Saved in:
Bibliographic Details
Main Authors: Liu, Yuxiang, Cao, Shengfan
Format: Preprint
Published: 2025
Subjects:
Online Access:https://arxiv.org/abs/2512.05335
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866917554202607616
author Liu, Yuxiang
Cao, Shengfan
author_facet Liu, Yuxiang
Cao, Shengfan
contents 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.
format Preprint
id arxiv_https___arxiv_org_abs_2512_05335
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle State-Conditional Adversarial Learning: An Off-Policy Visual Domain Transfer Method for End-to-End Imitation Learning
Liu, Yuxiang
Cao, Shengfan
Robotics
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.
title State-Conditional Adversarial Learning: An Off-Policy Visual Domain Transfer Method for End-to-End Imitation Learning
topic Robotics
url https://arxiv.org/abs/2512.05335