Signal Detection Algorithms And Implementations For Massive MIMO Systems

By increasing the number of transmitting and receiving antennas of the systems,massive multiple-input multiple-output(MIMO)technology can significantly improve the data rate,transmission reliability,and spectral efficiency.It becomes one of the key technologies of 5G wireless communication systems.Signal detection plays an important role in signal processing of massive MIMO systems,and its performance will affect the communication quality of the entire system.The maximum likelihood detection(ML)algorithm achieves the best signal detection performance with the highest computational complexity.In order to trade off the performance and complexity,a series of linear and non-linear detection algorithms have been explored and studied.In this paper,a low-complexity message passing detection(LCMPD)algorithm is proposed.By updating the transmitted information approximately and fixing the constellation points with the maximum probabilities,the calculation of information update can be significantly simplified and the required number of iterations can be reduced with negligible BER performance loss.This paper also proposes a serial maximum likelihood(SML)algorithm.After decomposing the signal vector and specifying the initial search point,the signals are updated through the serial single dimension searching to reduce the complexity.This algorithm can achieve better performance than MMSE,which is suitable for the massive MIMO systems with low-order modulation and multi-user.In addition,a hybrid preprocessing conjugate gradient(HPCG)detection algorithm is also proposed,where an approximate preprocessing matrix and an iterative serial update scheme are used,making the BER performance of the HPCG algorithm superior to that of the MMSE under different antenna configurations of the system.Compared with the conjugate gradient(CG)series detection algorithms,the HPCG algorithm has the lowest complexity.In this paper,an improved Gauss-Seidel algorithm with initialization(IGSI)is designed in cooperation with the hardware.The algorithm is optimized based on hardware implementation.First,the Taylor series expansion is used to approximate the inverse of diagonal matrix,and then the lower-triangle matrix performs group inversion to reduce the processing delay of the system.Based on the IGSI algorithm,an efficient hardware architecture of signal detector is designed and implemented on FPGA.The FPGA implementation results show that the throughput of the detector is 2.6 times higher than that of the hardware design based on GSI algorithm,and it has higher hardware efficiency.