Joint Task Offloading and Resource Allocation in Multi-User Mobile Edge Computing With Continuous Spectrum Sharing

This paper investigates the joint task offloading and communication/computation resource allocation in a multiuser mobile edge computing (MEC) system. In particular, we consider the multi-channel scenario, in which there are multiple channels each at a different frequency band, and the spectrum of each channel is shared by one or more mobile users (MUs) for offloading continuously. Our objective is to minimize a weighted sum of the energy consumption at MUs, by jointly optimizing the offloading time duration, the offloaded task data bits, the bandwidth allocations of different MUs at these channels, and the computing frequencies at both MUs and edge servers. Because the formulated problem has coupled optimization variables, it is non-convex and difficult to get the optimal solution. To address the challenge, we propose to decompose the formulated problem into two levels. For the lower-level problem with given offloading time duration, we employ the block coordinate descent framework to optimize two groups of optimization variables iteratively, for each of which the optimal solution is obtained in well structures with low complexity. For the upper-level problem for optimizing the offloading time duration, we find the optimal solution by performing variable manipulations and transforming the problem to a standard monotonic optimization problem. Numerical results show that our proposed method is effective in achieving the global optimality of the energy minimization problem and its advantages over other benchmark schemes.

I. Introduction

Thanks to the rapid growth of mobile devices and the surging of artificial intelligence technique in the recent decade, plenty of mobile applications have been developed recently, including virtual reality (VR) and augmented reality (AR). These applications need a significant amount of computation and are also time-sensitive [1]. On the other hand, limited by the computation capability, it is hard for a mobile device to complete the computation task of these mobile applications in time [2]. A possible solution is mobile cloud computing (MCC), which allows the mobile user (MU)¹

Terms “mobile device” and “mobile user” are used interchangeably in this paper.

to offload its computation task to a cloud center. But this may result in long time delay since the cloud center is usually far away from the MU. Alternatively, mobile edge computing (MEC) can overcome the challenge by permitting the MU to offload its computation task to a nearby edge server (e.g., at a base station or an access point) with high computation capability. Thus, the MU's computation task is able to be finished in time at a low energy cost [3].