Active User Detection And Channel Estimation In Non-orthogonal Multiple Access Of MMTC

As one of the important application scenarios of the fifth generation mobile com-munication technology(5G),massive machine type communication(mMTC)has tech-nical standards such as large-scale access,low latency,and high spectrum efficiency.Non-orthogonal multiple access(NOMA)is a multiple access technology that can serve multiple users through a single wireless resource,and can be an important solution for mMTC.NOMA can serve large-scale users with limited resources without the require-ment of resource application in the uplink,which avoids delay caused by signaling trans-mission.However,this also causes the base station side being unable to know which users the data is sent by.So it is necessary to detect the active users,and at the same time,in order to recover the received data,the channel state information needs to be estimated.This thesis mainly studies how to use the system prior statistical information to solve the problem of active user detection(AUD)and channel estimation(CE),and to solve these problems without part of the prior statistical information as prior activity probability.This thesis mainly includes the following two points:(1)When the priori statistical information,that is,user prior activity probability,is known,active user detection and channel estimation problems can be modeled as sparse reconstruction problem by Compressive Sensing(CS)theory.Based on the the-ory of variational approximation interference,variational approximation message pass-ing(VAMP)algorithm based on Expectation Propagation(EP),which uses a Gaussian distribution approximation to express complex target probability distributions,is pro-posed here.The messages transmitted between nodes in the factor graph can be trans-formed by Gaussian distribution parameters.So the computational complexity is re-duced to linear and the computational efficiency is greatly improved.At the same time,in order to ensure the convergence of VAMP algorithm,a damping factor is introduced to smooth the parameter updating in the iterative process.(2)For the active user detection and channel estimation problem without knowing the prior activity probability,a variational approximation message passing algorithm based on Expectation Maximun(EM)is proposed(Damping-EM-VAMP).The priori active probability is dealed as a hidden variable,the maximum likelihood estimation(MLE)of which is performed by the E step of EM algorithm.Then the estimated value is inputed as the prior information of the Damping-VAMP algorithm,where the estimated user channel information is utilized for the MLE.Repeat this way until convergence.The above two researches can reduce the computational complexity and improve the computational efficiency of the active user detection and channel estimation problem in NOMA of the mMTC scenario with higher active user detection accuracy and lower channel estimation mean square error than other algorithms.These researches have solved the problem of detecting the data source users in the uplink of NOMA and solved the channel estimation problem which is required to recover data,so that NOMA can be effectively applied in the mMTC scenes.