Profit optimized task scheduling for vehicular fog computing

Abstract

Vehicular fog computing has emerged as a promising paradigm that provisions computing at the network edge and alleviates the computation workload of static edge computing servers. In this regard, building computing facilities on top of jammed vehicles is particularly attractive and practically viable. However, the respective offloading mechanisms and resource sharing have been less explored. In this work, we propose a novel jammed vehicular cloudlet (JVC) assisted task offloading framework that aggregates and leverages underutilized communication and computation resources of congested vehicles and nearby road side unit to serve resource-intensive tasks of mobile users. To motivate resource provisioning by the JVCs in a non-competitive environment, we design an incentive mechanism that charges offloading user and rewards the serving JVC. With aim to maximize the total profit earned by JVCs, we formulate joint task assignment and resource allocation problem in presence of data segmentation, task deadline, and budget constraints. The formulated problem is mixed integer non-linear programming problem, and we directly obtain its solution using genetic algorithm (GA). We further devise a greedy fractional-knapsack based resource allocation scheme named profit-aware task scheduling (PATS). The extensive evaluation under realistic human mobility trajectories demonstrates that, GA outperforms other baseline schemes in maximizing the total profit of JVCs while PATS achieves comparable performance and serves more users with much lower computation complexity.