Performance Evaluation of a Video Streaming Architecture for Vehicle Teleoperation under Constrained Network Conditions

This thesis evaluates a video streaming pipeline designed for the teleportation of heavy vehicles in constrained environments. The analysis was divided into two parts: the first examined different video codecs together with transport layer protocols under challenging network conditions, while the second assessed the quality of information retained by video compression when bandwidth was limited and packet loss was high. Measurement systems were also considered, including a low-cost setup for Glass-to-Glass latency (G-2-G) evaluation and the use of an AI-based model (VMAF) to assess Quality of Experience (QoE). Video codec tests focused on AV1 and H.264, showing that AV1 provided higher quality but performed worse under constrained conditions. TCP with BBR congestion control demonstrated strong performance in limited networks, while RTP remained preferable under normal conditions. The ultimate goal of this study was to build a reusable test bench capable of evaluating newer solutions while maintaining comparability with those already tested.