Research On User Scheduling And Beam Selection Scheme Based On Millimeter Wave Beamforming System

The rapid development of mobile communication and the rapid popularization of smart phones,tablets and other mobile devices have brought unprecedented challenges for wireless service providers to overcome the problem of global bandwidth.This also promotes the development of wireless communication from the fourth generation to the fifth generation(5G).In order to overcome the shortage of bandwidth and meet the expected increasing data rate requirements of 5G system,it is considered a key technology to explore the millimeter wave(mmwave)frequency band(usually considered as the frequency band between 30 GHz and 300GHz)for cellular communication.Nevertheless,in the past few years,through a large number of measurement work carried out by academia and industry all over the world,people have obtained the models and parameters of millimeter wave propagation channels and the characteristics of millimeter wave propagation in various outdoor environments.However,5G channel modeling including millimeter wave system is still in progress,and the performance of millimeter wave system,In particular,the combination with multiple input multiple output(MIMO)technology needs to be comprehensively evaluated.Millimeter wave communication is a promising technology for future spectrum shortage issues.For the cost and energy saving,hybrid beamforming is considered in millimeter wave communication system.In this paper,we propose a novel machine learning-based joint user scheduling and analog beam selection method for millimeter wave communication.We consider user scheduling and beam selection as two multiclass-classification problems where the dataset consists of numbers of samples of millimeter-wave channel.Firstly,based on clustering algorithm,the users are clustered according to channel correlation.Secondly,we consider the analog beam selection problem as multi-classification problem where the dataset consists of samples of the millimeter-wave channel.Using the dataset,we build a statistical classification model via support vector machine(SVM)and then the optimal analog beams are selected for each user,which can obtain the maximum sum rates.Then we propose a gradient-free swarm intelligence method,namely whale optimization algorithm(WOA),to obtain a very general and efficient solution.The analysis and simulation results show that,compared with conventional methods,the ML-based method achieves a sum rate close to greedy searching with low complexity,and the WOA method has a fast convergence and a near global optimal performance.In the scenario that the beam selection and power allocation for Intelligent reflecting surface(IRS)assisted millimeter-wave(mm Wave)nonorthogonal multiple access(NOMA)system was considered.This research aimed to maximize the average sum rate of all users by optimizing the beam selection and power allocation at BS and the phase shifts at the IRS and the successive interference cancellation(SIC)decoding order.Due to the combinatorial nature of the SIC decoding order,the computation complexity of the optimal solution is generally very high.To solve the formulated nonconvex problem,a hybrid whale optimization algorithm(HWOA)and a trick method for decoding order constrain was proposed.Numerical results show that the achievable sum rate can be improved by the IRS with the proposed algorithm,the proposed IRSassisted mm Wave-NOMA systems obtain higher achievable sum rate than the IRSassisted mm Wave orthogonal multiple access(OMA)system.