Beamforming And Channel Estimation For Multiuser MIMO With Intelligent Reflecting Surface

The network capacity proliferation and ubiquitous wireless connectivity in the forth-coming wireless network pose great challenges to the existing wireless transmission tech-nologies.Seeking for new techniques with low cost,high spectral and energy efficiency is crucial for 6G networks.Intelligent reflecting surface（IRS）is able to achieve high spectrum-/energy-efficiency via low-cost passive reflecting elements,and thus has been brought forward as a promising solution.By judiciously adjusting its reflecting coeffi-cients with a preprogrammed controller,IRS can change the attenuation and scattering of the incident electromagnetic wave so that it can propagate in the desired way before reaching the intended receiver,which is called as programmable and controllable wireless environment.Despite the appealing advantages and great potential of IRS,there are new challenges that need to be addressed in IRS-aided wireless communications,mainly in-cluding:1)How to jointly optimize the active beamforming and the passive beamforming at the IRS with low complexity;2)How to obtain the perfect CSI with limited training overhead.Firstly,this dissertation studies IRS-aided multicast beamforming for multiuser multi-input single-output（MU-MISO）downlink systems.Depart from the traditional（rank-one）beamforming,the Alamouti space-time block coding（STBC）-aided rank-two beamform-ing is employed at the BS and a joint optimization of the rank-two beamforming at the BS and the passive beamforming at the IRS is formulated to minimize the transmit power with guaranteed signal-to-noise ratio（SNR）at the users.To tackle the nonconvex joint opti-mization problem,this dissertation custom-derives a proximal distance algorithm（PDA）with generalized power method（GPM）to iteratively optimize the rank-two beamforming and the IRS passive beamforming.Both PDA and GPM can be performed very efficiently than the semidefinite relaxation（SDR）-based method.Simulation results demonstrate that compared with the existing SDR-based method,the proposed PDA-GPM algorithm de-creases the runtime by more than 95%while guarantees the transmit power performance loss no more than 2%.Moerover,the rank-two beamforming attains much lower power than the rank-one beamforming.Specifically,when the number of users K=40,the transmit power gap between rank-two and rank-one schemes is about 3 d B.Next,consider that wireless transmission,with the nature of broadcast,are vulnera-ble to be eavesdropped by illegal receivers（eavesdroppers）,of which the CSI may not be perfect,this dissertation studies a joint robust design of secure multicast beamforming at the active transmitter and the passive beamforming at the IRS to guarantee the physical-layer security of the system and realize maximum robustness against eavesdroppers’CSI uncertainty.The CSI error is modeled by a moment-based random error model,in which the BS only knows the first-and second-order statistics of the error,but not the exact dis-tribution.Under this CSI error model,the resulting problem is an intractable semi-infinite chance-constrained optimizaiton problem.By applying recent advances in distribution-ally robust optimization,this dissertation shows that the challenging problem admits an explicit conic reformulation,which is then efficiently solved by utilizing techniques of SDR and the penalty convex-concave procedure.Theoretical analysis also shows that under some conditions,SDR delivers a rank-one optimal solution.Moreover,to further improve the secrecy performance,another rank-two beamformed Alamouti transmission scheme at the BS is studied.Simulation results demonstrate that the proposed designs are robust against the error distribution.Moreover,the 2.5 d B signal-to-noise ratio（SNR）gains can be achieved by the IRS-aided scheme over that without an IRS when the number of IRS reflecting elements is 24.Finally,for the IRS-aided 1-bit massive multi-input multi-output（MIMO）systems,the problems of beamforming design and channel estimation are investigated.For the beamforming design problem,consider a downlink MISO wireless communication sys-tem with 1-bit digital-to-analog converters（DACs）,a joint design of 1-bit symbol-level precoding（SLP）at the BS and the passive beamforming at the IRS is proposed with a goal of minimizing the worst-case symbol error probability（SEP）of the users under the PSK modulation.By analysing the special architecture of the problem,an alternating op-timization framework based on the dual optimizaiton and accelerated projected gradient method is custom-derived to acheive the joint beamforming design.Simulation results show that the gap between the 1-bit SLP designs with and without the aid of IRS is more than 10 d B at the bit-error rate（BER）level 10^-3.For the IRS channel estimation,this dissertation studies the IRS-aided multiuser MIMO uplink channel estimation with 1-bit quantized mearsurements at the BS.By exploiting the sparsity of the mm Wave channels in the angular domain,the IRS channel estimation is formulated as a structured sparse signal recovery problem.By considering different levels of structured sparsity of IRS cascade channels,including elementwise-level,row-level and row-column-level sparsity,three sparse Bayesian learning（SBL）-based channel estimation schemes are proposed,namely the SBL scheme,the block SBL scheme and the two-stage SBL scheme.For each scheme,a variational expectation-maximization algorithm is custom-derived to approxi-mately solve the intractable type-II maximum likelihood estimation problem arising from the one-bit quantization.Also,a complexity-reduced fast implementation of the SBL es-timator is developed.The proposed estimation algorithms consider the special structured sparsity of IRS cascaded channels and thus can achieve good performance with reduced training overhead.Simulation results show that compared with the schemes which do not exploit the special structured sparsity of IRS channels,our proposed schemes can improve the normalized mean square error performance at most 12 d B.