TD3 with Reverse KL Regularizer for Offline Reinforcement Learning from Mixed Datasets

TL;DR

This paper introduces a novel offline RL method using adaptively weighted reverse KL divergence as a behavior cloning regularizer, effectively balancing RL and BC signals and avoiding OOD actions in mixed datasets.

Contribution

It proposes a new approach with per-state adaptive weighting and mode-seeking reverse KL regularization to improve offline RL from mixed datasets.

Findings

Outperforms existing offline RL algorithms on MuJoCo tasks.

Effectively balances RL and BC signals with per-state weights.

Avoids out-of-distribution actions through mode-seeking regularization.

Abstract

We consider an offline reinforcement learning (RL) setting where the agent need to learn from a dataset collected by rolling out multiple behavior policies. There are two challenges for this setting: 1) The optimal trade-off between optimizing the RL signal and the behavior cloning (BC) signal changes on different states due to the variation of the action coverage induced by different behavior policies. Previous methods fail to handle this by only controlling the global trade-off. 2) For a given state, the action distribution generated by different behavior policies may have multiple modes. The BC regularizers in many previous methods are mean-seeking, resulting in policies that select out-of-distribution (OOD) actions in the middle of the modes. In this paper, we address both challenges by using adaptively weighted reverse Kullback-Leibler (KL) divergence as the BC regularizer based on the TD3 algorithm. Our method not only trades off the RL and BC signals with per-state weights (i.e., strong BC regularization on the states with narrow action coverage, and vice versa) but also avoids selecting OOD actions thanks to the mode-seeking property of reverse KL. Empirically, our algorithm can outperform existing offline RL algorithms in the MuJoCo locomotion tasks with the standard D4RL datasets as well as the mixed datasets that combine the standard datasets.