Research On Beam Selection Algorithms For Beamspace Mmwave Massive MIMO System

With the development of the 5th generation communication,the demand of people for high speed and high bandwidth makes millimeter-wave(mm Wave)largescale multi-input and multi-output(MIMO)systems become the focus of attention.As the research of discrete lens array(DLA)becomes more mature,the concept of beamspace is widely used in mm Wave communication systems.According to the sparsity of mm Wave in the beam domain,the beam selection algorithm can be used to select some beam subsets for transmitting,which can reduce the number of radio frequency(RF)links,thus further reducing the hardware complexity and system power consumption of MIMO transceiver.In this paper,the number of users and RF links will not be limited,and we can customize the number of beams.Firstly,inspired by machine learning,an adaptive cross-entropy(ACE)-based beam selection algorithm is designed in this paper.By constantly updating the probability distribution of each beam,a high probability solution is generated to solve the combinational optimization problem.This algorithm not only has low computational complexity and fast convergence speed,but also has obvious advantages in increasing the spectral efficiency(SE)of the system.In addition,considering the similarity with random optimization algorithms,a genetic algorithm(GA)-based beam selection algorithm is designed in this paper.The two algorithms are compared,and the computational complexity and convergence are proved.In addition,inspired by principal component analysis(PCA),we designed a codec-based beam selection algorithm.And we designed a new beam selection algorithm performance criteria,which is the distance between original channel matrix and the reconstruction matrix.We hope to use this criteria together with SE as the standard to measure the beam selection algorithm.Finally,we explored the impact of beam selection algorithms on the performance of mm Wave large-scale MIMO systems in beamspace when minimum mean square error(MMSE)precoding is adopted.Through the dual equivalent uplink,the SE formula of the system is derived.Simulation results show that when MMSE precoding is used,the SE of the system is not sensitive to the beam selection algorithm,and it is close to the optimal full beam transmission system performance.