Multiple Beam Selection and Near-Optimal Digital Precoding for Multiuser Millimeter-Wave Massive MIMO Systems

Abstract

Beamspace massive multiple-input multiple-output (MIMO) can reduce the number of required radio-frequency (RF) chains without obvious performance loss. However, the joint power leakage and multiuser interferences (MI) problem in beamspace channels degrades the system performance seriously, which is not considered in most recent works. To solve this problem, we propose a hybrid analog/digital precoding scheme with multiple beam selection. First, all the energy focusing beams are selected unshared to collect the channel power and alleviate the MI. Then, the analog precoding is determined by maximizing the array gains for each user. The digital precoding is designed by maximizing the system sum rate under the power constraint. However, the logarithms and fractions in the objective function make the optimization problem nonconvex. Hence, we use the Lagrangian transform to eliminate the logarithms, as well as fractional programming to remove the fractions. Finally, the digital precoding matrix is obtained via convex optimization. Simulation results verify that the proposed scheme can obtain the considerable sum rate performance, while achieving a higher energy efficiency than those of existing schemes.