Research On Millimeter Wave Communication Precoding Technoligy Base On Large-scale Antenna Array Hybrid Beamforming

With the popularization of wireless intelligent terminal equipment,demand of users for mobile traffic has increased dramatically,and traditional band spectrum resources have become increasingly scarce.Millimeter-wave bands rely on rich spectrum resources and have received widespread attention.However,the millimeter-wave signal path loss is very large,and it is very susceptible to shadows,which can cause interruptions.Large-scale antenna arrays and proper beamforming can increase antenna gain to compensate for path loss,but digital baseband beamforming requires a complete dedicated radio frequency link for each antenna unit in the transmitter and receiver,which greatly increases equipment costs and power consumption.While hybrid beamforming technology achieves similar performance to pure digital beamforming technology,it reduces device costs by substantially reducing the number of digital transceivers,but the number of transmitting power amplifiers still increases with the total number of transmitting antennas.Therefore,in this architecture,the power consumption of the power amplifier device is a key issue that restricts the energy efficiency of the entire system.In order to improve the energy conversion efficiency of the power amplifier device,the Peak-to-Average Power Ratio(PAPR)must be reduced to make the power amplifier work stably near the saturation region.This article mainly studies the signal precoding scheme to reduce the signal peak-to-average ratio to improve the energy efficiency of the transmitting power amplifier.The main work is as follows:This thesis introduces the peak-to-average ratio compression precoding technology of millimeter-wave signals based on mathematical optimization models,including the zero-forcing peak-to-average ratio compression precoding scheme as a baseline model,and the performance-optimized Fast-Iterative Shrink Threshold Algorithm(FISTA)and an iterative optimization scheme constructed by the Squash method for solving the unconstrained infinite norm regular problem.The trade-off effect of regularization coefficient on peak-to-average ratio performance and bit error rate performance was verified through simulation for the iterative optimization scheme.For the millimeter wave system with large-scale antenna array combined with hybrid beamforming architecture,a peak-to-average compression precoding scheme based on asymmetric auto-encoder is designed.Among the scheme,according to the QPSK modulation used in the transmission signal,a complex fully connected layer(CFC)is given,and the complex fully connected layer,the layer normalization scheme and the Swish activation function based on the large-scale search scheme are applied to the auto-encoder structure.The training process of the auto-encoder adopts a multi-task learning scheme.According to the requirements of two indicators,that is,the cumulative complementary error function(CCDF)that measures the compression ratio of the peak-to-average ratio and the bit error rate that measures the degree of signal distortion,a combined loss function for multi-task learning is designed,and the first-stage and second-stage programs are compared when training the model.Finally,zero-forcing and iterative optimization schemes based on mathematical models are simulated,as well as the auto-encoder scheme proposed in this thesis.Simulation shows that the scheme based on auto-encoder is significantly better than the previous schemes in the peak-to-average ratio performance and bit error rate performance.Besides,a brief analysis of the complexity of the inference of different schemes is made.The analysis shows that the time complexity of the scheme based on the auto-encoder scheme is greatly reduced when it is applied.