Gram-Schmidt precoding for two-tier cellular networks with massive MIMO

Abstract

The uplink in a cellular network where the base station (BS) and an advanced user equipment (UE) that is not the traditional mobile device is investigated. This is proposed as a method of antenna offloading for the typical UE which suffers from size constraints. The Gram-Schmidt (GS) precoding algorithm is used to design novel precoding solutions when the receiver and transmitter can have a large antenna array (e.g. N r = 64, N T = 32), as expected to be the case for deploying wireless backhaul links. Using the QR-decomposition on the Zero-Forcing (ZF) receiver, the proposed scheme is shown to essentially perform channel inversion with the help of the transmitter and receiver. With a moderate number of receive antenna elements, the matched filter (MF) was observed to incur inter-stream interference, while the ZF and minimum mean squared error (MMSE) receiver were more susceptible to transmitter side correlation compared to the GS precoding algorithm. When there is only channel correlation information (CCI) at the transmitter side, the GS precoding matrix is shown to be approximated by inverting the Cholesky decomposition of the CCI transmit matrix and outperforms the MF. Finally a block successive interference cancellation (SIC) detection scheme for the K-User MIMO channel is presented. It is shown for N T = 2 the GS precoder has a higher sum rate than the ZF receiver in the high signal to noise ratio (SNR) regime.