Antenna selection in massive MIMO using non-central Principal Component Analysis

Abstract

Massive MIMO has the potential to offer high throughput in today's fast wireless communication systems, however, the large number of antennas and RF chains needed at the transmitter brings the challenge of high system complexity and hardware energy consumption. In this paper, two semi-heuristic techniques are proposed for practical antenna selection in a multi-user MIMO broadcast system using principal components analysis (PCA) to reduce signal correlations at users and remove antennas that contribute least to system sum capacity, thereby reducing the number of RF chains needed. Zero forcing precoding is used and users are equipped with a single antenna. Using analytic methods PCA eigenvalues are decomposed into two components: the mean channel gain component and the channel correlation component. Using simulation we show that the proposed antenna selection methods perform much better using mean channel gain selection and show how antenna selection depends on the channel matrix structure.