Page 25 - ITU Journal Future and evolving technologies Volume 2 (2021), Issue 4 – AI and machine learning solutions in 5G and future networks

ITU Journal on Future and Evolving Technologies, Volume 2 (2021), Issue 4

SITE‑SPECIFIC MILLIMETER‑WAVE COMPRESSIVE CHANNEL ESTIMATION
ALGORITHMS WITH HYBRID MIMO ARCHITECTURES

4,5
2,3
2,3
6
1
Sai Subramanyam Thoota , Dolores Garcia Marti , Özlem Tuğfe Demir , Rakesh Mundlamuri , Joan Palacios , Cenk M.
6
7
4,5
7
1
Yetis , Christo Kurisummoottil Thomas , Sameera H. Bharadwaja , Emil Björnson , Pontus Giselsson , Marios
1
8
6
Kountouris , Chandra R. Murthy , Nuria González‐Prelcic , Joerg Widmer 2
1 2 3
Indian Institute of Science, Bangalore, India, IMDEA Networks, Madrid, Spain, Universidad Carlos III, Madrid, Spain,
4 5 6
KTH Royal Institute of Technology, Stockholm, Sweden, Linköping University, Linköping, Sweden, EURECOM,
8
7
Sophia‐Antipolis, France, Lund University, Lund, Sweden, North Carolina State University, Raleigh, USA
NOTE: Corresponding author: Sai Subramanyam Thoota, thoota@iisc.ac.in
Abstract – In this paper, we present and compare three novel model‑cum‑data‑driven channel estimation procedures in a
millimeter‑wave Multi‑Input Multi‑Output (MIMO) Orthogonal Frequency Division Multiplexing (OFDM) wireless communi‑
cation system. The transceivers employ a hybrid analog‑digital architecture. We adapt techniques from a wide range of signal
processing methods, such as detection and estimation theories, compressed sensing, and Bayesian inference, to learn the un‑
known virtual beamspace domain dictionary, as well as the delay‑and‑beamspace sparse channel. We train the model‑based
algorithms with a site‑speci ic training dataset generated using a realistic ray tracing‑based wireless channel simulation tool.
We assess the performance of the proposed channel estimation algorithms with the same site’s test data. We benchmark the
performance of our novel procedures in terms of normalized mean squared error against an existing fast greedy method and
empirically show that model‑based approaches combined with data‑driven customization unanimously outperform the state‑
of‑the‑art techniques by a large margin. The proposed algorithms were selected as the top three solutions in the “ML5G‑PHY
Channel Estimation Global Challenge 2020” organized by the International Telecommunication Union.
Keywords – Bayesian inference, channel estimation, compressed sensing, data‐driven, hybrid MIMO, mmWave
1. INTRODUCTION hybrid MIMO multiple antennas con‐
nected to an RF chain using a phase shifter network (RF
Millimeter‐Wave (mmWave) wireless communication is
precoder/combiner), digital precoder/combiner
one of the potential technologies proposed for the next
complex side of the
generation communication systems (5G and beyond) to
transceiver The RF and digital precoders/combiners are
meet the ever‐increasing demand for high data rates. The
igured system performance metric
mmWave frequency spectrum, ranging from 30 GHz to
or signal interference noise ra‐
300 GHz, is attractive because it offers large bandwidths Unlike a fully analog architecture, a hybrid architec‐
(∼ 2GHz), resulting in very high data rates and low la‐
ture allows one to reduce the number of RF chains, while
tency. These advantages come at a cost of higher path loss
multi‐stream multi‐user transmissions.
due to several factors, such as blockages and oxygen ab‐
sorption at mmWave frequencies, which in turn bring sev‐ The major challenges then are in estimating the mmWave
wireless channel and con iguring the RF and digital pre‐
eral engineering challenges in adopting this technology in
coders/combiners The
commercial wireless communication systems.
problem is exacerbated by the fact that only the low di‐
A potential solution to overcome this problem is beam‐ mensional RF combined signals at the baseband are avail‐
forming, which leverages the availability of multiple an‐ able for estimating the channel. Since the system does not
have any knowledge of the channel state during the chan‐
tennas at the transmitter and receiver. In particular,
nel estimation phase, the baseband precoders/combiners
millimeter wavelengths enable one to accommodate a
are set to the identity matrix and random phase shifts are
larger number of antennas into the same physical space,
chosen for the RF precoders/combiners.
and thereby attain high beamforming gains. However,
a fully digital architecture in a Multi‐Input Multi‐Output
(MIMO) system, i.e., one Radio Frequency (RF) chain per MmWave channel estimation in a hybrid MIMO architec‐
antenna, and one complex‐valued Analog‐to‐Digital Con‐ well studied provide brief
verter (ADC) per RF chain is less appealing both from overview of some of the key existing literature here. The
commercial and engineering perspectives due to its high simplest channel estimation method in hybrid MIMO sys‐
cost and energy requirements. Therefore, a hybrid MIMO tems is the Least Squares (LS)‐based approach [2], which
architecture is proposed in the literature as a potential so‐ is inherited from conventional MIMO A more re ined
lution to solve this problem [1]. solution to channel estimation is to exploit both the delay

© International Telecommunication Union, 2021 9