Research On Beam Allocation Algorithms In Millimeter-wave Massive MIMO Systems

With the rapid growth of multimedia services and intelligent terminals,mobile communication systems have been under pressure from several aspects such as system capacity,data rate and spectrum resources.Millimeter-wave communication provides a larger available spectrum,and it greatly alleviates the spectral resource pressure in the current mobile communication system.The Massive MIMO technology configures a large-scale antenna array at the base station so that the transmitting signal acquires large beamforming gain which can greatly compensate for the large path loss of the millimeter-wave.In addition,the short wavelength of millimeter-wave allows to equip numbers of antennas in a limited space.Therefore,the millimeter-wave Massive MIMO system is an inevitable technical solution of 5G.This work is supported by the research project "Research on Key Technology and Business Analysis Technology of New Generation Networks" cooperated with an enterprise.This paper focuses on the research of the beam allocation technology in millimeter-wave Massive MIMO systems,and proposes two beam allocation algorithms which are applicable to the initial access procedure and the data transmission procedure respectively.First of all,this paper describes the basic concepts and key technologies of millimeter-wave and Massive MIMO respectively,and analyzes the pros and cons of the current beam allocation algorithms in the millimeter-wave Massive MIMO systems.Then,in view of the large access delay brought by the beam sweeping schemes and the low accuracy brought by the current beam allocation algorithms in initial access procedure,a beam allocation algorithm with low access delay is proposed in this paper.It models the beam allocation problem in initial access procedure as a support vector machine multi-classification problem by using the location information of user equipment as features and the transmitting beams of millimeter-wave base station as classes.The overall procedure of the proposed algorithm can be divided into system initialization procedure and beam allocation procedure,corresponding to training classification model and predicting classification results.Simulation results show that the proposed algorithm can effectively reduce the average access delay of user equipments and has a high beam allocation accuracy.Finally,in view of the problem that current beam allocation algorithms in data transmission procedure try to optimize system capacity partially and neglect fairness among user equipments,a beam allocation algorithm with fair transmission rate is proposed in this paper.The proposed algorithm forms an alternative data transmitting beam set by using the LOS beam and NLOS beams that point at the target user equipment and establishes an optimization problem with an optimization target of maximizing system utility.Eventually,the proposed algorithm utilizes the distributed Q learning algorithm to obtain the optimal beam allocation scheme iteratively.Simulation results show that the proposed algorithm can effectively improve system capacity and the fairness of transmission rate among user equipments.