LRFusionPR: A Polar BEV-Based LiDAR-Radar Fusion Network for Place Recognition

TL;DR

LRFusionPR tackles GPS-denied place recognition by fusing LiDAR and radar through a unified polar BEV representation. It introduces a dual-branch architecture with cross-attention-based fusion and a radar-only distillation path guided by a structure-aware loss, producing rotation-invariant, robust descriptors. Extensive experiments across four datasets and adverse weather demonstrate state-of-the-art accuracy and robust performance, with real-time inference. The approach accommodates heterogeneous radar types and provides open-source code for reproducibility.

Abstract

In autonomous driving, place recognition is critical for global localization in GPS-denied environments. LiDAR and radar-based place recognition methods have garnered increasing attention, as LiDAR provides precise ranging, whereas radar excels in adverse weather resilience. However, effectively leveraging LiDAR-radar fusion for place recognition remains challenging. The noisy and sparse nature of radar data limits its potential to further improve recognition accuracy. In addition, heterogeneous radar configurations complicate the development of unified cross-modality fusion frameworks. In this paper, we propose LRFusionPR, which improves recognition accuracy and robustness by fusing LiDAR with either single-chip or scanning radar. Technically, a dual-branch network is proposed to fuse different modalities within the unified polar coordinate bird's eye view (BEV) representation. In the fusion branch, cross-attention is utilized to perform cross-modality feature interactions. The knowledge from the fusion branch is simultaneously transferred to the distillation branch, which takes radar as its only input to further improve the robustness. Ultimately, the descriptors from both branches are concatenated, producing the multimodal global descriptor for place retrieval. Extensive evaluations on multiple datasets demonstrate that our LRFusionPR achieves accurate place recognition, while maintaining robustness under varying weather conditions. Our open-source code will be released at https://github.com/QiZS-BIT/LRFusionPR.

Figures (6)

• Figure 1: The fusion of LiDAR and radar enables the acquisition of fine-grained scene structure while improving weather robustness. In LRFusionPR, multimodal data are aligned into a polar BEV format, enabling feature fusion via cross-attention and cross-modality distillation. This results in multimodal global descriptors for accurate and robust place recognition.
• Figure 2: The overall architecture of LRFusionPR. The LiDAR and radar data are unified into polar BEV representations, and then fed into the ResNet encoders. Subsequently, the fusion branch exploits cross-modality feature correlations, and the distillation branch generates radar-based descriptors. The discriminability of the radar-based descriptors is enhanced by structure-aware distillation. Ultimately, the descriptors from both branches are concatenated to achieve accurate and robust place recognition.
• Figure 3: The pipeline of the polar cross-attention module. It consists of two cross-attention branches to facilitate feature interaction of the feature map, thus exploiting cross-modality correlations.
• Figure 4: The variation of the Recall@1 as a function of visibility under fog on SQ (left) and Sejong (right) splits.
• Figure 5: Visualization of place recognition in foggy weather. Each row in the figure corresponds to a different scene. The green borders indicate correct retrieval results, while the red borders represent incorrect retrievals. We use red points to highlight the noise introduced by fog.