Navigating the Future — Machine Learning for Wireless Sensing and Localization
(2024)- Abstract
- Cellular-based localization and sensing pave the way for a variety of
applications across various domains, ranging from autonomous driv-
ing to emergency care and intelligent traffic management. Although
traditional methods have been effective, they still face challenges such as the
requirement for highly accurate models and the inherent complexity of the
algorithms. This thesis explores the potential of integrating machine learning
(ML) techniques to augment the performance of sensing and localization
systems. Three main topics are covered by this thesis, namely, ML-aided
channel estimation, sensing, and localization.
The first topic focuses on the calibration of the RF chain and the... (More) - Cellular-based localization and sensing pave the way for a variety of
applications across various domains, ranging from autonomous driv-
ing to emergency care and intelligent traffic management. Although
traditional methods have been effective, they still face challenges such as the
requirement for highly accurate models and the inherent complexity of the
algorithms. This thesis explores the potential of integrating machine learning
(ML) techniques to augment the performance of sensing and localization
systems. Three main topics are covered by this thesis, namely, ML-aided
channel estimation, sensing, and localization.
The first topic focuses on the calibration of the RF chain and the esti-
mation of the propagation channel, which serves as essential prerequisites
for numerous subsequent applications including arriving angle estimation,
radio localization, digital beamforming, and sensing. We introduce a novel
RF chain calibration algorithm for massive multiple-input multiple-output
(MIMO) systems, using uplink signals. We derive the maximum likelihood
estimator (MLE) and its corresponding Cramér-Rao Lower Bound (CRLB).
Additionally, we propose a novel ML-powered channel estimation pipeline for
orthogonal frequency division multiplex (OFDM) systems, which efficiently
estimates channel coefficients for all OFDM grids based on only using a
limited number of pilot signals.
The second topic addresses ML-based wireless sensing using massive
MIMO systems. We introduce a novel pipeline that first employs tensor-
decomposition algorithms to extract channel characteristics. Subsequently, a
fully connected neural network is deployed to classify human movements. Our pipeline is evaluated by a measurement campaign using a massive
MIMO testbed in an indoor environment, providing empirical evidence of
the system’s efficacy in wireless sensing applications. Our work shows the
sensing capacity of massive MIMO in such scenarios.
The third topic focuses on cellular localization, comprising three main
sections. The first section introduces classical radio localization algorithms,
including time of arrival (ToA), angle of arrival (AoA) and time difference
of arrival (TDoA). The second main section addresses ML-based localization
with a massive MIMO system. We introduce a novel pipeline composed of
several parallel processing chains. Each chain is trained on distinct channel
fingerprints, including the channel impulse response (CIR) and covariance
matrices. To improve localization accuracy, we model the position error in
each processing chain as a Gaussian distribution and combine the outputs to
compute localization uncertainty. Furthermore, we investigate the required
training density using the Nyquist theorem. Our pipeline is evaluated
through indoor measurements, showing a centimeter-level localization ac-
curacy. The third section investigates ML-aided cellular localization using
a 5G new radio (NR) system that operates in beam space. We present
an ML-based localization pipeline that integrates attention mechanisms and
advanced uncertainty estimation algorithms. Unlike in the previous section,
this uncertainty estimation approach is not restricted to a Gaussian assump-
tion. Validation of our pipeline is conducted through an extensive outdoor
measurement campaign encompassing both Line-of-Sight (LoS) and Non
Line-of-Sight (NLoS) propagation scenarios. Measurement results indicate
the sub-meter accuracy levels with our pipeline. (Less)
Please use this url to cite or link to this publication:
https://lup.lub.lu.se/record/65570162-08f7-4a2e-9ad9-904746bc9f97
- author
- Tian, Guoda LU
- supervisor
-
- Fredrik Tufvesson LU
- Ove Edfors LU
- Bo Bernhardsson LU
- Xuesong Cai LU
- opponent
-
- Prof. Tirkkonen, Olav, Aalto University, Finland.
- organization
- publishing date
- 2024-10-21
- type
- Thesis
- publication status
- published
- subject
- keywords
- Machine learning, Wireless localization, Massive MIMO system., Wireless sensing
- edition
- 1st
- pages
- 187 pages
- publisher
- Electrical and Information Technology, Lund University
- defense location
- Lecture Hall E:1406, building E, Ole Römers väg 3, Faculty of Engineering LTH, Lund University, Lund. The dissertation will be live streamed, but part of the premises is to be excluded from the live stream.
