Research On Beam Management For 5G Millimeter Wave System

In the current situation of increasingly tight wireless spectrum resources,how to further meet the continuous growth rate demand of 5G wireless communication becomes a key issue for future mobile communication technologies.The introduction of large-scale arrays of millimeter-waves in 5G systems helps meet the growing demand for mobile data,but there are many challenges,such as RF(Radio Frequency)hardware costs,power consumption,and so on.Millimeter wave signal transmission loss is large,and it is affected by rain and atmosphere.The transmission loss of signal can be compensated by the massive MIMO(Multiple Input Multiple Output)beamforming gain.Therefore,hybrid precoding technology and beam management in 5G millimeter-wave massive MIMO systems are the focus of academia and industry research.With the development of big data,the data computing and processing capabilities have been greatly improved,artificial intelligence is also in full swing,and is widely used in various technical fields.This thesis combines artificial intelligence technology,research on hybrid beamforming technology and beam distribution scheme.The main work of this thesis is as follows:1.Aiming at improving the sum rate of millimeter-wave massive MIMO systems by solving optimization problems,and further improve system channel capacity.We combined hybrid precoding with genetic algorithm and proposed a hybrid precoding scheme based on genetic algorithm.comparing with Adaptive Cross Entropy(ACE)algorithm and Coordinate Update Algorithms(CUA)algorithm,The spectrum utilization rate of the proposed algorithm is 2.6 bit/s/Hz higher than that of the ACE algorithm,and 1.5 bit/s/Hz higher than the CUA algorithm.In terms of complexity,the complexity of the proposed algorithm is higher than that of the CUA algorithm,and is equivalent to the complexity of the ACE algorithm.2.Utilizing limited airspace resources is the key to improving the performance of communication systems.First,this thesis transforms the beam assignment problem into a classification problem,and then models the problem with a random forest model in machine learning.Based on the data set collected by Bruce Force Algorithms(BF),a random forest beam allocation scheme using Attribute Combination is proposed by introducing direction distance ratio and identifying beam interference.The simulation results show that the proposed algorithm improves the classification accuracy by 2%~3% compared with the traditional random forest algorithm,and the fluctuation of classification accuracy is small with the increase of random forest size.The average sum rate based on the classification model is better than the Low Complexity Beam Allocation(LBA)algorithm,lower than the brute force matching algorithm.