Representation Alignment from Human Feedback for Cross-Embodiment Reward Learning from Mixed-Quality Demonstrations

TL;DR

This paper explores how to learn transferable reward functions from mixed-quality video demonstrations across different embodiments, emphasizing the role of human feedback in aligning representations for effective cross-embodiment reinforcement learning.

Contribution

It introduces techniques leveraging human feedback to improve representation alignment and reward transfer in cross-embodiment inverse reinforcement learning from mixed-quality demonstrations.

Findings

Prior methods struggle with mixed-quality data.

Human feedback improves reward representation alignment.

Different techniques lead to qualitatively different reward behaviors.

Abstract

We study the problem of cross-embodiment inverse reinforcement learning, where we wish to learn a reward function from video demonstrations in one or more embodiments and then transfer the learned reward to a different embodiment (e.g., different action space, dynamics, size, shape, etc.). Learning reward functions that transfer across embodiments is important in settings such as teaching a robot a policy via human video demonstrations or teaching a robot to imitate a policy from another robot with a different embodiment. However, prior work has only focused on cases where near-optimal demonstrations are available, which is often difficult to ensure. By contrast, we study the setting of cross-embodiment reward learning from mixed-quality demonstrations. We demonstrate that prior work struggles to learn generalizable reward representations when learning from mixed-quality data. We then analyze several techniques that leverage human feedback for representation learning and alignment to enable effective cross-embodiment learning. Our results give insight into how different representation learning techniques lead to qualitatively different reward shaping behaviors and the importance of human feedback when learning from mixed-quality, mixed-embodiment data.