On The Methods Of User Activity Detection And Channel Estimation In Non-Cooperative Multiple Access

With the rapid development of the Internet of Things(IoT),it is inadequate to apply current commu-nication protocols to suffice the need of Machine-to-Machine(M2M)communication as its remarkable re-quirements are highly distinct from the features of the Long-Tern Evolution(LTE).To enable a low-latency,high-reliable Machine-Type Communication(MTC)in the next generation wireless network,this thesis in-vestigated several approaches of pilot-assisted joint activity detection(AD)and channel estimation(CE)in a non-cooperative multiple access scenario.Firstly,we investigated MMV based orthogonal matching pursuit(OMP)and several its derivations in AD and CE of the non-cooperative multiple access scenarios.Firstly,based on the existing theoretical results of block OMP(BOMP),we derived the asymptotic behavior of OMP-MMV when the base station(BS)antennas are tend to infinity,which contains the sufficient condition that OMP-MMV can exactly identify all active users and the mean square error(MSE)bound that it can achieve.Secondly,we consider the fact that current literatures in this field mainly assume the perfect knowledge of the number of active users a prior,which is unreasonable in massive sporadic transmission.Thus,we models the number of active users a poisson distributed random variable.Under this circumstance,we proposed an improved OMP-MMV to detect active users and estimate their channel responses,which can be proved to be the Bayes optimal matching pursuit in detecting active users as it fully utilized the a prior knowledge of the distribution of user activity.Finally,we adjust our OMP-MMV and improved OMP-MMV algorithms to some practical scenarios that it may come across in a real wireless communication system.Simulation results of our proposed algorithms revealed a better performance in AD and CE compared to other existing approaches,which makes our algorithms good candidates in a scalable M2M non-cooperative multiple access system.In the second part of this thesis,we considered how to apply the generalized approximate message pass-ing(GAMP)algorithm into AD and CE in non-cooperative multiple access.Firstly,we use factor graph(FG)to represent the basic mathematical model of AD and CE.Secondly,we divide this graph into two partitions.Then we use GAMP into the first graph to conduct a minimize mean square error(MMSE)estimation of the channel responses.At the same time,message passing algorithm(MPA)is running on the second graph.By iteratively changing the information of these two graphs,we change the traditional GAMP algorithm into GAMP-MMV.GAMP-MMV can fully utilize the joint sparse property of the original signals and estimate them in a Bayesian manner,which is not only powerful but also scalable.Finally,in the case where BS has no knowledge of activity probability,we combine the expectation maximization(EM)algorithm into GAMP-MMV to enable a learning process within each iteration of GAMP.Simulation results reveal its high reliability in supporting a low latency M2M communication.With the fast development of CS,a new Khatri-Rao Product based compressed signal support recovery algorithm,along with its derivations,has been discussed in the sequel.From the insight of the Khatri-Rao Product based algorithm,the support of a signal with more sparsity can be recovered compared to other signal reconstruction algorithms,which is mainly results from the advantages that this algorithm can provide by fully utilizing the self-correlation matrix that is asymptotically approached by multiple measurement vectors(MMV).Due to the strict requirement of MTC in next generation wireless network,AD should be conducted within each time slot.If the active user number are beyond the pilot length,a majority of the signal re-construction algorithms can not guarantee the successfully identification of these active users.Fortunately,the Khatri-Rao product based support recover algorithm can provide an acceptable performance in AD in such overload scenario.Furthermore,simulation results illustrated its advantages in massive multiple-input-multiple-output(MIMO)non-cooperative multiple access system,which underlies the basic for subsequent channel estimation and signal detection.