Novel Algorithms For Multiuser Detection In M2M Communications

The future 5th Generation(5G)is a true sense of the integrated network.A key advantage of 5G is reflected in the Internet of Things(IoT).Massive Machine-to-Machine(M2M)communications is an important application of IoT.Noting that the total number of users is very large and the active rate is typically small,we exploit Low-Activity Code Division Multiple Access(LA-CDMA)for M2M communications.In this paper,we focus on solving the Multiuser Detection(MUD)problem supported by the LA-CDMA in M2M communications.Particularly,accurate multiuser detection in M2M communications helps us to address the challenges such as reducing the cost of machine terminals,resource allocation and low-cost M2M terminal coverage.In the literature,a number of multiuser detection algorithms have been proposed,such as the Zero-Forcing(ZF)detector,the Minimum Mean-Square-Error(MMSE)detector,and the Maximum a Posterior(MAP)detector.However,the ZF and the MMESE detectors do not have good detection performance.The MAP detector has been proposed to detect the active users when the user activity factor is known and small.However,such user activity factor is usually unknown and could be large in practice,which makes the MUD in M2M communications a significant and worth studying task.In order to address the difficulty of unknown activity factor and improve the detection ability,we will study the MUD for LA-CDMA uplink using Compressive Sensing(CS)theory and introduce appropriate algorithms,including Bayesian and iterative reweighed algorithms.The novelty and contributions of our thesis are summarized as follows.1)In M2M communications,in order to study the LA-CDMA uplink multiuser detection problem at any time,we introduce Single Measurement Vector(SMV)model.Since some users enter/leave the communication systems at some moment and remain unchanged within continuous time slots,we introduce Multiple Measurement Vector(MMV)model to study the multiuser detection problem for LA-CDMA uplink.Based on MMV model,the transmitted signal has block sparsity feature,which is used to improve signal detection performance.2)We study the multiuser detection problem for LA-CDMA uplink based on SMV model.Noting that the user activity factor is typically small in M2M communications,the transmitted signal has sparsity feature,we introduce Sparse Bayesian Learning(SBL)algorithm to detect the transmitted signal,which is suitable for low activity factor situation.Next,we add on the known finite-alphabet constraints and introduce Gaussian Mixture Model(GMM)method to detect the transmitted signal,which is applicable for any value of activity factor.Additionally,the SBL has very high computational complexity.To reduce the complexity,we develop a novel Fast Inverse-free Sparse Bayesian Learning(FI-SBL)method,which reduces the computational complexity largely.3)We study the multiuser detection problem for LA-CDMA uplink based on MMV model.In the real applications,some users enter/leave the communication systems at some moment and remain unchanged within continuous time slots.Based on MMV model,the transmitted signal has row sparsity feature,we propose Block Sparse Bayesian Learning(BSBL)algorithm and Pattern Coupled Sparse Bayesian Learning(PCSBL)algorithm.Furthermore,to reduce the computational complexity,we embed the Generalized Approximate Message Passing(GAMP)into PCSBL and develop a novel algorithm,called Generalized Approximate Message Passing Pattern Coupled Sparse Bayesian Learning(GAMP-PCSBL).The proposed GAMP-PCSBL method has superior performance both in recovery rate and running time.4)In Chapter 5,we study multiuser detection problem based on iterative reweighed approach.We propose two novel iterative algorithms for signal detection,respectively Iterative Reweighed(IR)algorithm for sparse signal and Minimum Mean-Square-Error Iterative Reweighed(MMSEIR)algorithm for non-sparse signal.On one hand,we replace 0-norm by log-sum function.Then we propose the IR algorithm based on Majorization-Minimization(MM)strategy and transform the non-convex MUD optimization problem into an iterative reweighed problem.On the other hand,we first convert the system into sparse framework,then embed the MMSE into IR,and propose MMSEIR algorithm.5)The proposed algorithms in this paper are applicable to multiuser detection problem with different user activity factor.Simulation results have shown that the proposed methods overcome the difficulties in the MAP algorithm and outperform the classical multiuser detection algorithms.