Research On Channel Estimation And Signal Detection In MmWave Massive MIMO Systems Assisted By Intelligent Reflecting Surface

Intelligent reflecting surface(IRS)is a new and promising B5G(6G)wireless communication technology.It is a large two-dimensional array composed of a lot of the passive transmitting elements.It can improve the signal transmission environment and reduce channel interference by changing the phase and amplitude of the signal.Millimeter wave belongs to very high frequency band and the singal hop communication distance is short.IRS can effectively control the propagation path of signals including phase,amplitude,frequency and even polarization,without complex decoding,coding and RF processing operations.The IRS studied in this paper is a passive transmitting element,so this IRS does not have the ability to process signals,which means that cascaded channel estimation needs to be performed according to the transmitted pilots,and a large number of reflection elements in the IRS make the pilot overhead of the signals required for channel estimation a major problem.At present,most channel estimation mehods are applied to the scene of single channel.Without a good combination of IRS assisted MIMO communication system,how to estimate the two hop channels of base station IRS and IRS client is also a difficult problem.The existing signal detection algorithms are generally used in traditional MIMO communication systems.The maximum likelihood detector is the optimal detector that user the minimum joint error probability to detect all symbols.The maximum likelihood detector is usually realized by an efficient search algorithm,but the computational complexity of the maximum likelihood detector is very high.This paper studied channel estimation and signal detection in IRS assisted millimeter wave MIMO communication system.The main research contents include the following two points:(1)Through tensor modeling under the time-domain model and making effective use of the sparse characteristics of millimeter wave channel,an aiternating iterative algorithm for estimating two hop channel is proposed.The algorithm is an iterative method that transfroms the channel estimation problem into sparse signal recovery.The method also involves the determination of regularization parameters.Simulation results show that the parameters have a great impact on the performance of the algorithm.In order to optimize the parameters,the estimation method is improved,and the parameters are adaptive through network training.A fully connected network is designed to improve the performance of channel estimation.In this paper,a fully connected network is designed by expanding the iterative algorithm to improve the performance of channel estimation.Simulation results show that compared with the existing channel estimation methods,the channel estimation method based on network optimization parameters had better performance.The comparison of different channel estimation schemes also shows the performance superiority of this scheme.(2)Moreover,the changes of channel characteristics before and after reflection are not considered.During signal detection,the amount signal between the base station and the user is reflected,and the channel characteristics will also change.For two hop different channels,a specific signal detection scheme is required to improve the detection performance.This paper also studied a signal detection method based on neural network.In massive MIMO systems,in order to avoid the high complexity caused by the inversion of high dimensional channel matrix,neural network has been applied to signal detection.However,signal detection schemes based on neural network can only train a single channel model.According to the research content of(1),it can be found that the two hop channels of base station IRS and IRS client can be estimated,according to the changes of the characteristic of the two hop channels before and after reflection,a novel signal detection gradient descent method according to the different channel characteristic of the two hop channels.The detection scheme combines model driven and data-driven,and can detect the signal more accurately.Finally,through the analysis of simulation results,the proposed signal detection method had better performance than single neural network and traditional signal detection methods.This paper also analyzes the influence of different transmitting antennas and IRS transmitting elements on the performance of this scheme,and carries out a large number of simulation experiments.