Research On Precoding Technology Base On Millimeter Wave Communication Large-scale Antenna Array

In recent years,the number of various wireless terminal devices around the world has exploded,causing the demand for mobile data to increase faster and faster.And the existing communication frequency bands are becoming more and more saturated.The millimeter wave communication with the high frequency band has abundant spectrum resources,which has aroused extensive attention from the academic community.Nevertheless,the high frequency and short wavelength of the millimeter wave cause that the signal has an excessive path loss during transmitting and it is susceptible to shadows.In severe cases,communication may be interrupted.In response to these challenges,researchers in academia have proposed many solutions.Among these solutions,the combination of hybrid beamforming and large-scale antenna arrays millimeter wave communication system model has become the mainstream architecture and continues to be optimized and developed.Large-scale antenna arrays have a great number of transmitting and receiving antennas,and the antenna gain can effectively compensate for the path loss.The hybrid beamforming structure that combines digital beamforming and analog beamforming can reduce the number of radio frequency links.However,in this millimeter wave communication architecture,the number of power amplifiers at the transmitter will still increase with the increase of transmitting antennas.Therefore,reducing the loss of the power amplifier and improving system efficiency are critical.For the purpose of streamlining the energy conversion efficiency of the power amplifier,it is necessary to reduce the signal peak-to-average power ratio(PAPR)of the millimeter wave signal at the transmitter to ensure that the power amplifier works stably near the saturation region.This paper conducts the research on precoding technology base on millimeter wave communication large-scale antenna array to realize the transmitted signal PAPR compression.The main contents of this paper are as follows:Aiming at the millimeter wave communication system with the peak power of the transmitting antenna is limited,this article introduces the traditional mathematical optimization model of millimeter wave signal PAPR compression scheme,namely fast iterative shrinkage-thresholding algorithm(FISTA),the improved Squash method is used in this scheme to solve the near-end optimization problem.On the basis of expounding the design scheme of complex neural network,this paper proposes a PAPR compression schemes of millimeter wave signal based on complex neural network(AE1d)for the scenario where the maximum transmit power of the transmitter antenna is not restricted.An autoencoder based on a one-dimensional convolutional network is designed to compress the PAPR of millimeter-wave signals.And a loss function based on multi-task learning is designed according to the two index parameters of complementary cumulative distribution function and system bit error rate(BER).The simulation shows that the AE1 d scheme can achieve better PAPR performance and BER performance,and illustrates the influence of the weight in the loss function on the PAPR performance of the millimeter wave signal and the system BER performance.Finally,for the scenario where the peak power of a single antenna at the transmitter is limited,this article improves the AE1 d scheme and proposes the AE2 d scheme.An autoencoder based on a two-dimensional complex convolutional network is designed.The loss function is optimized.A multi-task loss function based on neural network prediction weights is constructed to realize the PAPR compression of millimeter wave signals.Meanwhile,the signal encoded by the encoder is clipped to meet the maximum power constraint of the transmitting antenna.The final simulation results show that,compared to the FISTA algorithm,the PAPR of the millimeter wave signal of the AE2 d scheme is significantly reduced,and the system BER is relatively low.What's more,the complexity is also reduced.