Inversión aproximada de matrices en sistemas Massive MIMO correlados en tiempo o frecuencia

Abstract

Massive multiple-input multiple-output (MIMO) is expected to be one of the keys in 5G. In this technology, the base station is equipped with a big number of antennas serving multiple users simultaneously to improve spectral efficiency, coverage, and range. Zero-Forcing and Minimum Mean Square Error have been considered potential practical precoding and detection methods for large scale MIMO systems but require much larger dimensions of matrix inversion. This paper presents an architecture for approximate matrix inversion based on Neumann Series, thereby reducing the cost of hardware. In addition, we propose a solution for systems with time or frequency correlation among different channels where we are able to reach a much higher throughput.