Research On Low Complexity MPA Multiuser Detection Algorithm In SCMA System

The mobile communication technology has been developed to the 5th-Generation(5G).As a kind of Non-orthogonal Multiple Access(NOMA),Sparse Code Multiple Access(SCMA)technology can solve the problem of massive access by using sparsity of SCMA codeword.The Message Passing Algorithm(MPA)is main multi-user detection method of SCMA syste,m,because computational complexity of MPA is high.The existing improved algorithm mainly starts with simplifying the factor graph and speeding up convergence.However,the removal of the branch will cause the error to spread because some messages are not converged,while the performance of Bit Error Ratio(BER)will degrade in research.Even though MPA converges faster by using the serial update method,the computational delay is increased.Therefore,the complexity of MPA needs to be reduced,the balance between complexity,BER performance and computational delay needs to be obtained.This thesis will focus on the low complexity MPA multi-user detection algorithm.The main contents of research are as follows:(1)The research of MPA finds that the main reason for the high complexity of MPA is that it updates all the messages of the factor graph in iteration.Different codeword messages converge at different speeds,a certain proportion of slower-converging codeword messages are chosen to update to reduce error diffusion,Dynamically Select Message to Update-Message Passing Algorithm(DSM-MPA)is proposed in this thesis by this feature.The complexity of MPA is reduced by the total number of message updates is decreased.The spread of error caused by some messages are not converged is reduced.When the appropriate weight factor is chosen,the BER performance of the proposed algorithm is basically the same as that of the original MPA,better than algorithm such as Dynamic Factor Graph-MPA(DFG-MPA),and the complexity of algorithm is significantly lower.(2)However,the convergence of codeword messages selected for DSM-MPA algorithm has randomness and volatility,which cannot adapt to changes of channel conditions.Therefore,in response to this problem,by setting the specific convergence threshold to determine the actual convergence of the message and reducing the randomness and volatility of the message selection,Judgment-Message Passing Algorithm(J-MPA)is proposed in this thesis.All messages are closer to the original convergence value.This algorithm can adapt to changes of channel by compensating for the performance loss due to randomness and volatility,and the better the channel conditions is,the lower the complexity has.(3)Both DSM-MPA and J-MPA belong to parallel MPA,and the convergence speed of parallel MPA is slow.The fast-converging serial Residual-aided MPA(RA-MPA)is studied in this thesis.Aiming at the problem that the delay of RA-MPA is long,User Node-Message Passing Algorithm(UN-MPA)is proposed by research.The computational delay is reduced when branches is updated by grouping.The branches on each user node are updated one by one according to the confidence of each node group,branches of each group are updated in parallel.The proposed method maintains the fast convergence and low complexity characteristics of the serial MPA,while the BER performance of UN-MPA is basically the same with algorithm such as the RA-MPA and the original MPA,and is superior to the DSM-MPA.The delay is lower than the RA-MPA by updating the edge branches on the user node in parallel.UN-MPA has a good balance between the three properties of BER,computational complexity and delay.