Research On Hybrid Beamforming And Channel Estimation Technologies For Millimeter Wave Massive MIMO

By leveraging abundant spectrum resources and combining the advantages of antenna array gain and spatial diversity,millimeter wave(mm Wave)massive multiple-input multipleoutput(MIMO)technology can provide guarantee for ultra-high transmission rate and ultralow service delay of mobile Internet,which has become one of the key technologies across the evolution of 5G and 6G mobile communication eras.At present,mm Wave massive MIMO still faces numerous urgent key problems,mainly including: The employment of a large quantity of antennas and radio frequency chains increases the complexity and power consumption of system hardware;The expansion of wireless channel dimension increases the resource overhead and storage burden of channel estimation;The increase of path loss and occlusion probability reduces the coverage and quality of service of mobile networks.In order to realize the future vision of full spectrum and full coverage,and to construct a high-speed and intelligent wireless communication environment,this dissertation conducts in-depth research from the aspects of hybrid beamforming design,channel state information acquisition and cross fusion with new physical dimensions for mm Wave massive MIMO systems.Firstly,the design of hybrid MIMO transceiver based on the uniform channel decomposition(UCD)algorithm is studied,and a two-stage optimization scheme of hybrid beamforming is proposed.The objective of maximizing the geometric mean value of the baseband equivalent channel matrix singular values is established,for which the phase extraction and iterative update methods are applied to design the hardware-constrained analog beamformers;The objective of equalizing the subchannel signal-to-interference-plus-noise ratios is established,for which the UCD algorithm based on the minimum mean square error criterion is applied to design the unconstrained digital beamformers;By combining nonlinear processing techniques,i.e.,vertical-Bell Labs layered space-time and dirty paper coding,the interference between non-orthogonal subchannels is eliminated,and the asymptotic performance analysis of the equivalent subchannel gain and system diversity gain is also presented.Simulation results indicate that the proposed scheme can effectively reduce the demand for radio frequency chains,and efficiently reduce the transmission bit error rate with less capacity loss.Secondly,the channel estimation of mm Wave massive MIMO orthogonal frequency division multiplexing(OFDM)systems affected by spatial-frequency dual-wideband effects is investigated,and a channel reconstruction and parameter recovery scheme based on tensor decomposition is proposed.Based on the sparse characteristics of mm Wave propagation,a spatial-frequency wideband channel model is established,which is characterized by the multipath parameters.For scenarios with weak wideband effects,the single-frame multicarrier training signals are modeled as a low-rank trilinear canonical polyadic decomposition(CPD)tensor,where the factor matrix stores the phase shift information in frequency domain.By combining the spatial smoothing method,a structured CPD-based channel estimation algorithm is proposed;For scenarios with significant wideband effects,the multi-frame single-carrier training signals are modeled as a third-order CPD tensor,where the factor matrix contains the fading information in time domain.By combining specific beamforming methods,a structured CPD-based channel estimation oriented towards the dual-wideband effects is proposed;The uniqueness condition analysis for tensor factorization is presented to optimize the beam training design and time-frequency resource allocation.Simulation results indicate that the proposed scheme has strong robustness,and can achieve better channel estimation performance than traditional tensor decomposition schemes with lower computational complexity.Thirdly,the channel estimation of reconfigurable intelligent surface(RIS)-assisted mm Wave massive MIMO-OFDM systems is researched,and a hybrid RIS architecture,as well as,a channel reconstruction and parameter recovery scheme based on tensor completion are proposed.Passive reflectors and active antennas are employed to construct a hybrid RIS layout with the environment sensing ability,while the cascaded channel estimation task can be decomposed into two subchannel reconstruction problems;In the single-carrier training mode,the partially observed signal on the RIS is modeled as an incomplete third-order CPD tensor,and a fiber-sampling tensor completion problem is constructed.Based on the tensorial space transform,an algebraic channel estimation algorithm is proposed;In the multi-carrier training mode,the partially observed signal on the RIS is modeled as an incomplete fourthorder CPD tensor,and a slice-sampling tensor completion problem is constructed.Based on the channel structural characteristics and the concept of virtual array,two algebraic channel estimation algorithms are proposed respectively;The uniqueness condition analysis of tensor completion is presented to optimize the hybrid RIS layout.Simulation results indicate that the hybrid RIS architecture can eliminate the inherent ambiguities from passive RISs,and the proposed scheme can achieve better channel estimation performance than traditional tensor completion schemes with extremely low computational complexity even under the condition of high data missing ratio.After that,the channel estimation and user localization of RIS-aided mm Wave MIMOOFDM systems are investigated,and a twin-RIS architecture,as well as,a channel reconstruction and parameter recovery scheme based on pattern training,tensor decomposition and nonlinear optimization are proposed.The twin RIS layout is constructed by arranging two RIS planes with spatial rotation angles,which can transfer the threedimensional spatial information;The cascaded channel training signals are modeled as a third-order tensor with the number of components equaling to the quantity of RIS reflectors;Four training pattern designs of RIS reflection coefficients,i.e.,random pattern,structured pattern,grouping pattern and sparse pattern,are proposed,while the extraction of cascaded path parameters for two-hop channels is efficiently implemented by combining the tensor operation,array processing and atomic norm denoising.By leveraging the physical structure and electromagnetic properties of the twin RIS,magnitude and phase constraint equations with respect to the channel parameters are formulated,and two series/parallel parameter decoupling modes based on classical nonlinear solvers are proposed;The uniqueness condition analysis of channel estimation with different training patterns is presented,and a user positioning application is implemented based on the recovery results of channel parameters.Simulation results indicate that the twin RIS architecture can eliminate the inherent ambiguities from the single RIS device,and the proposed scheme can achieve accurate channel estimation and user localization performance.Finally,the channel estimation and environment mapping of RIS-empowered mm Wave MIMO-OFDM systems are studied,and a three-dimensional conformal RIS architecture,as well as,a channel reconstruction and parameter recovery scheme based on compressed sensing,nonlinear optimization and spatial spectrum estimation are proposed.The conformal RIS topology is constructed by regularly deploying reflection unit cells with different positions and orientations on curved substrates or carriers,which can transfer the threedimensional spatial information;The training signal is modeled as a third-order tensor,and the cascaded channel estimation is equivalent to a tensor factorization problem;In condition of sufficient training measurements,the three-dimensional cascaded path parameters are extracted based on the least squares criterion,and the trust region algorithms are employed to implement the parameter decoupling without ambiguities;In condition of limited pilot overhead,a single-carrier channel estimation algorithm based on intelligent algorithms is proposed by exploiting the RIS radiation pattern characteristics,and two multi-carrier channel estimation algorithms based on nonlinear optimization and kernel subspace projection are proposed by exploiting the RIS frequency selective properties;The analysis of coverage range and shadow effect of conformal RISs is presented,and an environment mapping application is implemented based on the recovery results of channel parameters.Simulation results indicate that the conformal RIS architecture can eliminate the inherent ambiguities of two-dimensional RISs,and the proposed scheme can achieve accurate parameter decoupling and environment mapping performance.