A Reliable Learning Based Task Offloading Framework for Vehicular Edge Computing

Abstract

Vehicular fog computing is an evolving solution for the delay sensitive computations at the vehicular edge. Due to the rapidly changing environment, effective resource utilisation becomes quite challenging. Centralised solution are proposed to improve the resource utilisation efficiency but with the added cost of central management and lower efficiency of the resource sharing environment. Distributed task offloading solutions are presented to address the issue; however, it results in an uneven workload distribution without considering the reliability of the communication between the nodes. In this work, we propose a fully distributed task offloading framework that minimises the residence time of the system under the task failure constraints. This overall improves the straggler effect by guaranteeing the task offloading delay at the vehicular edge by replicating the tasks at different vehicular destinations. The proposed work only keeps the tasks with the fastest response time and tasks with the slower response times are removed from the execution queues improving the task resource utilisation efficiency of the resource sharing environment.