Cross Domain Robot Imitation with Invariant Representation

TL;DR

This paper introduces an invariant representation-based imitation learning algorithm that enables cross domain robot imitation without human-labeled data, successfully generalizing to new robots in simulation.

Contribution

The paper proposes a novel invariant representation learning method for cross domain imitation learning that does not require pairwise labels and leverages cycle-consistency and domain confusion.

Findings

Successfully generalizes to unseen robots in simulation

Learns similar representations for different robots with similar behaviors

Does not require human-labeled pairwise data

Abstract

Animals are able to imitate each others' behavior, despite their difference in biomechanics. In contrast, imitating the other similar robots is a much more challenging task in robotics. This problem is called cross domain imitation learning~(CDIL). In this paper, we consider CDIL on a class of similar robots. We tackle this problem by introducing an imitation learning algorithm based on invariant representation. We propose to learn invariant state and action representations, which aligns the behavior of multiple robots so that CDIL becomes possible. Compared with previous invariant representation learning methods for similar purpose, our method does not require human-labeled pairwise data for training. Instead, we use cycle-consistency and domain confusion to align the representation and increase its robustness. We test the algorithm on multiple robots in simulator and show that unseen new robot instances can be trained with existing expert demonstrations successfully. Qualitative results also demonstrate that the proposed method is able to learn similar representations for different robots with similar behaviors, which is essential for successful CDIL.