Deep learning for intelligent and autonomous wireless networks

Amorosa, Lorenzo Mario (2025) Deep learning for intelligent and autonomous wireless networks, [Dissertation thesis], Alma Mater Studiorum Università di Bologna. Dottorato di ricerca in Ingegneria elettronica, telecomunicazioni e tecnologie dell'informazione, 37 Ciclo. DOI 10.48676/unibo/amsdottorato/12227.
Documenti full-text disponibili:
[thumbnail of amorosa_lorenzo_mario_tesi.pdf] Documento PDF (English) - Richiede un lettore di PDF come Xpdf o Adobe Acrobat Reader
Disponibile con Licenza: Salvo eventuali più ampie autorizzazioni dell'autore, la tesi può essere liberamente consultata e può essere effettuato il salvataggio e la stampa di una copia per fini strettamente personali di studio, di ricerca e di insegnamento, con espresso divieto di qualunque utilizzo direttamente o indirettamente commerciale. Ogni altro diritto sul materiale è riservato.
Download (15MB)

Abstract

In recent years, distributed and ubiquitous intelligence has become a central force driving groundbreaking advances in wireless networks. Enabled by emerging autonomy, next-generation wireless networks are set to undergo substantial evolution. The main objective for the future consists in designing and creating cohesive communication and learning frameworks to achieve intelligent and autonomous wireless networks, pursuing the ambitious goal of achieving human-out-of-the-loop artificial intelligence (AI). Meeting this challenge signifies a critical transformation toward AI-native wireless networks that can dynamically self-adapt to complex scenarios. Addressing these future challenges requires a comprehensive integration of extensive knowledge in fields such as wireless communication and deep learning, creating opportunities for data-driven wireless network design, management, and optimization. This thesis investigates promising areas in wireless communications where AI paves the way for the achievement of intelligent and autonomous wireless networks. The first chapter introduces adaptive optimization in wireless communication networks and addresses the goal of learning effective and robust radio resource management strategies in complex scenarios. The second chapter explores the impact of generative AI in next-generation wireless networks, focusing on reliable and uncertainty-aware data generation processes enabled by approximate Bayesian learning. It illustrates how generative AI can represent an effective means of aiding data-driven algorithms in generalization and reducing the need for costly data collection. The third chapter then delves into distributed learning over wireless networks for radio resource management, which has the potential to meet the scalability demands of modern data-driven applications. This chapter emphasizes the importance of incorporating graph structures as an efficient way to introduce a relational inductive bias into the learning process, thus enhancing system performance across various metrics. Finally, the last chapter presents a holistic performance analysis of AI-native applications in 5G industrial internet-of-things networks, particularly within safety-critical scenarios, highlighting practical considerations on AI-native wireless network architectures.

Abstract
Tipologia del documento
Tesi di dottorato
Autore
Amorosa, Lorenzo Mario
Supervisore
Co-supervisore
Dottorato di ricerca
Ciclo
37
Coordinatore
Settore disciplinare
Settore concorsuale
Parole chiave
5G New Radio, Artificial Intelligence, Bayesian Learning, Deep Learning, Graph Neural Network, Reinforcement Learning, Industrial Internet of Things, Machine Learning, Multi-Agent System, Radio Access Network, Wireless Network
DOI
10.48676/unibo/amsdottorato/12227
Data di discussione
4 Aprile 2025
URI

Altri metadati

Statistica sui download

Gestione del documento: Visualizza la tesi

^