Efficient UAV-Based MEC Using GPU-Based PSO and Voronoi Diagrams

Abstract

Mobile-Edge Computing (MEC) displaces cloud services as closely as possible to the end user. This enables the edge servers to execute the offloaded tasks that are requested by the users, which in turn decreases the energy consumption and the turnaround time delay. However, as a result of a hostile environment or in catastrophic zones with no network, it could be difficult to deploy such edge servers. Unmanned Aerial Vehicles (UAVs) can be employed in such scenarios. The edge servers mounted on these UAVs assist with task offloading. For the majority of IoT applications, the execution times of tasks are often crucial. Therefore, UAVs tend to have a limited energy supply. This study presents an approach to offload IoT user applications based on the usage of Voronoi diagrams to determine task delays and cluster IoT devices dynamically as a first step. Second, the UAV flies over each cluster to perform the offloading process. In addition, we propose a Graphics Processing Unit (GPU)-based parallelization of particle swarm optimization to balance the cluster sizes and identify the shortest path along these clusters while minimizing the UAV flying time and energy consumption. Some evaluation results are given to demonstrate the effectiveness of the presented offloading strategy.