Task Offloading and Trajectory Scheduling for UAV-Enabled MEC Networks: An Optimal Transport Theory Perspective

In this letter, we propose a task offloading framework for UAV-enabled MEC networks, where all smart mobile devices are scattered randomly following an arbitrary distribution. By exploiting the hovering and mobilizing capability of a UAV, task offloading service can be provided to the devices in all sub-areas of the network. To extend UAV operating time and associated network lifetime, we formulate an optimization problem to minimize the total energy consumption of the UAV through joint region partitioning and UAV trajectory scheduling. To solve it, we decompose it into two independent sub-problems, i.e., region partitioning and UAV trajectory optimization. For the first sub-problem, we model it as a semi-discrete optimal transport problem by considering the traffic balance among different sub-areas and propose an iterative algorithm to achieve the optimal solution. Then, the UAV trajectory optimization sub-problem is modeled as a traveling salesman problem to determine the shortest route. Simulation results show that the proposed scheme can significantly reduce the energy consumption of the UAV while achieving proper load balance.

I. Introduction

Advancements in 5G technologies are enabling many new emerging applications such as virtual/augmented reality (VR/AR) and immerse media that typically require intensive computation, which, however, restricts their use on smart mobile devices (SMDs) due to the limited computing capability and battery capacity. To deal with this dilemma, mobile edge computing (MEC) has been deemed as a promising solution by offloading tasks from mobile edge devices to static or mobile edge servers. Mobile edge servers have been widely adopted in a lot of scenarios, such as industrial Internet of Thing networks. For those scenarios where there are no terrestrial networks or there are but destroyed by natural disaster, unmanned aerial vehicle (UAV)-enabled MEC networks are indispensable. Integrating UAV into MEC networks not only has flexible deployment, but also can bring similar computation and communication performance with terrestrial MEC networks. Recently, UAV-enabled MEC networks have received significant attention from both academia and industry [1]–[3]. In [1], a task offloading and trajectory optimization approach was proposed to minimize the maximum energy consumption among all mobile terminals for fairness. In [2], a cooperative computation offloading scheme among UAVs was proposed to maximize the long-term utility of the whole network. In [3], a three-dimensional deployment of UAVs was investigated to minimize the total time for the UAVs to complete the offloaded tasks. However, all the aforementioned works haven’t considered the UAV’s energy consumption which is a very important design metric due to the limited onboard energy of UAV.