A Low-Complexity Soft-Output Signal Data Detection Algorithm for UL Massive MIMO Systems

In massive multiple-input multiple-output (MIMO) systems, although the performance of maximum likelihood (ML) is the optimum, it introduces extremely high computational complexity, while minimum mean square error (MMSE) receivers can achieve quasi-optimal performance. Unfortunately, it requires a matrix inverse which increases the computational complexity in high loaded environments. Several methods have been proposed to avoid the matrix inversion such as the accelerated over relaxation (AOR). In the AOR algorithm, the initial solution and the optimum parameters have a great impact on the performance, computational complexity, and the convergence rate. In this paper, a detector based on AOR and a stair matrix is proposed to iteratively avoid the inverse of equalization matrix and expediting the convergence rate. In order to obtain high performance and low complexity, suitable schemes for the selection relaxation and acceleration parameters are also proposed. Numerical results show that the computational complexity of the proposed AOR approach is dramatically reduced from O(K-3) to O(K-2) where K is the number of users. It is also shown that the proposed detection algorithm outperforms the Neumann series method and achieves a quasi-optimal performance with a relatively small number of iterations.

Automated summary

This paper proposes a detector based on AOR and a stair matrix to iteratively avoid the inverse of equalization matrix and expedite the convergence rate. Suitable schemes for selecting relaxation and acceleration parameters are also proposed for high performance and low complexity. Numerical results show a dramatic reduction in computational complexity and better performance compared to the Neumann series method.