Online vs. Offline Adaptive Domain Randomization Benchmark

TL;DR

This paper introduces a comprehensive benchmark comparing online and offline adaptive domain randomization methods for sim-to-real transfer in robotics, highlighting their strengths and limitations across different settings.

Contribution

It provides the first thorough comparison of adaptive domain randomization methods in both offline and online scenarios, guiding future research and application choices.

Findings

Online methods are limited by policy quality in iterative updates.

Offline methods may fail when replaying trajectories with open-loop commands.

The benchmark helps identify suitable methods for specific tasks and settings.

Abstract

Physics simulators have shown great promise for conveniently learning reinforcement learning policies in safe, unconstrained environments. However, transferring the acquired knowledge to the real world can be challenging due to the reality gap. To this end, several methods have been recently proposed to automatically tune simulator parameters with posterior distributions given real data, for use with domain randomization at training time. These approaches have been shown to work for various robotic tasks under different settings and assumptions. Nevertheless, existing literature lacks a thorough comparison of existing adaptive domain randomization methods with respect to transfer performance and real-data efficiency. In this work, we present an open benchmark for both offline and online methods (SimOpt, BayRn, DROID, DROPO), to shed light on which are most suitable for each setting and task at hand. We found that online methods are limited by the quality of the currently learned policy for the next iteration, while offline methods may sometimes fail when replaying trajectories in simulation with open-loop commands. The code used will be released at https://github.com/gabrieletiboni/adr-benchmark.