Beam Selection for Hybrid Beamforming with Multi-Path Propagation: Novel Learning Architectures and Sufficient Statistics
| Main Authors: | Antón-Haro, Carles, Mestre, Xavier |
|---|---|
| Format: | Proceeding Journal |
| Bahasa: | eng |
| Terbitan: |
, 2020
|
| Subjects: | |
| Online Access: |
https://zenodo.org/record/4459437 |
Daftar Isi:
- In this paper, we investigate the applicability of deep and machine learning (ML/DL) techniques to beam selection problems. Specifically, we adopt a hybrid beamforming architecture comprising an analog beamforming (ABF) network followed by a zero-forcing (ZF) baseband processing block. The goal is to select the element in the ABF codebook yielding the highest sum-rate. The multi-antenna system operates in 5GNR’s Frequency Range 2 and, accordingly, the ML/DL-based architecture has been designed to explicitly consider a number of practical aspects of such mmWave communication systems. In particular, the presence of multi-path propagation along with the use of multi-carrier signals precludes the use of (single) angle-ofarrival information as an input to the learning system. Therefore, we investigate here alternative sufficient statistics (SS) such as the singular vector/values of the (multi-carrier) channel matrix, the average covariance matrix, or the covariance matrix at a given subcarrier. Besides, the novel ML/DL architecture enables a continuous operation of the system and avoids the spectral efficiency losses associated to periodically switching to a dedicated ABF for SS estimation. Computer simulation results illustrate the performance of several ML/DL approaches (k-nearest neighbors, support vector classifiers, multi-layer perceptron) in realistic 5G scenarios.
- Grant numbers : ARISTIDES - Aprendizaje Estadístico e Inferencia para Sistemas de Comunicación de Alta Dimensionalidad (RTI2018-099722-B-I00) and IRACON - Inclusive Radio Communication Networks for 5G and beyond ( 01 January 2016 - 01 January 2020) projects.