- defense date
- 2024-11-18 09:15:00
- ISBN
- ISBN 978-91-8104-120-0
- ISBN 978-91-8104-119-4
- language
- English
- LU publication?
- yes
- id
- 65570162-08f7-4a2e-9ad9-904746bc9f97
- date added to LUP
- 2024-10-21 23:30:17
- date last changed
- 2025-04-04 15:28:01
@phdthesis{65570162-08f7-4a2e-9ad9-904746bc9f97,
abstract = {{Cellular-based localization and sensing pave the way for a variety of<br/>applications across various domains, ranging from autonomous driv-<br/>ing to emergency care and intelligent traffic management. Although<br/>traditional methods have been effective, they still face challenges such as the<br/>requirement for highly accurate models and the inherent complexity of the<br/>algorithms. This thesis explores the potential of integrating machine learning<br/>(ML) techniques to augment the performance of sensing and localization<br/>systems. Three main topics are covered by this thesis, namely, ML-aided<br/>channel estimation, sensing, and localization.<br/><br/>The first topic focuses on the calibration of the RF chain and the esti-<br/>mation of the propagation channel, which serves as essential prerequisites<br/>for numerous subsequent applications including arriving angle estimation,<br/>radio localization, digital beamforming, and sensing. We introduce a novel<br/>RF chain calibration algorithm for massive multiple-input multiple-output<br/>(MIMO) systems, using uplink signals. We derive the maximum likelihood<br/>estimator (MLE) and its corresponding Cramér-Rao Lower Bound (CRLB).<br/>Additionally, we propose a novel ML-powered channel estimation pipeline for<br/>orthogonal frequency division multiplex (OFDM) systems, which efficiently<br/>estimates channel coefficients for all OFDM grids based on only using a<br/>limited number of pilot signals.<br/><br/>The second topic addresses ML-based wireless sensing using massive<br/>MIMO systems. We introduce a novel pipeline that first employs tensor-<br/>decomposition algorithms to extract channel characteristics. Subsequently, a<br/>fully connected neural network is deployed to classify human movements. Our pipeline is evaluated by a measurement campaign using a massive<br/>MIMO testbed in an indoor environment, providing empirical evidence of<br/>the system’s efficacy in wireless sensing applications. Our work shows the<br/>sensing capacity of massive MIMO in such scenarios.<br/><br/>The third topic focuses on cellular localization, comprising three main<br/>sections. The first section introduces classical radio localization algorithms,<br/>including time of arrival (ToA), angle of arrival (AoA) and time difference<br/>of arrival (TDoA). The second main section addresses ML-based localization<br/>with a massive MIMO system. We introduce a novel pipeline composed of<br/>several parallel processing chains. Each chain is trained on distinct channel<br/>fingerprints, including the channel impulse response (CIR) and covariance<br/>matrices. To improve localization accuracy, we model the position error in<br/>each processing chain as a Gaussian distribution and combine the outputs to<br/>compute localization uncertainty. Furthermore, we investigate the required<br/>training density using the Nyquist theorem. Our pipeline is evaluated<br/>through indoor measurements, showing a centimeter-level localization ac-<br/>curacy. The third section investigates ML-aided cellular localization using<br/>a 5G new radio (NR) system that operates in beam space. We present<br/>an ML-based localization pipeline that integrates attention mechanisms and<br/>advanced uncertainty estimation algorithms. Unlike in the previous section,<br/>this uncertainty estimation approach is not restricted to a Gaussian assump-<br/>tion. Validation of our pipeline is conducted through an extensive outdoor<br/>measurement campaign encompassing both Line-of-Sight (LoS) and Non<br/>Line-of-Sight (NLoS) propagation scenarios. Measurement results indicate<br/>the sub-meter accuracy levels with our pipeline.}},
author = {{Tian, Guoda}},
isbn = {{ISBN 978-91-8104-120-0}},
keywords = {{Machine learning; Wireless localization; Massive MIMO system.; Wireless sensing}},
language = {{eng}},
month = {{10}},
publisher = {{Electrical and Information Technology, Lund University}},
school = {{Lund University}},
title = {{Navigating the Future — Machine Learning for Wireless Sensing and Localization}},
url = {{https://lup.lub.lu.se/search/files/197992389/Thesis_Digital_Nailing.pdf}},
year = {{2024}},
}