Title: Enhancing Adverse Weather Resilience in Cooperative Perception through the Integration of 4D Radar, LiDAR, and Cameras

Abstract: While cooperative perception is a critical component of autonomous driving, its reliability often diminishes when camera and LiDAR sensors suffer performance degradation under harsh weather conditions. To mitigate this vulnerability, this study incorporates 4D imaging radar as a resilient sensing modality into collaborative perception frameworks, supplemented by a novel Doppler-guided spatial attention mechanism designed to facilitate multi-agent data fusion. The proposed methodology builds upon two established backbone architectures: one where radar acts as a direct replacement for LiDAR in a radar-camera configuration, and another where radar serves as a complementary sensor to LiDAR in a LiDAR-radar setup. To facilitate rigorous assessment, we introduce OPV2V-R and Adver-City-R, two new benchmarks augmented with radar data and featuring physics-based simulations of LiDAR degradation. Experimental results indicate significant improvements in robustness during fog and rain, particularly in scenarios where radar compensates for compromised LiDAR signals. Further validation using the MAN TruckScenes dataset confirms the model’s effectiveness in real-world environments beyond simulated settings. These findings underscore the potential of 4D imaging radar to enable reliable, all-weather collaborative perception. The associated code and datasets can be accessed at: https://url.fzi.de/SlimComm.

Source: arXiv Generated at: 2026-06-02 00:00:00 UTC