Fine-grained latency analysis of real-world 5G-enabled multi-tier edge computing for school zone traffic safety

Abstract This article presents a real-world evaluation of 5-Safe, an innovative multi-tier 5G edge computing system designed to enhance traffic safety in school zones. The study addresses the critical challenge of implementing reliable and low-latency communication, data processing, and decision-making in complex urban environments. We employ a diverse methodology, including User Datagram Protocol (UDP) transmission analysis, Graphic Processing Unit (GPU) evaluation, and comparative studies of the messaging protocol Message Queueing Telemetry Transport (MQTT) and the Apache Kafka streaming platform. Our research revealed significant insights into system performance, such as latency variability in UDP transmissions, consistent GPU processing efficiency, and the superior performance of Kafka in data transmission. End-to-end system latency measurements highlighted both the potential and limitations of current edge computing technologies in meeting real-time traffic management requirements. We observe that average latencies remain below 58 ms, while maximum spikes reach up to 800 ms, surpassing typical thresholds (e.g.,<100 ms) required for preemptive traffic alerts in pedestrian zones. These findings contribute valuable benchmarks for future research and development in smart city applications. The significance of the study lies in its analysis in real-world implementation, which offers critical insights into the challenges and opportunities of developing responsive, reliable traffic monitoring systems. Our results pave the way for advancements in edge computing architectures and communication protocols, with broad implications for the design of Cyber-Physical Systems (CPS) requiring precise, low-latency data processing in dynamic urban environments.