Research On The Low Complexity Detection Algorithms For Large-scale MIMO Systems

The rapid development of mobile Internet and Internet of Things provides a broad application prospect for the 5th Generation Mobile Communication System?5G?.Large-scale multi-input multiple-output?MIMO?technology is recognized as one of the core technologies in 5G systems because of its good link stability,high spectrum utilization and so on.However,the signal detection in receiver is more complex when the number of antennas reaches tens or even hundreds,thus whether the transmission signal can be correctly detected at the receiver is an important factor affecting the practical application of the large-scale MIMO systems.The research contents of this paper mainly include two aspects:one is the linear detection algorithm solving the issues in scenarios that the number of transmitter antennas is far less than the number of receiver antennas or the number of users is far less than the number of base station antennas.The second is that the linear detection algorithm performs poor where the number of transmitter antennas is close to the number of receiver antennas or the number of users is close to the number of base station antennas,thus the following nonlinear algorithms are studied in this paper,namely the compressed sensing detection algorithm based on sparsity,likelihood ascent search algorithm,reactive tabu search algorithm and reduced neighborhood algorithm.First,the linear detection algorithm is applied in the case where the number of antennas at the transmitter is much smaller than the number of antennas at the receiver or the number of users is much smaller than the number of base station antennas.The detection performance of the Minimum Mean Square Error?MMSE?algorithm is better than the matched filter?MF,Matched Filter?algorithm and the Zero Forcing?ZF?algorithm,but the computational complexity is high.To reduce the complexity of the weighted matrix inverse in MMSE algorithm,this paper analyzes two improved algorithms:Neumann series likelihood algorithm and iterative method,iterative method includes Gauss-Seidel iteration method,Richardson iterative method,successive over relaxation?SOR?iterative method,symmetric successive over relaxation?SSOR?iterative method and so on.Under the premise that the expansion term of Neumann series likelihood method is not more than 2,the computational complexity of Neumann series likelihood algorithm and various iterative methods is O?7?N _t²?8?.Compared with MMSE algorithm,the complexity can be reduced by an order of magnitude,which is more conducive to the practical application of large-scale MIMO systems.In order to speed up the convergence of the iterative method,this paper analyzes the influence of relaxation factor on Richardson,SOR and SSOR iterative method,and employs the diagonal estimation method and domain estimation method to obtain the initial solution.Through the comprehensive comparison,the three optimal ones in the iterative method are Gauss-Seidel,SOR and SSOR iterative methods.In the case of using the estimated initial solution for the Gauss-Seidel,SOR and SSOR iteration methods,Gauss-Seidel and SOR iterate twice,and the SSOR iterate once,which can achieve the performance of the MMSE detection algorithm and the computational complexity is very close.However,considering the influence of the relaxation factor on the SOR and SSOR iterative methods,the following conclusions can be drawn:if the optimal relaxation factor of the SOR and SSOR iterative methods can be estimated,the Gauss-Seidel,SOR and SSOR iteration methods can be applied;if the SOR and SSOR cannot obtain the optimal relaxing factor,the Gauss-Seidel iteration method should be employed.Then,when the number of transmitter antennas is close to the number of receiver antennas or the number of users is close to the number of base station antennas,the performance of the linear detection algorithm is not good,thus this paper analyzes the compressed sensing detection algorithm based on sparsity.Based on the generalized orthogonal matching pursuit algorithm,this paper proposes spartity-boosted iterative extend?SBIE?and concatenated sparstity-boosted iterative extend?CSBIE?detection algorithm.With the increase of the number of antennas,the detection performance of the SBIE algorithm is better and better.Through simulation analysis of SBIE-MF,SBIE-ZF,SBIE-MMSE algorithm,the SBIE-MF detection algorithm has excellent performance and low complexity.When the number of antennas is less,the performance of the concatenated SBIE-MF algorithm is lower than before.The performance of the concatenated SBIE-ZF and SBIE-MMSE algorithms is improved.When the bit error rate is 10^-3,the CSBIE-ZF and CSBIE-MMSE algorithms can obtain about 1 dB gain compared with the SBIE-ZF and SBIE-MMSE algorithms.Finally,the paper studies the basic principles and implementation steps of likelihood ascent search algorithm?LAS?and reactive tabu search?RTS?in detail,and analyzes and simulates its improved algorithm.Through simulation comparison,when the number of transmitter antennas is close to the number of receiver antennas or the number of users is close to the number of base station antennas in the large-scale MIMO systems,the LAS and RTS algorithms show better detection performance than the SBIE-MF algorithm.This paper analyzes the reduced neighborhood algorithm and proposes a method to select neighborhood metrics.Based on the reduced neighborhood,the detection performance of LAS and RTS does not degrade,nevertheless,the reduced neighborhood algorithm reduces the computational complexity by reducing the number of search vectors.The number of LAS algorithm search vectors decreases by more than 90%.The number of search vectors for the RTS algorithm drops by approximately 90%.It can be seen that the LAS and RTS algorithms based on the reduced neighborhood are more suitable for the detection of practical large-scale MIMO systems than the LAS and RTS algorithms.