Research On Key Technologies Of High Order SCMA Systems

In the future Internet of Things(IoT)age,not only people but also things will con-nect to the networks.People and people,machine and machine,people and machine can communication with each other.The fifth generation(5G)communication systems are facing so many challenges as follows:the contradiction between ever-increasing links and limited spectrum resources,millisecond end to end delay and the link per-formance under high user load.Non-orthogonal multiple access(NOMA)techniques emerge at the right moment and sparse code multiple access(SCMA)has the best link performance among numerous NOMA techniques.However,the existing studies about SCMA are mainly restricted to low order systems.The decoding complexities of corre-sponding schemes grow exponentially which also bring higher symbol error rate(SER)especially under high user load.This dissertation mainly studies the high order SCMA systems,explores the primary problems when applied in real physical scenarios and proposes the corresponding key technologies,whose main work contents include three aspects as follows:(1)Based on the idea of hierarchical coding,a scheme of serial SCMA systems,which converts the problem of enabling a high order SCMA system to the problem of enabling several low order subsystems,is proposed and the corresponding mapping modules which can recover the original user information from each low order subsystem are also built.Moreover,the dissertation also discusses the codebook design criteria of each low order subsystem and the receiver design for the serial SCMA.Besides,the dis-sertation compares and analyzes several performance indices of the proposed scheme to enable high order SCMA such as overloading factor,energy efficiency,block error rate(BLER)and decoding complexity with original SCMA and low density signature(LDS).Several low order systems make the adjustment of system parameters more flex-ible so as to adapt to dynamic network environments(such as the variable number of users).(2)Based on the idea of Gaussian approximation,an iterative decoding algorithm is proposed in which the message update adopts the propagation of expectation and variance instead of the propagation of beliefs during each iteration.The decoding com-plexity of a high order SCMA system increases exponentially as the system order grows,and this is attributed to the message passing algorithm(MPA)which is used in the ex-isted SCMA systems and needs to search all the nodes in the solution space during each iteration.By studying the message passing processes among all the nodes and the prop-erties of the solution space,optimizing the decoding procedure and cutting down the computational complexity,enabling high order SCMA systems will also become more feasible.Therefore,this dissertation uses Gaussian distribution to match the real distri-bution of each message in solution space during MPA iterations.Simultaneously,the operations of the sum of the messages in each node are converted to the operations of computing expections and variances.(3)A parallel SCMA codebook design scheme is proposed while the strategies of codebook assignment among users and user activity detection are also devised regarding to user features and activities.SCMA codebooks directly influence the error probability bounds of decoding and the difficulty of SCMA codebook acquisition is also the key factor which restricts the high order SCMA realization.In order to obtain the optimal codebooks for all the users,the user diversities need to be considered rather than adopt-ing the methods of randomly assigning codebooks among users as the existed SCMA systems.Therefore,the dissertation divides SCMA codebook design into the codebook assignment among users,the factor graph design and multi-dimensional constellation design.The codebook assignment and factor graph can be obtained simultaneously by solving the convex optimization problem which is built taking the detection complexity or system power consumption as the optimization obiective.On the multi-dimensional constellation design,the dissertation proposes a scheme which generates the high order codebooks through several low order codebooks and aims to control the system perfor-mance bounds,i.e.parallel SCMA.Besides,in practical scenarios,the simultaneously activated users only make up a small proportion of all the users.In order to further im-prove the link performance and save the system overhead,the dissertation designs the corresponding scheme to conduct the user activity detection.