ÃÛ¶¹ÊÓƵ

The introduction of Multiple-Input Multiple-Output (MIMO) systems has dramatically transformed wireless communication systems, in particular in the Fifth Generation (5G) New Radio (NR) systems, fundamentally changing how signals are transmitted and received. MIMO technology deploys numerous antennas to transmit and receive multiple data streams simultaneously. The presence of obstructions and scatterers in wireless environments, varying in location, size, and shape, contributes to a high-dimensional feature space. As user devices move, the interaction of electromagnetic radio waves with surrounding objects and devices generates distinct patterns, called spatial fingerprints. By analyzing the behavior of the radio channel in real time through these spatial fingerprints and their temporal evolution, MIMO systems unlock significant opportunities for deeper insight into channel dynamics. These insights lay the groundwork for previously unforeseen functionalities in the Radio Access Network (RAN) domain of cellular networks, moving beyond the constraints of traditional approaches based on mathematical models and solutions.
This thesis aims primarily to utilize channel measurements generated in commercial MIMO systems to explore the underlying statistical structures from real-time data, eliminating or mitigating the need for precise mathematical modeling. Recognizing that conventional solutions are unlikely to deliver the performance enhancements required for future wireless networks, the research presented in this thesis explores innovative approaches and tools to push the boundaries of this field. Over the past decade, a new era of Machine Learning (ML) and Artificial Intelligence (AI) techniques, particularly Deep Learning (DL), has emerged as a powerful alternative for designing and optimizing wireless networks, as demonstrated in this thesis. The subsequent chapters of this thesis begin with an introductory section that outlines the theoretical background that serves as the foundation for the research topic. This is followed by a collection of papers that detail the conducted studies. The six papers included in this thesis encompass three key research areas: user device clustering, user positioning, and traffic pattern-related predictions.
The first and fourth papers focus on user classification, or grouping, based on channel fingerprints derived from measurement data. The first paper explores the feasibility of using channel measurements as a data source to classify users based on their spatial proximity, density, and velocity. In contrast, the fourth paper demonstrates grouping based on user position and direction using commercial 5G measurement data.
The second research area focuses on cellular positioning, with the second paper among the first to demonstrate the feasibility of user positioning using commercial 5G NR beam measurement data. Building upon the findings of the second paper, the fifth paper refines positioning accuracy through an attention-based AI model and advanced statistical post-processing techniques. The results showcase a sub-meter level of positioning accuracy.
As part of the last research area, traffic pattern-related predictions, the third paper proposed a customized cell handover prediction strategy for dense urban environments. This work emphasizes a user-context-aware handover process, with the aim of improving the efficiency and reliability of handovers in complex network scenarios. Finally, the sixth paper provides novel insights into 5G beam management strategies for long- and short-term channel predictions. This innovative approach introduces a highly accurate, attention-based prediction model capable of deriving the complete downlink transmission chain in a commercial-grade 5G system. The model demonstrates precise beam predictions extending far beyond coherence time, specifically addressing the challenges posed by Non Line-of-Sight (NLOS) environments characterized by complex, high-dimensional channel dynamics.