Font Size: a A A

Research On Message Passing Based Channel Estimation And Multiuser Detection In Grant-free NOMA System

Posted on:2021-09-06Degree:DoctorType:Dissertation
Country:ChinaCandidate:Y Y ZhangFull Text:PDF
GTID:1528306620477744Subject:Information and Communication Engineering
Abstract/Summary:PDF Full Text Request
As one of the key technologies of 5G communication networks,Non-orthogonal multiple access(NOMA)realizes the superposition of multi-user on the same resource block,so it can accommodate more users under the same resource,and can realize system overload to meet the needs of massive access in massive machine type communications(mMTC)scenario in 5G communication networks.In addition,the conventional grant-based transmission mechanism requires handshaking procedure to establish wireless transmission link,which will lead to serious reduction of signaling resource utilization efficiency and terminal energy efficiency,and this is unacceptable in mMTC as the communication becomes inefficient due to the small amount of payload data.Therefore,grant-free transmission without handshaking procedure is highly desirable.In grant-free NOMA system,the handshaking procedure is not required to establish wireless transmission link,and users can send signals to the base station randomly in a period of time,so the base station does not know which users are active before data detection.The major tasks for the access point of the system includes user activity detection,channel estimation and data detection of active users.To solve the above problem of signal estimation and detection in grant-free NOMA system,the dissertation studies message passing based Bayesian inference theory.By fully exploiting the statistical characteristics of the system model,iterative receivers based on factor graph and hybrid message passing algorithms are designed to realize joint channel estimation,user activity detection and data detection.The main contents of the dissertation are as follows:1.Considering the problem of low-complexity and efficient algorithm design for joint user activity detection and channel estimation without the knowledge of noise variance in uplink low density signature orthogonal frequency division multiplexing(LDS-OFDM),we formulate the joint user activity detection and channel estimation as a block sparse signal recovery problem by making full use of the sparse distribution of active users and the LDS structure,and propose a low-complexity message passing based block sparse Bayesian learning(MP-BSBL)algorithm,where belief propagation(BP)and mean field(MF)are combined for message passing.In the MP-BSBL algorithm,some auxiliary variable nodes and function nodes are introduce to decompose the observation factor into some sub-factors,which enables the use of both BP and MF to tackle different sub-factors,leading to considerable performance improvement,and some careful approximations are introduced in some message computation to achieve low complexity.Simulation results show that,compared with the block orthogonal matching pursuit(BOMP)and least squaresapproximate message passing-sparse Bayesian learning(LS-AMP-SBL)algorithm,the proposed MP-BSBL algorithm can reach the performance of Genie-aided MMSE algorithm,where the knowledge of user activity are perfect known,and maintain a lower computational complexity.2.Considering the problem of receiver design without the assistance of pilot signal in uplink LDS-OFDM,we formulate the joint channel estimation,user activity detection and data detection of active users without the use of pilots as a novel structured signal estimation problem.The elements of the symbol matrix are discrete valued.The effective channel matrix has a structure due to the LDS sequences for subcarrier allocation in LDS-OFDM.We encode the structure of the effective channel matrix to a probabilistic form,which allows a Bayesian treatment for exploiting the structure of the channel matrix and the development of efficient message passing algorithms.Some auxiliary variable nodes and function nodes are introduce to decompose the observation factor into some sub-factors,which enables the use of both BP and MF to tackle different sub-factors,leading to considerable performance improvement,and some careful approximations based on expecation propagation(EP)are introduced in some message computation to achieve high performance with low complexity.Simulation results show that the active-user-identification error rate(AER),bit error rate(BER)and the convergence of the message passing algorithm based on combine BP-MF are improved compared with the message passing algorithm based on MF only.In addition,with the increase of signal-noise ratio(SNR),the BER performance of the proposed algorithm is close to that of the pilot based receiver,while there is no additional pilot overhead.3.Considering the problem of receiver design without channel estimation for non-coherent detection in uplink MIMO-NOMA,we formulate the user activity and their symbol non-coherent detection as the estimation of a row sparse multiple measurement vector(MMV)model’s support and non-zero element.With the constrain of MIMO and differential modulation,a message passing algorithm based on BPMF is proposed to solve this problem.The proposed algorithm is divided into two steps:active user detection and symbol detection.Firstly,the sparse Bayesian learning method based on message passing is used to detect active users.Then,combined with the differential modulation constraints between adjacent time slots,the message passing algorithm is used to detect the transmitted symbols of active users.Some careful approximations based on moment matching are introduced in some message computation to achieve low complexity.Simulation results demonstrate the effectiveness of the proposed message passing based receiver design for non-coherent detection.
Keywords/Search Tags:Grant-free, non-orthogonal multiple access, factor graph, message passing algorithm, sparse Bayesian Learning, channel estimation, active user detection, non-coherent detection
PDF Full Text Request
Related items