Search papers, labs, and topics across Lattice.
This paper introduces RadarMOT, a 3D multi-object tracking framework that explicitly leverages radar point clouds to enhance state estimation and recover missed detections, particularly at long ranges and in adverse weather. RadarMOT fuses radar data as an additional observation, rather than treating it as a learned feature, to maintain robustness when other sensors degrade. Experiments on the MAN-TruckScenes dataset demonstrate a 12.7% absolute improvement in AMOTA at long range and a 10.3% improvement in adverse weather conditions.
Radar data, often relegated to a learned feature, can unlock significant gains in 3D multi-object tracking robustness, especially when other sensors falter in adverse conditions or at long ranges.
The challenge of 3D multi-object tracking (3D MOT) is achieving robustness in real-world applications, for example under adverse conditions and maintaining consistency as distance increases. To overcome these challenges, sensor fusion approaches that combine LiDAR, cameras, and radar have emerged. However, existing multi-modal fusion methods usually treat radar as another learned feature inside the network. When the overall model degrades in difficult environmental conditions, the robustness advantages that radar could provide are also reduced. We propose RadarMOT, a radar-informed 3D MOT framework that explicitly uses radar point cloud data as additional observation to refine state estimation and recover detector misses at long ranges. Evaluations on the MAN-TruckScenes dataset show that RadarMOT consistently improves the Average Multi-Object Tracking Accuracy (AMOTA) with absolute 12.7% at long range and 10.3% in adverse weather. The code will be available at https://github.com/bingxue-xu/radarmot
