Constraint-based offline reinforcement learning (RL) involves policy constraints or imposing penalties on the value function to mitigate overestimation errors caused by distributional shift. This paper focuses on a limitation in existing offline RL methods with penalized value function, indicating the potential for underestimation bias due to unnecessary bias introduced in the value function. To address this concern, we propose Exclusively Penalized Q-learning (EPQ), which reduces estimation bias in the value function by selectively penalizing states that are prone to inducing estimation errors. Numerical results show that our method significantly reduces underestimation bias and improves performance in various offline control tasks compared to other offline RL methods
@article{arxiv.2405.14082,
title = {Exclusively Penalized Q-learning for Offline Reinforcement Learning},
author = {Junghyuk Yeom and Yonghyeon Jo and Jungmo Kim and Sanghyeon Lee and Seungyul Han},
journal= {arXiv preprint arXiv:2405.14082},
year = {2024}
}
Comments
10 technical page followed by references and appendix. Accepted to Neurips 2024 as spotlight paper