Varyani, Nitin2024-08-222024-08-222024-06https://hdl.handle.net/11299/265181University of Minnesota Ph.D. dissertation.June 2024. Major: Computer Science. Advisor: Zhi-Li Zhang. 1 computer file (PDF); ix, 84 pages.The remote operation, or teleoperation, of autonomous vehicles (AVs) via emerging 5G networks is a critical application area.This thesis provides an in-depth exploration of the challenges and opportunities surrounding autonomous vehicle (AV) teleoperation, focusing on the integration of emerging technologies such as 5G networks, mobile edge computing, and inter-data-center backbone networks. This thesis embarks on a systematic investigation into the feasibility of AV teleoperations over commercial 5G networks, focusing on cross-layer and end-to-end perspectives. Emphasizing the timely delivery of sensor data, such as camera and LIDAR data, we analyze the impact of low-layer 5G radio network factors, including channel conditions, radio resource allocation, and handovers, on end-to-end latency performance and video perceptual quality. Furthermore, we explore the effectiveness of data compression and adaptive bitrate (ABR) adaptation mechanisms in reducing end-to-end latency while quantifying latency-and-perceptual-quality trade-offs and their impacts on downstream AI tasks such as object detection and recognition. Moreover, we investigate the potential benefits of additional end-system mechanisms in improving tail latency performance. Our investigation reveals the limitations of existing sensor data streaming mechanisms and the challenges posed by 5G networks, providing valuable insights to guide the co-design of future-generation wireless networks, end/edge cloud systems, and applications aimed at overcoming low-latency barriers in AV teleoperations. Mobile edge computing, with resources situated closer to the network edge, facilitates low-latency data processing and decision-making, crucial for AV teleoperation. Recognizing the imperative of low latency and high bandwidth for AV teleoperation, we propose an efficient Quality of Service (QoS)-based routing algorithm tailored for software-defined overlay networks in Mobile Edge Cloud environments. The introduction of software-defined networking (SDN) offers centralized and streamlined traffic management, vital for ensuring optimal QoS. However, existing routing algorithms encounter difficulties in adapting to the dynamic overlay link QoS characteristics and scaling effectively with respect to route computation time and routing entries. In response, we introduce QROUTE, an efficient routing scheme employing a directed-acyclic-graph (DAG)-based approach to minimize route computation time and a QoS-metric-based forwarding scheme to reduce forwarding table entries. As AV teleoperation increasingly relies on inter-data-center backbone networks, we confront the evolving needs of these infrastructures. CAVs generate vast amounts of sensor data requiring real-time processing and analysis for safe and efficient vehicle operation. Inter-data-center backbone networks provide infrastructure for processing and storing this data in distributed data centers equipped with high-performance computing resources. By leveraging these capabilities, complex algorithms for perception, decision-making, and control can be executed efficiently, contributing to the overall teleoperation process. Such transition of traffic in inter-data center backbone networks from bandwidth-intensive to latency-sensitive traffic closely tied to end-user experience requires rethinking traditional fair bandwidth allocation policies. Traditional fair bandwidth allocation policies are no longer adequate for latency-sensitive traffic, necessitating the introduction of the `fair share of latency' concept. Our linear-programming-based routing algorithm for inter-data-center backbone networks integrates both fair share of latency and fair allocation of bandwidth. Simulation results on networks of major providers demonstrate significant improvements in meeting latency Service Level Objectives (SLOs) with a modest reduction in bandwidth allocation fairness. In summary, this thesis investigates the feasibility of autonomous vehicle teleoperations over commercial 5G networks and proposes efficient routing mechanisms for both mobile edge cloud and inter-data-center backbone networks. We address the challenges posed by dynamic network conditions and evolving infrastructures, proposing innovative solutions such as QoS-based routing algorithms for mobile edge cloud environments and fair share of latency concepts for inter-data-center backbone networks. By providing insights into the limitations of existing sensor data streaming mechanisms and offering efficient strategies for optimizing end-to-end QoS performance, our work aims to guide the development of future cellular networks, edge cloud systems, and inter-data-center backbone networks tailored for AV teleoperation, ultimately enhancing the safety and efficiency of autonomous vehicle operations.enThesis or Dissertation