LiDAR Teach, Radar Repeat: Robust Cross-Modal Navigation in Degenerate and Varying Environments

2026-05-04

AI Summary

This work addresses environmental degradation in long-term autonomous navigation caused by adverse weather, dynamic disturbances, and structural changes. The authors propose LTR², a novel system that constructs high-precision maps using LiDAR during a teaching phase and achieves robust localization with a 4D millimeter-wave radar during repeated traversals. Key innovations include the first cross-modal, cross-platform LiDAR-teaching/radar-repetition navigation framework, a cross-modal registration network integrating Doppler motion priors with physical models of radar power and LiDAR intensity, and an unsupervised adaptive online fine-tuning strategy. Evaluated over 40+ kilometers across six months on multiple platforms, LTR² achieves centimeter-level localization accuracy and significantly outperforms existing methods under challenging conditions such as nighttime and smoke. It also sets a new state-of-the-art in cross-modal registration on public benchmarks.

Abstract

Long-term autonomy requires robust navigation in environments subject to dynamic and static changes, as well as adverse weather conditions. Teach-and-Repeat (T\&R) navigation offers a reliable and cost-effective solution by avoiding the need for consistent global mapping; however, existing T\&R systems lack a systematic solution to tackle various environmental variations such as weather degradation, ephemeral dynamics, and structural changes. This work proposes LTR$^2$, the first cross-modal, cross-platform LiDAR-Teach-and-Radar-Repeat system that systematically addresses these challenges. LTR$^2$ leverages LiDAR during the teaching phase to capture precise structural information under normal conditions and utilizes 4D millimeter-wave radar during the repeating phase for robust operation under environmental degradations. To align sparse and noisy forward-looking 4D radar with dense and accurate omnidirectional 3D LiDAR data, we introduce a Cross-Modal Registration (CMR) network that jointly exploits Doppler-based motion priors and the physical laws governing LiDAR intensity and radar power density. Furthermore, we propose an adaptive fine-tuning strategy that incrementally updates the CMR network based on localization errors, enabling long-term adaptability to static environmental changes without ground-truth labels. We demonstrate that the proposed CMR network achieves state-of-the-art cross-modal registration performance on the open-access dataset. Then we validate LTR$^2$ across three robot platforms over a large-scale, long-term deployment (40+ km over 6 months), including challenging conditions such as nighttime smoke. Experimental results and ablation studies demonstrate centimeter-level accuracy and strong robustness against diverse environmental disturbances, significantly outperforming existing approaches.

Problem

• long-term autonomy
• environmental variations
• cross-modal navigation
• teach-and-repeat
• robustness

Innovation

• Cross-Modal Registration
• Teach-and-Repeat Navigation
• 4D Millimeter-Wave Radar
• LiDAR-Radar Fusion
• Adaptive Fine-Tuning