Task offloading using GPU-based particle swarm optimization for high-performance vehicular edge computing

Abstract

With mobile edge computing, vehicles can obtain nearby network resources and computability and meet the growing demand for vehicular service at large scales. However, as a result of vehicle mobility and offloading of an extensive number of tasks, the congestion in wireless networks and the insufficient computing resources of edge servers make it difficult to maintain good service quality for users. Moreover, network access point selection is not often considered a factor in task execution latency. In this paper, we propose a smart metaheuristic optimization model to address the problem of low service quality due to vehicle movements and limited edge coverage. Then, the proposed model is used to characterize the overall latency of vehicle task offloading by considering resource utilization, workload at edge servers and vehicle movement characteristics. There are two advantages of our proposed framework. First, the design of the optimization model offers an adaptive task offloading strategy by automatically providing preallocation decisions with respect to the edge server workload state. Second, the proposed metaheuristic approach benefits from recent advances in graphics processing unit (GPU) architectures. In fact, the design of the PSO on a GPU shifts the optimization of the task offloading process to a promising area in terms of time and precision. Extensive experimental results are presented to demonstrate the effectiveness of the proposed framework.