Polynomial Expansion Multiuser Detection For Massive MIMO

Multiple input multiple output (MIMO) technique is a mature technology incorporated into the emerg-ing wireless broadband standard, as a key technology in modern wireless communication system standards, such as long-term evolution (LTE) project. In general, if the BS (Base station) or the terminals equip more antennas, the system can achieve better performance both in throughput and stability, because the transmis-sion channel can provide a higher degree of freedom. In the next generation mobile communication system, MIMO technology requires more large-scale configuration of the antenna at the BS side (more than hundreds of antennas) to serve more users at the same time. As a result, the greatly increased consumption of system become challenges. And one of the key problems is the design of the low-complexity receiver. This paper has done research on the signal detection technology based on the polynomial series expansion theorem for the massive MIMO systems, and the main contents and contributions are as follows:Firstly, the knowledge of signal detection technology for massive MIMO system is introduced, and also the feasibility of linear detection technology is analyzed in the new situation. The computational complexity of the linear signal detection technology is very high in the massive MIMO systems because of the large-scale matrix inversion of the correlated matrix of channel response matrix. To this end, the approximative matrix inversion method based on the Neumann series expansion (NSE) is proposed and then applied in the proposed MMSE-DNSE signal detection algorithm. The proposed MMSE-DNSE method has the advantage of the low complexity and it is very hardware friendly.Then, in order to further exploit the advantages of NS method, MMSE-INSE signal detection method is proposed. This method combines the characteristics of MMSE signal detection and NS expansion, greatly saving the hardware resources by the way of iterative calculation method. Then the corresponding VLSI architecture is proposed for the MMSE-INSE algorithm, and the results of simulation present better hardware performance compared with the classical Cholesky method. For the scenes where require the high order NS items, the MMSE-RNSE algorithm is proposed, which can achieve the fast iterative calculation of NS method and also improve the hardware efficiency.Furthermore, we focus on the sceneries of unideal channel transmission and unfavorable system con-figuration. For the analysis of correlated MIMO channel, the Kronecker module is applied and developed for the new condition. The analysis is shown that the spatial correlation of antenna array reduces the BER performance of the NS method for the signal detection. To this end, the developed MMSE-TNSE method is proposed, which can improve the convergence stability and speed of applied NS even in the unfavorable transmission condition. Besides, the combination way of MMSE-TNSE method and MMSE-RNSE method can achieve an efficient performance.Finally, different from the previous algorithms, a soft-output MMSE signal detection algorithm is pro-posed. This linear algorithm can avoid the the operation of matrix inversion by the NS expansion, to the goal of the low computational complexity. And then, to achieve the demand for the soft-output information for the soft channel decoder, the LLR calculation algorithm is proposed which is also based on the NS expansion theorem. The overall computational complexity of the proposed soft-output signal detection algorithm has the advantage of O(K2).