Imitation Learning from Observations by Minimizing Inverse Dynamics Disagreement

TL;DR

This paper introduces a novel approach called IDDM for imitation learning from observations, which minimizes inverse dynamics disagreement to bridge the gap with traditional learning from demonstration, showing improved performance on benchmarks.

Contribution

The paper provides a theoretical analysis of the difference between LfO and LfD, and proposes a practical, model-free method to minimize inverse dynamics disagreement, enhancing LfO performance.

Findings

IDDM reduces the gap between LfO and LfD.

Empirical results show IDDM outperforms other LfO methods.

The method is effective on challenging benchmarks.

Abstract

This paper studies Learning from Observations (LfO) for imitation learning with access to state-only demonstrations. In contrast to Learning from Demonstration (LfD) that involves both action and state supervision, LfO is more practical in leveraging previously inapplicable resources (e.g. videos), yet more challenging due to the incomplete expert guidance. In this paper, we investigate LfO and its difference with LfD in both theoretical and practical perspectives. We first prove that the gap between LfD and LfO actually lies in the disagreement of inverse dynamics models between the imitator and the expert, if following the modeling approach of GAIL. More importantly, the upper bound of this gap is revealed by a negative causal entropy which can be minimized in a model-free way. We term our method as Inverse-Dynamics-Disagreement-Minimization (IDDM) which enhances the conventional LfO method through further bridging the gap to LfD. Considerable empirical results on challenging benchmarks indicate that our method attains consistent improvements over other LfO counterparts.