Cost Optimization of Partial Computation Offloading and Pricing in Vehicular Networks

For vehicles with limited computation resources offloading their computational tasks to a mobile edge computing (MEC) server has been studied in the past as an effective means for improving their computational capabilities. However, most of these studies do not consider, in a meaningful way, the economic aspects related to both the computation offloading of the vehicles and the MEC service providers. In order to fill this gap, in this paper, a new cost based optimization methodology which jointly considers the cost of partial offloading vs. the pricing of the MEC server is proposed and its performance is analyzed. In particular, we first formally establish the cost model for vehicles and then, by setting a service price, the revenue model for MEC server. Secondly, optimal vehicle offloading strategies are identified and through a cost minimization partial computation offloading algorithm vehicles can configure, in an optimal way, the local CPU frequency and task partition based on the service price. Thirdly, by considering its computation resource limitations, the resource allocation and pricing mechanism for the MEC server is presented. It is shown that, through the development of an appropriate pricing algorithm, the MEC server can obtain the service price which maximizes its revenue while at the same time satisfying the server's resource constraints. Numerical results have verified that the proposed scheme is indeed more cost effective as compared to local execution with dynamic voltage scaling (DVS) technique, full computation offloading and other partial computation offloading schemes. Furthermore, various performance evaluation results obtained by means of computer simulations have shown that the proposed pricing scheme achieves higher revenue as compared to other previously known fixed and random pricing schemes.