Outcome-Driven Action Flexibility for Robust Offline Reinforcement Learning

Ke Jiang, Xiaoyang Tan

Keywords: Offline reinforcement learning; Robust reinforcement learning;Outcome-driven action flexibility;

Abstract:

We address the challenge of offline reinforcement learning using realistic data, specifically non-expert data collected through sub-optimal behavior policies. A primary concern is that the learned policy must be conservative enough to manage \textit{distribution shift} while maintaining sufficient flexibility for generalization. To tackle this issue, we introduce a novel method called Outcome-Driven Action Flexibility (ODAF), which seeks to reduce reliance on the empirical action distribution of the behavior policy.
Specifically, we develop a new reward mechanism that evaluates whether the subsequent states, following the current policy, meet specified performance requirements (e.g., safety—remaining within the state support area), rather than solely depending on the characteristics of the actions taken (e.g., whether the action imitates the behavior policy).
Besides theoretical justification, we provide empirical evidence on widely used D4RL benchmarks, demonstrating that our ODAF method, implemented using uncertainty quantification techniques, effectively tolerates unseen transitions for improved "trajectory stitching," while enhancing the agent's ability to learn from realistic non-expert data.

Primary Area: reinforcement learning

Code Of Ethics: I acknowledge that I and all co-authors of this work have read and commit to adhering to the ICLR Code of Ethics.

Submission Guidelines: I certify that this submission complies with the submission instructions as described on https://iclr.cc/Conferences/2025/AuthorGuide.

Reciprocal Reviewing: I understand the reciprocal reviewing requirement as described on https://iclr.cc/Conferences/2025/CallForPapers. If none of the authors are registered as a reviewer, it may result in a desk rejection at the discretion of the program chairs. To request an exception, please complete this form at https://forms.gle/Huojr6VjkFxiQsUp6.

Anonymous Url: I certify that there is no URL (e.g., github page) that could be used to find authors’ identity.

No Acknowledgement Section: I certify that there is no acknowledgement section in this submission for double blind review.

Submission Number: 5683