Research On The LDPC-coded Sparse Code Multiple Access System

Sparse code multiple access(SCMA)is a spectral efficient non-orthogonal transmission scheme.It has become a promising candidate for multiple access techniques in 5G communications,which aims to support massive capacity and connections in emerging application scenarios such as machine-type communications.Compared with other non-orthogonal multiple access schemes,SCMA systems obtain shaping gain and diversity gain with the use of multi-dimensional constellations.Meantime,the resource allocation strategies in SCMA systems are described by a sparse non-orthogonal superposition matrix,which the message passing algorithm(MPA)with near-optimal performance can be performed on.However,current researches about codebook design have met their bottlenecks,which is restricted to extending to large-size constellation and indicator matrix design.In this thesis,design for an SCMA system under a coded scenario is studied with emphasis on the codebook design.The main contributions are as follows.To begin with,the basic principle of SCMA system is introduced,along with elaboration of the resource allocation schemes for users and multi-user detection algorithms represented by message passing algorithm(MPA)at the receiver.Then simulations for SCMA systems with different loads are presented,which demonstrate SCMA systems can achieve near single user bound even under the overloading setting.In order to decrease the detection complexity while exploiting the feedback from decoders,a joint detection decoding algorithms based on approximate message passing(AMP)is proposed.With acceptable performance loss,the proposed algorithm can achieve linear complexity.After a general introduction of the SCMA system,the applications of SCMA techniques in coded systems are discussed.An incremental redundant hybrid ARQ(IR-HARQ)mechanism based on rateless codes is proposed,with the joint detection decoding algorithm,the error performance can be substantially improved.Besides,a nonbinary LDPC-coded system is proposed for high-order modulation,which exhibits 1d B performance gain compared to its binary counterpart.When SCMA techniques are employed in multi-cell settings,it can cope with the problem of interference management and seamless handover,even though it shows early error floor under the influence of large-scale fading.Based on the coded SCMA system,the codebook design is studied,which consists of highorder multi-dimensional constellations designed for multi-user systems and indicator matrix construction.A method to construct low-projection multi-dimensional constellation based on Cartesian product is proposed,which results in better Euclidean distance of superposed signals with the reduction of detection complexity and MPA iteration rounds.In this thesis,a comparison of state-of-art constellation design in terms of principles,key performance indicators(KPI)and error rates is investigated.The proposed constellation outperforms other constellations with more than 1.2d B gain under a coded system with code rate 0.5,while it diverges in the uncoded system.Finally,design for the system of larger size is discussed,where the matrix is constructed by lifting or spatial-coupling based on protographs with PEG or algebraic-aided algorithms.The resulting matrix demonstrates 1d B gain over the original6 user-4 resource one which is optimal in term of sum rate.