Cameras and LiDAR degrade in rain, fog, and snow, while millimeter-wave radar remains largely unaffected. We align a radar encoder to frozen SigLIP vision embeddings and decode structured scene captions through a frozen vision-language model (VLM) with approximately 7M trainable parameters. On K-RADAR with held-out fog, light snow, and heavy snow sequences, all radar configurations outperform a camera baseline that collapses to over 90% hallucination. We identify a token-norm mismatch as the dominant failure mode when bridging radar to a frozen VLM and show that projector-output LayerNorm resolves it. Analysis of encoder complexity, caption format, and pooling strategy reveals tradeoffs that inform future radar-VLM pipeline design.
Cite
@article{arxiv.2605.07367,
title = {Weather-Robust Scene Semantics with Vision-Aligned 4D Radar},
author = {Kali Hamilton and Christoffer Heckman},
journal= {arXiv preprint arXiv:2605.07367},
year = {2026}
}
Comments
5 pages + references, 2 appendix pages. ICRA 2026 Radar in Robotics Workshop