An Experimental and Feasibility Study of Teleoperated Autonomous Vehicles

Abstract

Remote driving, or teleoperating autonomous vehicles (AVs), is a key application that emerging 5G networks aim to support. In this report, we conduct a systematic feasibility study of AV teleoperations over commercial 5G networks from both cross-layer and end-to-end perspectives. Given the critical importance of the timely delivery of sensor data, such as camera and LIDAR data, for AV teleoperations, our focus is on the performance of uplink sensor data delivery. We evaluate the effectiveness of data compression and adaptive bitrate (ABR) adaptation mechanisms - commonly used in today's real-time video streaming applications - in reducing end-to-end latency. We also quantify the impact of latency on video perceptual quality. Additionally, we analyze the impacts of low-layer 5G radio network factors, including channel conditions, radio resource allocation, and handovers, on end-to-end latency performance. Furthermore, we explore the potential benefits of additional end-system mechanisms, such as selective frame dropping and leveraging multiple 5G operators, in improving tail latency performance. Finally, we investigate the challenges posed by multiple AVs competing for radio resources. Our study reveals the limitations of existing sensor data streaming mechanisms and the challenges posed by 5G networks. The insights obtained will help guide the co-design of future-generation wireless networks, end/edge cloud systems, and applications to overcome the low-latency barriers in AV teleoperations.