Abstract
Unmanned aerial vehicle is a promising technology in upcoming wireless networks due to its potential aerial feature and line-of-sight capability. Mobile edge computing systems that rely on unmanned aerial vehicles consider offloading computationally intensive tasks to unmanned aerial vehicle to be executed via a powerful edge server. In this paper, we exploit the power of device-to-device communications in terms of low latency and low-power transmission as an additional option to edge offloading. Specifically, any smart device that have a nearby device partner can offload part of its task to be executed by his partner. Additionally, we aim to minimize the total energy consumption during the offloading and local computing procedures at smart devices via jointly optimizing the trajectory of the unmanned aerial vehicle, number of bits allocated for both unmanned aerial vehicle and the partner device, and finally task partitioning. We divide the non-convex major problem into two sub-problems which are efficiently solved via alternative optimization. Simulation results reveal the superiority of device-to-device-assisted mobile edge computing systems that rely on unmanned aerial vehicles over counterparts with no device-to-device assistance.