Skip to main content

Resource Allocation for Mobile Edge Computing Systems

Time: Mon 2020-10-12 13.00

Location: Zoom link for online defence (English)

Subject area: Electrical Engineering

Doctoral student: Ming Zeng , Nätverk och systemteknik

Opponent: Professor Di Yuan, Uppsala Universitet

Supervisor: Viktória Fodor, Mikroelektronik och informationsteknik, IMIT, Nätverk och systemteknik; Gunnar Karlsson, Mikroelektronik och informationsteknik, IMIT


The rapid development of mobile information industry has led to the emergence of various mobile applications in areas such as industrial automation, health care or transportation. These applications often require heavy computations with strict latency requirement, which may surpass the processing abilities of the mobile devices. A promising technology to support such applications is mobile edge computing (MEC), which places edge servers in the vicinity of mobile devices to enable computation offloading. MEC has the potential to substantially augment the computation capacities of the mobile devices. However, computation offloading requires to transfer the input data from the mobile devices to edge servers, resulting in extra transmission latency and energy consumption. Meanwhile, the limited computing power at the edge servers needs to be shared by multiple users, which may in turn lead to non-negligible computing time. Moreover, the effect of the available computing and communication resources is coupled in MEC systems, and thus, a joint allocation of these two resources is of significance. In this thesis we propose centralized and decentralized algorithms for effective resource allocation for various MEC systems that adopt orthogonal multiple access (OMA) or non-orthogonal multiple access (NOMA) for computation offloading.

In the first part of the thesis, we propose to apply NOMA to MEC for facilitating computation offloading. We first study power allocation for a multi-cluster NOMA network. We derive optimal solution for cross-cluster and inner-cluster power allocation to maximize the sum rate, and demonstrate the superiority of NOMA over OMA. On this basis, we investigate the resource allocation of NOMA-assisted single-cell MEC systems. We first consider a wirelessly powered MEC system. We propose a greedy offloading decision combined with optimal resource allocation, and show that while NOMA obtains the same computation rate as OMA, it achieves increased fairness. We then consider a MEC system, which employs multi-carrier NOMA for offloading. We propose low-complexity near-optimal solutions for sub-channel assignment, and optimal joint computing resource and transmission power allocation.

In the second part of the thesis, we study the problem of transmission energy minimization for multi-cell MEC systems, now employing OMA. We first consider the scenario, where each user only offloads to its nearest base station. We transform the network-wide resource allocation into a convex optimization problem, and propose a distributed algorithm that achieves optimal solution. We then investigate the more general scenario, where each user can offload to multiple base stations for parallel processing. We first show that the joint resource allocation problem is non-convex. We study the complexity of optimizing a part of the system parameters, and based on these results propose an iterative algorithm that converges to a local optimum. 

The results in this thesis reveal the importance of advanced transmission techniques for computation offloading and the necessity of joint communication and computing resources allocation in various MEC systems.