ETD

• 6G networks
• low earth orbit satellites
• networked music performance
• non-terrestrial networks
• quality of experience
• reinforcement learning
• software-defined wide area networks

This doctoral thesis investigates the transition toward resilient, data-driven 6G networks to bridge the Digital Divide. Following a comprehensive survey of Software-Defined Wide Area Networks (SD-WAN), we present simulation, emulation, and testbed environments integrating single- and multi-agent reinforcement learning techniques.
The framework is then extended to address a cornerstone of 6G: the integration of Non-Terrestrial and Terrestrial Networks. By leveraging SD-WAN, we propose a novel architecture enabling the seamless integration of Low Earth Orbit (LEO) satellite networks.
Driven by the integration requirements identified in the first part of the thesis, we conducted measurement campaigns on the Starlink network across Europe and California. These trials include in-flight connectivity studies and stress tests during solar storms. Additionally, a real-world deployment for wildlife monitoring in San Rossore Park proves the viability of LEO-backhauled Edge AI in off-grid scenarios.
To investigate the compatibility of LEO networks with the most stringent real-time multimedia requirements, we selected Networked Music Performance (NMP) as a critical use case. In collaboration with Stanford University, we conducted measurement and live jamming sessions using JackTrip over Starlink, highlighting the human-centric dimension of QoE.
The thesis concludes with insights into Quantum and Semantic networking, outlining a design framework for reliable, human-aware 6G connectivity.