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

Large-scale multiple-input multiple-output(Large-MIMO)technology can significantly improve spectral and power efficiencies,which is considered as one of the most key technologies in the next generation wireless communications(5G),and therefore,has very high research value.Large-MIMO systems with tens or hundreds of antennas at the base station(BS)make the computational complexity increase sharply so that low complexity detection algorithms become the key to practical application.This paper studies and analyzes some signal detection algorithms in two aspects of complexity and performance in order to explore the algorithms which are more practical.This paper first introduces the advantages and difficulties of the Large-MIMO technology and points out the research direction and significance.Several kinds of traditional MIMO detection algorithms are simply introduced,and the related system suitability is deeply analyzed by theoretical derivation and simulation.Traditional nonlinear detection algorithms like sphere decoding(SD)can achieve the near-optimal performance but they are practical only up to certain dimensions.Traditional linear detection algorithms like minimum mean squared error(MMSE)are usually not adopted due to its poor performance,however,they will become a promising choice because of the channel particularity in Large-MIMO systems.Then,for the case where the number of base station antennas is far larger than the users,this paper focuses on an accelerated overrelaxation(AOR)iterative method.For Large-MIMO systems,the column vectors of the channel matrix are asymptotically orthogonal,thus,the linear signal detections like MMSE can achieve the near-optimal performance.Based on the analysis of channel hardening characteristics in Large-MIMO systems,this paper introduces Neumann series likelihood method and conjugate gradient method,which reduce the complexity of MMSE algorithm at the cost of performance.The AOR iterative method proposed in this paper,which can be regarded as the generalized definition of recent Jacobi and Gauss-Seidel method,is detailed studied and analyzed in respects of the principle of algorithm,the optimal performance parameters selection,the initial solution and LLR approximation.Gauss-Seidel estimation formula is proved to be the best estimate through formula deduction about AOR iterative method.In addition,the simulation results show that the two schemes including initial solution and LLR approximation can effectively reduce the complexity,and achieve near-optimal performance at the same time.Finally,in view of the nonlinear signal detection algorithms in Large-MIMO systems,this paper briefly introduces two kinds of typical low complexity algorithms,ie.,likelihood ascent search(LAS)and reactive tabu search(RTS).After a brief introduction of those detection principle,implementation steps and the optimization algorithms,a common defect is concluded that the performances of them are far away from SISO-AWGN performance in high-order modulation.Focused on parallel interference cancellation(PIC)algorithm,this paper proposes a reduced complexity iterative parallel interference cancellation(R-IPIC)algorithm,which can achieve better performance than MMSE algorithm with a lower complexity than AOR iterative method when the number of base station antennas is far larger than users.In the case of the number of base station antennas is close to users,R-IPIC algorithm can achieve better performance than ZF-LAS algorithm for low-order modulation.To conquer its defect of poor performance in high-order modulation,a thought of partial decision is introduced to R-IPIC algorithm,which effectively improves the performance of R-IPIC algorithm.In addition,Inspired by MMSE-ISDIC algorithm,this paper also proposes an improved MF-IPIC algorithm by applying the iterative mechanism in recent literature,and effectively improves the performance of the MF-PIC algorithm.