| In recent years,the increased demand for data throughput has resulted in the rapid development of mobile communication technology.Millimeter-Wave(mmwave)massive Multiple-input Multiple-output(MIMO)technology,with its abundant spectrum resources and extremely high spatial multiplexing gain,has become one of the key supporting technologies for the fifth-generation of mobile communication systems(5G)and even for the future sixthgeneration of mobile communication systems(6G).However,the mmwave band is high,lossy,has limited coverage and is easily obscured by obstacles.Intelligent Reflecting Surfaces(IRS)can improve these shortcomings.The superior performance of IRS-aided mmwave MIMO systems relies on high quality Channel State Information(CSI).However,obtaining accurate CSI is extremely difficult due to the high dimensionality of the channel matrix and the limited guide frequency resources.To address these problems,this thesis investigates the channel estimation problem for IRSaided millimetre-wave massive MIMO systems.The details of the study are as follows:1)Sparse representation of IRS-aided millimeter-wave massive MIMO signal models.Firstly,the received signal and the channel model are established;secondly,for the problem that the angular information such as the angle of arrival is difficult to estimate,the angular domain isometric quantization is performed and the channel model after the angular domain quantization is reconstructed;finally,based on the IRS switching strategy,the sparse representation of the signal model is performed by using the sparsity of the mmwave propagation path.2)On the basis of Research Component 1,a compressed-aware IRS-aided channel estimation scheme for mmwave massive MIMO systems is investigated.Firstly,a three-layer Bayesian compressive sensing model with a Laplace prior is designed to address the problem that the gamma prior is not sparse enough;secondly,an a priori hyperparametric iterative formulation is derived and the accuracy of channel estimation is improved by a variational Bayesian approach;finally,the performance of the proposed scheme is analysed in three aspects,including normalised mean square error,through simulations.The numerical results show that the proposed scheme outperforms the existing algorithms in the literature.3)Based on Research Component 1,a machine learning based channel estimation scheme for IRS-assisted millimeter wave massive MIMO systems is investigated.Firstly,based on the IRS switching strategy and the sparse representation model of the channel,two self-encoder neural networks are designed to estimate the channel of the direct link and the IRS reflective link respectively;secondly,to address the problem that the received signal noise is difficult to be directly transformed into channel sparse vector noise,a regularised least squares method is used to pre-process the data so as to generate the training data set;then,the Adam algorithm is used to train the network model;finally,the performance of the proposed scheme is analysed in four aspects,including normalised mean square error,through simulations,which show that the self-encoder-based channel estimation scheme outperforms existing algorithms in the literature and exhibits a lower computational complexity. |
