(745.34 KB) Abstract
Autonomous driving has made significant progress in recent years, but adverse weather conditions still remain a major challenge for perception and decision-making algorithms. This work presents a multimodal data acquisition prototype designed to enhance autonomous vehicle perception in challenging environments such as fog, heavy rain, and snow. The system is mounted on a Dacia Duster and features a diverse sensor suite including visible, Short-Wave Infra-Red, thermal, and polarimetric cameras with a solid-state LiDAR and automotive radars. The whole system is calibrated with dedicated calibration boards developed on purpose to enable data fusion across all modalities. Initial results demonstrate the effectiveness of our approach, showing that critical environmental features can be detected reliably across various weather conditions. Future work includes the release of a fully annotated multimodal dataset to support further research in adverse weather perception.