Research On Transmission Rate Analysis And Channel Estimation Performance Of IRS-Assisted Wireless Communication System

Using intelligent reflective surface(IRS)is a promising solution for improving the coverage and speed of future wireless communication systems.The IRS consists of a large number of near passive elements that interact with incoming signals to reflect them in an intelligent way that improves the performance of wireless systems.Previous work has mainly focused on the design of IRS reflection matrices assuming full channel knowledge.However,estimating these channels in IRS is a challenging issue.Due to the large number of IRS elements,channel estimation or reflected beam training will be related to the following factors:1)If all IRS elements are passive(not connected to the baseband),the training overhead is significant;2)If all elements are connected to the baseband through an all digital or hybrid analog/digital architecture,the hardware complexity and power consumption are too large.This article will use deep learning tools to propose effective solutions to these issues.Firstly,a novel IRS architecture based on sparse channel sensors is utilized.In this architecture,with the exception of a few active(connected to baseband)elements,all IRS elements are passive.However,in performance analysis,the influence of the geometric distribution of active elements on the results is ignored.Therefore,this paper discusses three geometric distributions of active elements on the IRS,namely,random distribution,uniform distribution in the same line,and eight queens distribution,and analyzes their impact on the achievable rate of IRS assisted wireless communication.Specifically,the optimal reachable rate is obtained from a predefined codebook determined by a reflected beamforming codeword related to the geometric distribution of active elements in the IRS.The simulation results show that different geometric distributions of active elements have different effects on the reachability rate.The eight queens distribution proposed in this paper has the highest reachable rate compared to random distribution and uniform distribution.On the surface of passive IRS,the distribution of a small number of active elements is limited by the Eight Queens method,and the achievable rate of wireless communication systems is further improved.This method can increase the achievable rate by7% compared to traditional methods.Integrating IRS into millimeter wave(mm Wave)massive multi input multi output(MIMO)communication is a promising method for improving coverage and throughput.Most existing work assumes ideal channel estimation because cascaded MIMO channels are high-dimensional and have a large number of passive reflective elements,which poses some challenges for channel estimation.Therefore,this paper proposes an improved deep denoising neural network assisted compression channel estimation mm Wave IRS system to reduce training overhead.Specifically,a hybrid active/passive IRS architecture is used to estimate the uplink user to IRS channel using a small number of receiver chains.During the channel training phase,a few elements are activated to sense the channel.In addition,the complete channel matrix is reconstructed based on compressed sensing limited measurements,and the common sparsity of millimeter wave MIMO channels between different subcarriers is utilized to improve accuracy.Based on this,this paper proposes a complex-valued denoising convolution neural network based on wavelet transform(Wt Cv-Dn CNN).The simulation results show that the proposed solution is superior to the existing solutions.