics discrepancy perspective. In this work, we reveal the limitations of these methods and explore the problem from the value difference perspective via a novel insight on the value consistency across domains. Specifically, we present the Value-Guided Data Filtering (VGDF) algorithm, which selectively shares transitions from the source domain based on the proximity of paired value targets across the two domains. Empirical results on various environments with kinematic and morphology shifts demonstrate that our method achieves superior performance compared to prior approaches.

Author Information

Kang Xu (Fudan University)
Chenjia Bai (Shanghai AI Laboratory)
Xiaoteng Ma (Department of Automation, Tsinghua University)
Dong Wang (Shanghai AI Laboratory)
Bin Zhao (Northwestern Polytechnical University)
Zhen Wang (Northwestern Polytechnical University)
Xuelong Li (Northwestern Polytechnical University)
Wei Li (Fudan University)

More from the Same Authors