Beyond Non-Expert Demonstrations: Outcome-Driven Action Constraint for Offline Reinforcement Learning

We address the challenge of offline reinforcement learning using realistic data, specifically non-expert data collected through sub-optimal behavior policies. Under such circumstance, the learned policy must be safe enough to manage distribution shift while maintaining sufficient flexibility to deal with non-expert (bad) demonstrations from offlinethis http URLtackle 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, hence reducing the negative impact of those badthis http URLbe specific, a new conservative reward mechanism is developed to deal with distribution shift by evaluating actions according to whether their outcomes meet safety requirements - remaining within the state support area, rather than solely depending on the actions' likelihood based on offlinethis http URLtheoretical justification, we provide empirical evidence on widely used MuJoCo and various maze 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.

@article{jiang2025_2504.01719,
  title={ Beyond Non-Expert Demonstrations: Outcome-Driven Action Constraint for Offline Reinforcement Learning },
  author={ Ke Jiang and Wen Jiang and Yao Li and Xiaoyang Tan },
  journal={arXiv preprint arXiv:2504.01719},
  year={ 2025 }
}