Data Analogies Enable Efficient Cross-Embodiment Transfer

TL;DR

This paper investigates how different types of demonstration data, especially data analogies, improve cross-embodiment transfer in generalist robot policies, showing that structured paired demonstrations significantly enhance transfer success.

Contribution

The study reveals that data analogies with paired demonstrations outperform unstructured data in enabling robot embodiment transfer, and demonstrates a practical method to improve real-world transfer success.

Findings

Data analogies significantly improve transfer success.

Broader diversity benefits perceptual shifts more than morphology shifts.

Structured paired demonstrations outperform unstructured data.

Abstract

Generalist robot policies are trained on demonstrations collected across a wide variety of robots, scenes, and viewpoints. Yet it remains unclear how to best organize and scale such heterogeneous data so that it genuinely improves performance in a given target setting. In this work, we ask: what form of demonstration data is most useful for enabling transfer across robot set-ups? We conduct controlled experiments that vary end-effector morphology, robot platform appearance, and camera perspective, and compare the effects of simply scaling the number of demonstrations against systematically broadening the diversity in different ways. Our simulated experiments show that while perceptual shifts such as viewpoint benefit most from broad diversity, morphology shifts benefit far less from unstructured diversity and instead see the largest gains from data analogies, i.e. paired demonstrations that align scenes, tasks, and/or trajectories across different embodiments. Informed by the simulation results, we improve real-world cross-embodiment transfer success by an average of $22.5\%$ over large-scale, unpaired datasets by changing only the composition of the data.