Research On Channel Estimation And Signal Detection In Uplink Grant-free NOMA System

With the development of mobile communication technology,massive machine-type communication(MTC)with the characteristics of a large number of device accesses,short data packets,low-rate transmission and sporadic communication has become one of the research hotspots in the fifth generation mobile Communication System(5G).Therefore,the emergence of grant-free non-orthogonal multiple access(NOMA)technology has received widespread attention.Each device can send data without base station scheduling,so as to reduce the control signaling overhead and transmission delay.Since most of the users are in an inactive state at the transmission terminal,and only a few users send data to the base station,the multi-user detection in the uplink grant-free NOMA system is a direction worth studying.How to perform channel estimation and multi-user detection quickly and accurately is the key research issue.Compressive Sensing(CS)technology can effectively realize the reconstruction of sparse signals,which is very suitable for completing the channel estimation and multi-user detection in the uplink grant-free NOMA system.In compressive sensing technology,how to design an efficient CS reconstruction algorithm and how to design an optimized observation matrix are the key factors to improve the effect of sparse signal reconstruction.In this paper,we mainly research on performing the channel estimation and multi-user detection in the uplink grant-free NOMA system based on the compressed sensing model.The main research work and contents are as follows:1)Discuss the channel estimation and multi-user detection technology of the uplink grant-free NOMA system based on one-frame transmission.Under the framework of multiple measurement vector-compressive sensing(MMV-CS),a thresholod aided-distributed weak selection stagewise adaptive matching pursuit(TA-DWSStAMP)algorithm is used to jointly solve the channel estimation(CE)and multiuser detection(MUD)problems.The algorithm terminates at precise iterations under the criterion and introduces a new identification parameter.When the identification parameter represents a large step,a variable step size method based on power function is designed.Simulation results show that,as compared with the existing algorithm,the proposed TA-DWSStAMP algorithm can obtain similar successful activity detection rate of the users,symbol error rate of the user data and the normalized mean squared error performance of the channel estimation with the computational complexity only accounting for about 10% of the existing algorithm.2)Based on the compressed sensing technology,the spread spectrum matrix of CE and MUD model in the grant-free NOMA system is equivalent to the recovery matrix in the compressed sensing theoretical model.Designing the spread spectrum matrix by the principle of designing the recovery matrix in CS theory can improve the signal reconstruction performance.Based on the Mutual Incoherence Property(MIP)criterion,we design a spread spectrum matrix optimization scheme based on random Gaussian matrix.Simulation results show that the channel estimation performance,successful active detection rate of users and the symbol error rate of user data are slightly improved by employing the optimized spread spectrum matrix based on Gaussian random matrix.3)Based on the criterion of mutual incoherence,we also propose a partial Fourier matrix-based spread spectrum matrix optimization method,using a random search algorithm to select several rows from the Fourier matrix as the spread spectrum matrix.Simulation results show that,as compared with employing the unoptimized spread spectrum matrix,employing the optimized partial Fourier matrix as the spread spectrum matrix can obtain better channel estimation performance,higher user successful active detection rate and lower user data Symbol error rate.