On the Role of Sensor Fusion for Object Detection in Future Vehicular Networks

Fully autonomous driving systems require fast detection and recognition of\nsensitive objects in the environment. In this context, intelligent vehicles\nshould share their sensor data with computing platforms and/or other vehicles,\nto detect objects beyond their own sensors' fields of view. However, the\nresulting huge volumes of data to be exchanged can be challenging to handle for\nstandard communication technologies. In this paper, we evaluate how using a\ncombination of different sensors affects the detection of the environment in\nwhich the vehicles move and operate. The final objective is to identify the\noptimal setup that would minimize the amount of data to be distributed over the\nchannel, with negligible degradation in terms of object detection accuracy. To\nthis aim, we extend an already available object detection algorithm so that it\ncan consider, as an input, camera images, LiDAR point clouds, or a combination\nof the two, and compare the accuracy performance of the different approaches\nusing two realistic datasets. Our results show that, although sensor fusion\nalways achieves more accurate detections, LiDAR only inputs can obtain similar\nresults for large objects while mitigating the burden on the channel.\n