Beyond OOD State Actions: Supported Cross-Domain Offline Reinforcement Learning

TL;DR

This paper introduces BOSA, a method for cross-domain offline RL that addresses OOD transition dynamics, significantly improving data efficiency by leveraging source domain data from different environments.

Contribution

The paper proposes BOSA, a novel approach to handle OOD transition dynamics in cross-domain offline RL, enhancing data efficiency and compatibility with existing techniques.

Findings

BOSA achieves 74.4% of SOTA performance using only 10% of target data.

BOSA can be integrated with model-based offline RL and data augmentation methods.

Cross-domain data can significantly improve offline RL efficiency.

Abstract

Offline reinforcement learning (RL) aims to learn a policy using only pre-collected and fixed data. Although avoiding the time-consuming online interactions in RL, it poses challenges for out-of-distribution (OOD) state actions and often suffers from data inefficiency for training. Despite many efforts being devoted to addressing OOD state actions, the latter (data inefficiency) receives little attention in offline RL. To address this, this paper proposes the cross-domain offline RL, which assumes offline data incorporate additional source-domain data from varying transition dynamics (environments), and expects it to contribute to the offline data efficiency. To do so, we identify a new challenge of OOD transition dynamics, beyond the common OOD state actions issue, when utilizing cross-domain offline data. Then, we propose our method BOSA, which employs two support-constrained objectives to address the above OOD issues. Through extensive experiments in the cross-domain offline RL setting, we demonstrate BOSA can greatly improve offline data efficiency: using only 10\% of the target data, BOSA could achieve {74.4\%} of the SOTA offline RL performance that uses 100\% of the target data. Additionally, we also show BOSA can be effortlessly plugged into model-based offline RL and noising data augmentation techniques (used for generating source-domain data), which naturally avoids the potential dynamics mismatch between target-domain data and newly generated source-domain data.