Research On The Grant-free Meschanism Of Uplink SCMA System

Since 3GPP released the "independent Release 15(R-15)" standard,as a new architecture with complete functions and independent deployment capability in the first stage,non-orthogonal multiple access technology has been included in the standardization formulation.SCMA is one of the non-orthogonal multiple access technologies designed for 5G.For the low delay and high reliability communication scenario of one of the three test environments in the R-15 standard,and in the wireless interface function module,it is proposed that the grant-free access mode should be implemented.In this paper,an uplink SCMA system based on grant-free transmission mechanism is proposed,we focus on the design of pilot sequence,codebook preallocation algorithm at the transmitter,and the multi-user detection algorithm with fast convergence at the receiver.The specific works are as follows:In the pilot design and active user detection module,since compressed sensing observation matrix constituted by pilot sequence,two kinds of pilot sequence design schemes based on observation matrix optimization are proposed.Scheme 1 takes advantage of the good correlation of ZC sequences,dynamically allocate root sequence resources and adjust sequence length,improve detection performance of greedy detection algorithm.Due to the limitation of the number of ZC root sequences,the scheme may bring interference problems of neighboring cells.Scheme 2 reduces the cross-correlation between the column vectors of the observation matrix by optimizing the Gram matrix,and also ensures the performance of the detection algorithm.Then we propose a DSMbased BOMP algorithm,and verify that the AUD performance of the two dynamic pilot schemes is better than the Gaussian random pilot sequences and the fixed length ZC sequences.We use active users' CSI calculated by the previous module,an improved codebook pre-allocation algorithm based on interference graph is proposed,for the optimal CSI algorithm never consider the interference caused by sparse overload,the algorithm uses exhaustive search to obtain optimal solution of codebook allocation,which maximizes system capacity.Because the time complexity of the algorithm is too high when the number of users is very large,we propose a suboptimal interference graph algorithm to reduce the complexity.In the uplink SCMA decoding part,we focus on the fast convergence MPA algorithm,and propose two improved algorithms,GMPA and IS-GMPA.The GMPA algorithm uses the orthogonality between VN nodes to divide them into multiple resource orthogonal groups,using serial and parallel MPA decoding method.The IS-GMPA algorithm integrate a confidence ranking process.The SINR values in the group are sorted by the interference graph,simulation results show that the proposed code book preallocation algorithm has significantly improved the performance of decoding BER,and IS-GMPA algorithm has reduced the number of iterations by 1 and 2 respectively compared with GMPA and traditional MPA algorithm.