Channel Estimation For Millimeter Wave Massive Mimo

In this thesis,traditional channel estimation approaches and deep learning channel estimation approaches are studied for millimeter wave(mm Wave)massive multiple-input and multiple-output(MIMO)systems,where traditional approaches include over sampling approach and estimating signal parameters via the rotational invariance techniques(ESPRIT)approach.First,a framework of over-sampled channel estimation in mm Wave massive MIMO system is proposed.The framework includes the design of hybrid precoding and combining matrix as well as the search method for the largest entry of over-sampled beamspace receiving matrix.Then based on the framework,three channel estimation schemes including identity matrix approximation-based scheme,scattered zero off-diagonal-based scheme and concentrated zero off-diagonal-based scheme are proposed.Simulation results show that the proposed schemes can approach the performance of the ideal case.Second,two channel estimation schemes based on the ESPRIT method for mm Wave massive MIMO systems are proposed.One scheme is based on two-dimensional ESPRIT,which includes three stages of pilot transmission.This scheme first estimates the angles of arrival(Ao A)and angles of departure(Ao D)and then pairs the Ao A and Ao D.To reduce the pilot transmission from three stages to two stages,the other scheme based on one-dimensional ESPRIT and minimum searching is proposed.It first estimates the Ao D of each channel path and then searches the minimum from the identified mainlobe.To improve the robustness of channel estimation,we also develop a hybrid precoding and combining matrices design method.Deep learning channel estimation approaches are also studied in addition to traditional channel estimation approaches.A deep learning compressed sensing(DLCS)channel estimation scheme is proposed.The neural network for the DLCS scheme is trained offline using simulated environments to predict the beamspace channel amplitude.Then the channel is reconstructed based on the obtained indices of dominant beamspace channel entries.Aside of channel estimation,hybrid precoding is also studied.A deep learning quantized phase(DLQP)hybrid precoder design method is developed after channel estimation.The training neural network for the DLQP method is obtained offline considering the approximate phase quantization.Then the deployment hybrid precoding neural network(DHPNN)is obtained by replacing the approximate phase quantization with ideal phase quantization and the output of the DHPNN is the analog precoding vector.Finally,the analog precoding matrix is obtained by stacking the analog precoding vectors and the digital precoding matrix is calculated by zero-forcing.