Simulation-based comparative analysis of offloading algorithms for vehicular edge computing in urban mobility environments

Abstract

The rise of Vehicular Edge Computing (VEC) has been catalysed by the increasing demands on urban mobility infrastructures, where rapid data processing and reduced latency are pivotal. As an extension of edge computing, Autonomous Edge Computing (AEC) particularly highlights the importance of data and computing offloading to enhance the responsiveness and efficiency of autonomous vehicles within these ecosystems. However, despite the proliferation of algorithms designed to optimise data offloading, a challenge remains: these algorithms often lack direct comparability, complicating the determination of the most effective approach under varying urban conditions.

This study compares existing approaches for offloading in VEC, focusing on the offloading efficiency (latency) of electric vehicles in urban environments. Two approaches, including First-In, First-Out (FIFO) and Ant Colony Optimisation (ACO), are investigated. Utilising two established simulators — Simulation of Urban MObility (SUMO) and EdgeSimPy — the comparison is executed across diverse traffic scenarios. The study evaluates the performance of these algorithms under three primary scenarios: Traffic Density, Computational Capability, and Data Generation Rate. The Traffic Density scenario examines how varying numbers of vehicles affect latency, revealing that FIFO can outperform ACO in very low-density conditions due to its simplicity and lower overhead. The Computational Capability scenario assesses the impact of different CPU core counts on latency, showing that while ACO generally excels with limited computational resources, FIFO becomes more competitive as resources increase. Lastly, the Data Generation Rate scenario explores how different rates of data generation influence latency, indicating that FIFO remains viable at lower data rates, whereas ACO is more effective at higher rates.

The analysis provides critical insights into the conditions under which each evaluated approach optimally utilises edge server resources, informing strategic decisions for deploying VEC systems in smart cities. These findings underscore the importance of considering specific environmental and computational contexts when selecting offloading algorithms for VEC systems, highlighting the need for adaptive and context-aware strategies to optimise performance and sustainability.

Date of Award	2024
Original language	English
Awarding Institution	Western Sydney University
Supervisor	Rodrigo Neves Calheiros (Supervisor), Bahman Javadi Jahantigh (Supervisor) & Arindam Pal (Supervisor)