Energy-Efficient Zero-Forcing Precoding Design for Small-Cell Networks

• Published in: IEEE transactions on communications 2016-02, Vol.64 (2), p.790-804
• Main Authors: Vu, Quang-Doanh, Tran, Le-Nam, Farrell, Ronan, Hong, Een-Kee
• Format: Article
• Language: English
• Subjects: ADMM; Algorithm design and analysis; Algorithms; Antennas; branch-and-bound; Computing time; energy efficiency; Energy management; joint design; Mathematical model; Mathematical models; MIMO; Networks; Optimization; Power demand; Quality of service; Quality of service architectures; Small-cell networks; Transformations; Wireless communication
• ISSN: 0090-6778; 1558-0857
• DOI: 10.1109/TCOMM.2015.2502941

We consider small-cell networks with multiple-antenna transceivers and base stations (BSs) cooperating to jointly design linear precoders to maximize the network energy efficiency, subject to a sum power and per-antenna power constraints at individual BSs, as well as user-specific quality of service (QoS) requirements. Assuming zero-forcing precoding, we formulate the problem of interest as a concave-convex fractional program to which we proposed a centralized optimal solution based on the prevailing Dinkelbach algorithm. To facilitate distributed implementations, we transform the design problem into an equivalent convex program using Charnes-Cooper's transformation. Then, based on the framework of alternative direction method of multipliers (ADMM), we develop a decentralized algorithm, which is numerically shown to achieve fast convergence. Since BSs are generally power-hungry, it may be more energy-efficient if some BSs can be shut down, while still satisfying the QoS constraints. Toward this end, we investigate the problem of joint precoder design and BS selection, which is a mixed Boolean nonlinear program, and then provide an optimal solution by customizing the branch-and-bound method. For real-time applications, we propose a greedy algorithm which achieves near-optimal performance in polynomial time. Numerical results are provided to demonstrate the effectiveness of the proposed algorithms.