Centralized Radio Resource Management for Metropolitan Area NetworksPublic Deposited
In the last decade, global mobile data traffic increased by more than a hundred fold times while maintaining essentially the same monthly charge to the average mobile user. Cisco predicts that overall mobile data traffic will continue to grow rapidly at a compound annual growth rate of 60 percent between 2017 and year 2021. Anticipating future demands, the wireless industry set an ambitious goal to increase the capacity per unit area by three orders of magnitude through the deployment of next generation technologies, often referred to as 5G or IMT-2020. ', 'The capacity gain will be achieved mainly using three means: 1) increased spectral efficiency, primarily through physical-layer improvements and the use of efficient resource management; 2) extreme network densification to improve the area spectral efficiency; and 3) increased bandwidth, primarily by exploiting the millimeter wave band. There has been broad consensus that next-generation networks are going to be heterogeneous with dense deployment of small cells under the umbrella of macro cells. ', 'The objective of this thesis is to formulate the general centralized resource management problems as a class of optimization problems and to provide computationally tractable resource management methods. In this thesis, the general resource management problems are addressed in three aspects: 1) spectrum allocation and user association over multiple radio access technologies. 2) scalable large-scale resource management with guaranteed near-optimal performance. 3) joint spectrum allocation, user association, and power control for large networks.', 'First, we study spectrum allocation in downlink heterogeneous networks (HetNets) with multiple radio access technologies (RATs) over different bands using the average packet sojourn time as the performance metric. In addition to the licensed band, a queueing model with vacation has been proposed to model the additional delay associated with the unlicensed band. Two optimization-based schemes have been proposed and shown to be highly effective through simulation results.', 'The thesis then focuses on the design of scalable centralized resource allocation algorithms for large-scale networks consisting many hundred access points (APs) and user devices. Instead of solving a convex optimization problem with an exponential number of variables in the network size, a scalable reformulation is obtained by exploiting the geometric graph nature of the network and provable sparsity of the optimal solution. A pattern pursuit algorithm with low complexity is proposed to solve the reformulated problem with guaranteed gap to the global optimum. ', 'Finally, the joint spectrum allocation, user association, and power control problem for large-scale networks is studied. We develop a scalable reformulation by exploiting the hidden sparse structure of the optimal solution. An efficient algorithm is proposed to solve the reformulated problem with guaranteed convergence. Moreover, each iteration is performed in closed form, which makes centralize resource allocation practically feasible even for a very large network.