Saved in:
Bibliographic Details
Main Authors: Sawant, Omkar, Zanatta, Luca, Malczyk, Grzegorz, Alexis, Kostas
Format: Preprint
Published: 2026
Subjects:
Online Access:https://arxiv.org/abs/2603.22182
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866917358256259072
author Sawant, Omkar
Zanatta, Luca
Malczyk, Grzegorz
Alexis, Kostas
author_facet Sawant, Omkar
Zanatta, Luca
Malczyk, Grzegorz
Alexis, Kostas
contents This paper presents a cross-modal learning framework that exploits complementary information from depth and grayscale images for robust navigation. We introduce a Cross-Modal Wasserstein Autoencoder that learns shared latent representations by enforcing cross-modal consistency, enabling the system to infer depth-relevant features from grayscale observations when depth measurements are corrupted. The learned representations are integrated with a Reinforcement Learning-based policy for collision-free navigation in unstructured environments when depth sensors experience degradation due to adverse conditions such as poor lighting or reflective surfaces. Simulation and real-world experiments demonstrate that our approach maintains robust performance under significant depth degradation and successfully transfers to real environments.
format Preprint
id arxiv_https___arxiv_org_abs_2603_22182
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle Cross-Modal Reinforcement Learning for Navigation with Degraded Depth Measurements
Sawant, Omkar
Zanatta, Luca
Malczyk, Grzegorz
Alexis, Kostas
Robotics
This paper presents a cross-modal learning framework that exploits complementary information from depth and grayscale images for robust navigation. We introduce a Cross-Modal Wasserstein Autoencoder that learns shared latent representations by enforcing cross-modal consistency, enabling the system to infer depth-relevant features from grayscale observations when depth measurements are corrupted. The learned representations are integrated with a Reinforcement Learning-based policy for collision-free navigation in unstructured environments when depth sensors experience degradation due to adverse conditions such as poor lighting or reflective surfaces. Simulation and real-world experiments demonstrate that our approach maintains robust performance under significant depth degradation and successfully transfers to real environments.
title Cross-Modal Reinforcement Learning for Navigation with Degraded Depth Measurements
topic Robotics
url https://arxiv.org/abs/2603.22182