Vayam Agarwal

Autonomous vehicles are an emerging technology. We haven’t seen full scale deployment of AVs because they haven’t proven their safety. The core problem is perception. AVs collect raw data through sensors such as images from cameras or point cloud data from LiDAR. This data must be perceived to perform functions like object tracking, object detection and localization. The challenge is that current sensors cannot collect data efficiently, the algorithms do not perform with high accuracy, and the visibility by a vehicle is limited to its immediate surroundings. To resolve these problems, one of the proposed solutions is fusing raw data of multiple vehicles. This requires transferring data from one vehicle to another. The problem, however, is that current communication technologies are unable to transfer heavy weight data in a time sensitive manner. A suitable region to perform such data transfers is intersections. Intersections comprise existing hardware and have dedicated communication channels. As autonomous vehicles will be deployed, additional hardware may be available in the region. In this paper, we explore opportunities and challenges of data collection, processing, and transfer in intersections. Data collection involves pre-processing raw data by vehicles to core features and transferring it to intersections. The data collected then gets processed to a suitable format after which it is broadcasted to other vehicles.