Liao, Wei-Cheng2016-02-122016-02-122015-11https://hdl.handle.net/11299/177139University of Minnesota Ph.D. dissertation. 2015. Major: Electrical Engineering. Advisor: Zhi-Quan Luo. 1 computer file (PDF); ix, 136 pages.In this dissertation, we consider the base station (BS) and the resource management problems for the cloud-based radio access network (C-RAN). The main difference of the envisioned future 5G network architecture is the adoption of multi-tier BSs to extend the coverage of the existing cellular BSs. Each of the BS is connected to the multi-hop backhaul network with limited bandwidth. For provisioning the network, the cloud centers have been proposed to serve as the control centers. These differences give rise to many practical challenges. The main focus of this dissertation is the distributed strategy across the cloud centers. First, we show that by jointly optimizing the transceivers and determining the active set of BSs, high system resource utilization can be achieved with only a small number of BSs. In particular, we provide efficient distributed algorithms for such joint optimization problem, under the following two common design criteria: i) minimization of the total power consumption at the BSs, and ii) maximization of the system spectrum efficiency. In both cases, we introduce a nonsmooth regularizer to facilitate the activation of the most appropriate BSs, and the algorithms are, respectively, developed with Alternating Direction Method of Multipliers (ADMM) and weighted minimum mean square error (WMMSE) algorithm. In the second part, we further explicitly consider the backhaul limitation issues. We propose an efficient algorithm for joint resource allocation across the wireless links and the flow control over the entire network. The algorithm, which maximizes the utility function of the rates among all the transmitted commodities, is based on a decomposition approach leverages both the ADMM and the WMMSE algorithms. This algorithm is shown to be easily parallelizable within cloud centers and converges globally to a stationary solution. Lastly, since ADMM has been popular for solving large-scale distributed convex optimization, we further consider the issues of the network synchronization across the cloud centers. We propose an ADMM-type implementation that can handle a specific form of asynchronism based on the so-called BSUM-M algorithm, a new variant of ADMM. We show that the proposed algorithm converges to the global optimal solution.enalternating Direction Method of Multipliers (ADMM)asynchronous algorithmbase station activationcloud-RANweighted minimum mean square error (WMMSE)Thesis or Dissertation