Study On Hybrid Precoding For Mm Wave Massive MIMO Systems

Millimeter wave and massive MIMO can significantly improve the spectrum efficiency of mobile communication system,so they have attracted more and more researchers’ attention and they are the core technology of fifth generation mobile communication system.All-digital precoding can effectively utilize the characteristics of millimeter wave and Ma-MIMO.However,considering the construction space and energy consumption of the base station system,the all-digital precoding scheme is not suitable for such a scenario.This paper studies hybrid precoding scheme that is more suitable for millimeter wave MaMIMO systems.That is,antennas share multiple radio frequency chain,respectively design analog precoding and baseband digital precoding,and use the 3D gain of Ma-MIMO array to compensate for millimeter wave path loss and enhance User received signal strength.By analyzing the design process of hybrid precoding in different scenarios,and design hybrid precoding algorithms for single-cell single-user and multi-cell multi-user millimeter-wave Ma-MIMO systems respectively.1)Researching the hybrid precoding of single-cell single-user millimeter-wave MaMIMO system,a hybrid precoding algorithm that equivalents the analog precoder and the analog combiner to the physical channel is proposed.The algorithm aims at maximizing spectrum efficiency and decouples the precoding optimization problem at the transmitting station and the receiving station.Firstly,by alternately searching the codebook,a suitable analog beamforming matrix and analog combination matrix are obtained.Then,the radio frequency chain at the transmitting station and the radio frequency chain at the receiving station are equivalent to a physical channel.To obtain a baseband precoder that eliminates interference between multiple data streams,singular value decomposition is performed on the equivalent channel.Finally,a solution model based on convolutional neural network is designed.The input of this model is the channel state matrix,and the output is the analog combiner weight.Through training and learning the implicit features of the channel state information,the CNN model can predict an analog combiner with the maximum spectral efficiency.In the training process,different synthetic noises are added to each input data to generate different channel state information,and an analog combiner with better performance can still be obtained under the limited hardware and the CSI is affected by noise.2)Researching the hybrid precoding of multi-cell and multi-user millimeter wave Ma-MIMO system,a two-stage hybrid precoding scheme is proposed.In the first stage of the solution,since the signals from other base stations have nothing to do with the connection structure of the antenna and radio chain of the cell,the multi-cell interference is projected to the user precoding null space by coordinating and sharing the user channel state information to eliminate the multi-cell interference;In the second stage,the multiuser hybrid precoding problem is decoupled to the single-user hybrid precoding problem.Firstly,the analog precoder and combiner pair are designed in turn for each user through an iterative algorithm.According to the equivalent physical channel,the digital precoder that make the multi-user interference channel located in the interference null space of the target user is designed.The simulation results verify the effectiveness of the scheme,especially when using a 2-bit low-resolution ADC/DAC,it can also achieve a compromise between energy consumption and performance.