Performance Analysis And Intelligent Transmission Design For Massive MIMO Communication Systems

The popularization of mobile terminals leads to the skyrocketing development of the mo-bile Internet and Internet of things.The requests from wireless users take on the new character-istics of high density,delay-sensitive,high mobility,diversity and so on,which drives the revo-lutionary progress to the fifth generation mobile communication systems(5G).Compared to the current fourth generation mobile communication systems,5G can thoroughly meet the diverse demands by dramatically improving the user experience rate,peak rate,connection density,reli-ability,delay performance and so on.As one of the key technologies to realize the performance leap for 5G,massive multiple-input multiple-output(MIMO)can exploit the spatial resource of wireless channels sufficiently by equipping the large-scale antenna arrays,which significantly improves the spectral efficiency and energy efficiency.This dissertation focuses on the pressing issues for massive MIMO systems from the theoretical and practical perspectives,including the analysis and design of low-cost and energy efficient massive MIMO systems,the deployment of massive MIMO relay systems,the coexistence of direct transmission users(DTUs)and relayed transmission users(RTUs)in massive MIMO systems,and the low-overhead and robust design of channel estimation and hybrid processing for massive MIMO systems.The major work and contributions are summarized as follows.· For massive MIMO systems with low-resolution analog-to-digital converters(ADCs),the existing performance analysis assumes independent and identically distributed chan-nels and thus causes the discrepancy compared to the actual performance.Therefore,this dissertation derives the uplink achievable rate for the massive MIMO system with low-resolution ADCs in spatially correlated channels for the first time,based on which it reveals how the achievable rate loss caused by low-resolution ADCs varies with the spatial correlation level.Specifically,the nonlinear ADC quantization is first approxi-mated as a linear operation by using the additive quantization noise model,after which the linear minimum mean-squared error(LMMSE)channel estimation is derived.Then,by considering the channel spatial correlation,quantization error,and imperfect CSI,this dissertation derives the closed-form approximation of the achievable rate with maximal ratio combining(MRC).Finally,further analysis reveals that the achievable rate loss caused by low-resolution ADCs decreases as the spatial correlation level increases re-sorting to majorization theory.Simulation results verify the theoretical analysis and show that low-resolution ADC is more suitable for massive MIMO systems with inherent spa-tial correlation.· To guarantee simultaneous high data rate service for an arbitrary number of remote users in the high energy and cost efficiencies,this dissertation designs and investigates mas-sive MIMO relay systems with low-resolution ADCs in different types of fading chan-nels.First,considering the massive MIMO relay with low-resolution ADCs in the rich scattering environment,this dissertation derives the closed-form approximations of the achievable rate with MRC for both perfect and imperfect CSI,based on which the gener-alized power scaling laws are extracted.The analysis reveals that two to three bits ADCs only cause a limited performance loss.Then,the theoretical analysis is extended to the scenario with line-of-sight path between the base station(BS)and the relay station(RS)based on perfect CSI.Finally,considering massive antennas and low-resolution ADCs at both the BS and the RS in spatially correlated channels,this dissertation derives the LMMSE channel estimate and its tractable equivalent form for ease of subsequent anal-ysis,based on which the closed-form approximation of the achievable rate with MRC is derived.The further analysis demonstrates that the spatial correlation does not impact the asymptotic behavior of the power scaling law and guides us how to deploy massive MIMO relay systems with low-resolution ADCs.Simulation results validate the correct-ness of the theoretical analysis.· To guarantee the coexistence of DTUs and RTUs in the same time-frequency resource,this dissertation proposes a simple and energy-efficient solution from the perspective of power scaling.Specifically,considering the overlap of pilot and data signals from dif-ferent types of devices,this dissertation first establishes the LMMSE channel estimation process for the complicated multi-cell scenario.Then,based on the estimated CSI,this dissertation derives the closed-form lower bound and approximation of the achievable rate for both DTUs and RTUs with MRC and ZF detections.By using the achievable rate approximation,a simple and energy-efficient power scaling law is proposed to en-able DTUs and RTUs to coexist well.Finally,by solving a sequence of geometric pro-gramming problems,an efficient power allocation scheme is proposed to guarantee the maximization of multi-cell capacity and the fairness between the two types of users si-multaneously.Simulation results verify the analytical derivation and the effectiveness of the proposed power control.· As the channel estimation and transceiver algorithms for millimeter wave massive MIMO systems still need to be improved in terms of performance,overhead,and robustness,this dissertation proposes the intelligent signal processing based channel estimation and hy-brid processing schemes.Accordingly,resorting to the unique advantage of deep convo-lutional neural network(CNN)in correlation extraction,this dissertation first proposes the spatial-frequency CNN and spatial-frequency-temporal CNN based millimeter wave channel estimation schemes with high accuracy,short computation time,and good robust-ness to different channel scenarios.Furthermore,by grouping several channel coherence intervals as a channel estimation unit,channel frequency and temporal correlation is ex-ploited to propose another CNN based channel estimation scheme,which saves the pilot overhead significantly at the cost of limited performance loss and slightly increased com-putational complexity.Finally,this dissertation develops a deep neural network based end-to-end optimization framework to realize the joint design for the analog and digital processing at the transceiver.It can be flexibly applied to the narrowband system and wideband orthogonal frequency division multiplexing system with high reliability,short computation time,and good robustness.Simulation results demonstrate the advantages of the proposed intelligent signal processing based schemes in terms of performance,ro-bustness,pilot overhead,and computation time.