Research On Multi-user Detection For 5G Non-Orthogonal Multiple Access System

With the rapid development of mobile communication and Internet of Things,IMT-2020(5G)Advanced Group proposes that 5G will be located in wireless networks with higher spectral efficiency,faster speed and larger capacity.The orthogonal multiple access technology can no longer meet the demands of 5G.By introducing Non-orthogonal multiple access technology,massive user access and higher spectral efficiency can be achieved.One of these NOMA technologies is Sparse Code Multiple Access(SCMA),it can achieve high overload and high throughout capacity of the system by highdimensional codebooks,and it has become one of the research hotspots.Firstly,this paper introduces the basic theory of SCMA and some key technologies.Then give the basic uplink system model,the factor graph and Multi-dimensional mother constellation design of SCMA system.At the same time,the traditional multi-user detection scheme of SCMA system is introduced,including Maximum A Posterior(MAP)and Message Passing Algorithm(MPA).The MAP algorithm is the optimal multi-user detection algorithm in SCMA,and its complexity of computation is too high to be realized in practical application.Then the traditional MPA is introduced,simulated and analyzed.The results show that MPA as a suboptimal multi-user detection scheme,it can reduce the complexity of computation within a certain range and approach the performance of the MAP algorithm by the sparsity of the user codewords.However,as the number of users and the codebook dimension increase,the decoding complexity of the MPA is still growing at an exponential rate.Therefore,it is worthwhile to investigate to reduce the complexity of the MPA in the SCMA system.Secondly,this paper proposes a Symbol Flipping Algorithm(SFA),similar to the bit-flipping decoding algorithm in the Low-density Parity Check(LDPC)code field.Based on the reliability definition introduced in the SFA,the MPA algorithm based on partial marginalization is improved.In the MPA iterative process,the variable nodes that are marginalized can be dynamically determined,so it can reduce the times of iteration.This algorithm is called PM-based dynamic MPA algorithm(DPM-MPA).The simulation results show that the proposed SFA can greatly reduce the complexity of the detector,while DPM-MPA has lower complexity and better system performance than PM-MPA.Finally,based on Sphere Decoding(SD)algorithm,this paper proposes an improved multi-user detection algorithm called List Sphere Decoding based MPA(LSD-MPA),which can output the bit data soft decision of each user to the corresponding on the channel decoder.The simulation results show that compared with MPA,LSD-MPA can reduce the receiver complexity and also obtain significant gain.Because LSD can be regarded as a depth-first tree search algorithm,by exploring the SCMA codeword lattice structure,an optimization method is proposed to prun the redundant access nodes,and reducing the size of the search tree.We call this improved algorithm is NP-LSDMPA.At the same time,the optimal initial spherical radius is given,so that the algorithm can meet the nodes pruning condition faster.The simulation results show that the proposed LSD-based improved algorithm can greatly reduce the decoding complexity,and the performance loss is closer to MPA compared with the existing algorithms.