Learning-based Max-Min Fair Hybrid Precoding for mmWave Multicasting
Abstract
This paper investigates the joint design of hybrid transmit precoder and analog receive combiners for single-group multicasting in millimeter-wave systems. We propose LB-GDM, a low-complexity learning-based approach that leverages gradient descent with momentum and alternating optimization to design (i) the digital and analog constituents of a hybrid transmitter and (ii) the analog combiners of each receiver. In addition, we also extend our proposed approach to design fully-digital precoders. We show through numerical evaluation that, implementing LB-GDM in either hybrid or digital precoders attain superlative performance compared to competing designs based on semidefinite relaxation. Specifically, in terms of minimum signal-to-noise ratio, we report a remarkable improvement with gains of up to 105% and 101% for the fully-digital and hybrid precoders, respectively.
Cite
@article{arxiv.2002.00670,
title = {Learning-based Max-Min Fair Hybrid Precoding for mmWave Multicasting},
author = {Luis F. Abanto-Leon and Gek Hong and Sim},
journal= {arXiv preprint arXiv:2002.00670},
year = {2020}
}
Comments
7 pages. To be appear in IEEE ICC 2020 Proceedings