Theory And Key Technologies Of Multiple-antenna Advanced Receiver Using Message Passing

The fifth-generation mobile communication systems(5G)elementarily achieve new radio enhancement and spectrum expansion by exploiting massive multiple-input-multiple-output(MIMO)and millimeter wave communications.With the large-scale deployment of 5G commercial system,it is realized that the current system design has problems of prohibitive hardware and power costs.Moreover,the fast-growing mobile services and applications demand further improvements on spectral efficiency,network coverage,localization and environment sensing,and so on.Many of enabling technologies has come into view,in which the multiple-antenna advanced receiver attracts considerable attentions.Recently,message passing method is introduced into the researches on detection and estimation due to its generality,high performance,and low complexity,and the corresponding researches is just started.On the basis of message passing method,this thesis investigates the theory and key technologies of multiple-antenna advanced receiver.Firstly,for low-precision massive MIMO relaying systems,the expectation consistent with multiple measurement vector(EC-MMV)algorithm is proposed to solve the distortion problem of quantized signal.Using the posterior expectation estimators,EC-MMV extracts soft information of transmitted symbols from the quantized signal,and realizes joint detection through the maximum ratio combining of soft information.In order to improve the convergence,the partially parallel scheduling is used to update the soft information of direct and relay links.According to the self-average property,the state evolution(SE)equations are derived to characterize the symbol error rate of EC-MMV.On the basis of SE equations,the dynamic analysis is studied to measure the performance loss caused by low-precision quantization and the diversity gain provided by relay link.Simulation results show that EC-MMV has Bayes-optimal performance,and the quantized systems realize the tradeoff between performance and cost with 2-to 4-bit analog-to-digital converter(ADC)with EC-MMV.Secondly,for reconfigurable intelligent surface(RIS)assisted millimeter wave systems,MIMO transmission scheme is used to exploit multiplexing gain,and the generalized expectation consistent with coding(GEC-C)algorithm is proposed to realize MIMO detection by adapting channel characteristics.By deploying multiple RISs in different locations,a synthetic channel with high spatial diversity is formed between user and base station.However,the synthetic channel is generally not a full-rank matrix.Similar to EC-MMV algorithm,GEC-C extracts soft information of modulated symbols by using the posterior expectation estimators.In addition,GEC-C uses soft channel decoder to efficiently demodulate and recover transmitted bit-stream and solve the problem of posterior expectation estimators in low-rank channels.Leveraging by the tricks of statistical physics,the SE equations with ascent and descent processes are derived for GEC-C.The impacts of RIS locations,user locations,and low-cost hardware on bit error rate of GEC-C are extensively studied through SE analysis.Simulation results show that the low-precision ADC and low-cost RIS only cause moderate performance loss under the support of GEC-C.Thirdly,for millimeter wave communications,the expectation propagation simultaneous localization and mapping(EP-SLAM)algorithm framework is proposed to realize localization and environment sensing under the complex multipath propagation environments.Based on the topology conditions and the soft propagation parameters,the soft positions are approximately computed in the Gaussian distribution family.According to the characteristics of Gaussian distributions,data fusion of soft positions is achieved through Gaussian reproduction formula,and the cost function is designed to recognize multipath and realize data association between soft positions.By exhausting all the possible combinations between multiple topology conditions and soft propagation parameters,standard EP-SLAM algorithm tracks user position and senses surrounding environments in continuous time sequence.On the basis of standard algorithm,simplified algorithm significantly reduces complexity by computing and reserving a small number of combinations with lower costs.Simulation results show that EP-SLAM realizes good localization performance with low complexity.Finally,for millimeter wave communications,the soft channel estimation and localization algorithms are proposed to extract channel state information(CSI)and user position information,and realize integrated communication and localization.For multipath channel,the Newtonized variational inference spectral estimation(NVISE)algorithm is proposed,where the multivariate Gaussian distribution is used to extract soft propagation parameters from the received signal and reconstruct CSI.For NVISE,Newton methods are used to reduce the distance between the posterior probability density function(pdf)of received signal and the surrogate pdf of soft propagation parameters.According to the soft propagation parameters extracted by NVISE,the soft user position and clock bias estimates are obtained through topology condition.On the basis of EP-SLAM algorithm framework,the soft localization algorithms are presented to recognize direct paths from multipath,fuse the information of direct paths,and simultaneously estimate user position and clock bias.For integrated communication and localization,the user position estimate that obtained from soft localization is fed back to NVISE to improve the estimates of soft propagation parameters and CSI.For propagation parameters and user position,simulation results show that the estimation performance of proposed algorithms approach the corresponding theoretical lower bounds with low complexity.