A Workflow for Offline Model-Free Robotic Reinforcement Learning

TL;DR

This paper presents a practical workflow for offline reinforcement learning in robotics, including metrics and guidelines to improve policy performance without online tuning, validated on simulated and real robotic tasks.

Contribution

It introduces a set of metrics and a systematic workflow for offline RL that guides design choices and hyperparameters without online evaluation, bridging the gap to real-world robotic applications.

Findings

Effective policies learned without online tuning

Workflow improves offline RL performance in robotic tasks

Validated on simulated and real robots with raw image inputs

Abstract

Offline reinforcement learning (RL) enables learning control policies by utilizing only prior experience, without any online interaction. This can allow robots to acquire generalizable skills from large and diverse datasets, without any costly or unsafe online data collection. Despite recent algorithmic advances in offline RL, applying these methods to real-world problems has proven challenging. Although offline RL methods can learn from prior data, there is no clear and well-understood process for making various design choices, from model architecture to algorithm hyperparameters, without actually evaluating the learned policies online. In this paper, our aim is to develop a practical workflow for using offline RL analogous to the relatively well-understood workflows for supervised learning problems. To this end, we devise a set of metrics and conditions that can be tracked over the course of offline training, and can inform the practitioner about how the algorithm and model architecture should be adjusted to improve final performance. Our workflow is derived from a conceptual understanding of the behavior of conservative offline RL algorithms and cross-validation in supervised learning. We demonstrate the efficacy of this workflow in producing effective policies without any online tuning, both in several simulated robotic learning scenarios and for three tasks on two distinct real robots, focusing on learning manipulation skills with raw image observations with sparse binary rewards. Explanatory video and additional results can be found at sites.google.com/view/offline-rl-workflow