M.Eng. Ghazaleh Kia defends her doctoral thesis "Utilization of Machine Learning Algorithms and Radio Frequency Signals to Remove the Barriers of Low-Cost Accurate Seamless Localization" on Tuesday the 9th of April 2024 at 13 o'clock in the University of Helsinki Chemicum building, Auditorium A129 (A.I. Virtasen aukio 1, 1st floor). Her opponent is Associate Professor Stephan Sigg (Aalto University) and custos Professor Hannu Toivonen (University of Helsinki). The defence will be held in English.
The thesis of Ghazaleh Kia is a part of research done in the Department of Computer Science at the University of Helsinki and the Electrical Engineering unit at Tampere University. Her supervisor has been University Lecturer Jukka Talvitie (Tampere University).
Utilization of Machine Learning Algorithms and Radio Frequency Signals to Remove the Barriers of Low-Cost Accurate Seamless Localization
The rapid evolution of wireless technologies opens up new possibilities and avenues for advancement in the field of positioning. However, as these technologies introduce increasingly complex signal data, they also give rise to methodological challenges within the realm of machine learning. This thesis embarks on a comprehensive exploration of the gaps in achieving precise positioning, capitalizing on the potential of emerging wireless technologies, and harnessing the power of innovative machine learning algorithms to tackle these challenges with an economically viable solution.
Recognizing the profound impact of wireless technology on positioning methods, this study underscores the imperative of enhancing position solutions. From a positioning perspective, the primary objective of this work revolves around offering cost-effective, fast and accurate alternatives for scenarios where existing solutions prove financially prohibitive and not accurate enough for critical scenarios. Viewed through a telecommunications perspective, the thesis delves into the realm of novel Radio Frequency (RF) technologies that exhibit the capacity to overcome positioning challenges. Within the domain of machine learning, the research endeavors to employ the most suitable algorithms to effectively address positioning challenges rooted in RF data. Specifically, the thesis strives to present a solution that is both cost-effective and capable of delivering accuracy in localization with utmost efficiency.
Divided into three distinct stages, the thesis addresses the concept of seamless localization. The initial phase delves into the complexities inherent in achieving accurate indoor positioning and propose a pioneering solution founded upon a novel methodology and algorithm, centered around sensor fusion. The second stage presents an innovative low-cost solution using deep neural networks and 5G NR signals for accurate outdoor positioning. Finally, the last stage introduces an innovative infrastructure-independent mechanism for environment detection, thereby facilitating seamless localization transitions between indoor and outdoor settings. The empirical findings of this research demonstrate that machine learning methodologies possess the potential to yield economical solutions to intricate positioning challenges by discerning underlying patterns within RF signals. Through its multifaceted exploration, this thesis not only contributes to the advancement of positioning techniques but also establishes a framework for synergy between wireless technologies, machine learning, and cost-effective solutions, thereby shaping the future landscape of seamless localization.
Availability of the dissertation
An electronic version of the doctoral dissertation will be available on the e-thesis site of the University of Helsinki at http://urn.fi/URN:ISBN:978-952-84-0112-4.
Printed copies will be available on request from Ghazaleh Kia: ghazaleh.kia@helsinki.fi.