Design and validation of a radar–camera fusion framework for object detection

Details

Supervisors

Macq, Benoît ; Craeye, Christophe

Faculty

Ecole polytechnique de Louvain

Degree label

Master [120] : ingénieur civil électricien, à finalité spécialisée

Abstract

enThis thesis studies radar-camera multimodal fusion for object detection in autonomous driving. Reliable perception remains a major challenge for autonomous systems, especially when they must operate in complex or degraded environments. While single-modality approaches can achieve strong performance, they are inherently limited by the weaknesses of the sensor on which they rely. Multimodal fusion offers a promising way to combine complementary information, but its benefits depend on the ability of the architecture to effectively exploit each modality. Through a theoretical and experimental study of recent perception methods, this work analyzes the potential and limitations of radar-camera fusion compared with camera-only perception. The study provides insight into the conditions under which multimodal fusion can be beneficial, while also highlighting the challenges that remain for designing robust perception systems