State-Conditional Adversarial Learning: An Off-Policy Visual Domain Transfer Method for End-to-End Imitation Learning

TL;DR

This paper introduces State-Conditional Adversarial Learning, a novel off-policy adversarial method for visual domain transfer in end-to-end imitation learning, addressing challenges with scarce, expert-free target data.

Contribution

It provides a theoretical analysis linking imitation loss to a state-conditional KL divergence and proposes a new adversarial framework to effectively align latent distributions conditioned on system state.

Findings

SCAL achieves robust transfer in diverse autonomous driving environments.

It demonstrates strong sample efficiency in off-policy visual domain transfer.

The method outperforms existing approaches in challenging visual imitation tasks.

Abstract

We study visual domain transfer for end-to-end imitation learning in a realistic and challenging setting where target-domain data are strictly off-policy, expert-free, and scarce. We first provide a theoretical analysis showing that the target-domain imitation loss can be upper bounded by the source-domain loss plus a state-conditional latent KL divergence between source and target observation models. Guided by this result, we propose State- Conditional Adversarial Learning, an off-policy adversarial framework that aligns latent distributions conditioned on system state using a discriminator-based estimator of the conditional KL term. Experiments on visually diverse autonomous driving environments built on the BARC-CARLA simulator demonstrate that SCAL achieves robust transfer and strong sample efficiency.