Heuristic Search Inspired Beam Selection Algorithms for mmWave MU-MIMO System With Discrete Lens Array

Abstract

By employing beam selection in beamspace, the costs of hardware and power consumption in millimeter wave (mmWave) massive multiple input multiple output (MIMO) system can be significantly reduced without obvious performance loss. In most existing schemes, there is a limitation that the number of beams (radio frequency (RF) chains) and users must be equal, which leads to degradation of system performance and flexibility. To overcome this limitation, we formulate beam selection as a discrete combinational optimization problem of binary selection vector to maximize the spectral efficiency of multi-user MIMO (MU-MIMO) system. In order to solve this problem, we propose enhanced CE (ECE)based and GA-based algorithms inspired by cross entropy (CE) method and genetic algorithm (GA) in heuristic search. Moreover, we analyze the complexity of proposed algorithms and provide a graphical representation of convergence property for ECE-based algorithm. Simulation results demonstrate that the proposed algorithms are able to achieve good performance with variable number of beams, and verify the validity of graphical representation of convergence property for ECE-based algorithm.