Behavior Regularized Offline Reinforcement Learning

TL;DR

This paper introduces a general framework called BRAC for offline reinforcement learning, demonstrating that simpler methods can perform as well as complex recent approaches across various continuous control tasks.

Contribution

The paper presents a unified framework for offline RL and shows that many recent complex techniques are unnecessary for strong performance.

Findings

Many recent complex methods are unnecessary for good offline RL performance.

Simple baselines perform comparably to sophisticated algorithms.

Design choices significantly impact offline RL effectiveness.

Abstract

In reinforcement learning (RL) research, it is common to assume access to direct online interactions with the environment. However in many real-world applications, access to the environment is limited to a fixed offline dataset of logged experience. In such settings, standard RL algorithms have been shown to diverge or otherwise yield poor performance. Accordingly, recent work has suggested a number of remedies to these issues. In this work, we introduce a general framework, behavior regularized actor critic (BRAC), to empirically evaluate recently proposed methods as well as a number of simple baselines across a variety of offline continuous control tasks. Surprisingly, we find that many of the technical complexities introduced in recent methods are unnecessary to achieve strong performance. Additional ablations provide insights into which design choices matter most in the offline RL setting.