mso-ansi-language:EN-US">This thesis addresses the optimisation of integrated Terrestrial and Non-Terrestrial Networks (TN-NTNs) in the context of 5G and beyond. As user demand for seamless, high-capacity connectivity intensifies, conventional terrestrial networks face limitations in coverage—particularly in remote or underserved regions. Non-Terrestrial Networks, including low-earth orbit (LEO) satellites, offer a promising complement by extending coverage and providing support for the terrestrial network. This work proposes and analyses three major contributions for the large-scale optimisation of TN-NTNs.
mso-ansi-language:EN-US">First, a utility-based optimisation framework is developed for high-traffic scenarios, enhancing user association and resource allocation by dynamically adjusting power levels and bandwidth sharing between terrestrial and satellite tiers. Then, the BLASTER algorithm is introduced, enabling traffic-aware management by optimising user association, macro base station activation, and bandwidth allocation throughout the day. This method significantly reduces energy consumption while enhancing sum log-throughput (SLT). Finally, a decentralised online learning framework based on the Bandit-feedback Constrained Online Mirror Descent (BCOMD) algorithm is presented. This distributed solution adapts to stochastic traffic conditions, ensuring Quality of Service (QoS) while reducing energy consumption without requiring centralised control.
mso-ansi-language:EN-US">Through theoretical analysis and extensive simulations, this thesis demonstrates how integrated TN-NTNs can dynamically adapt to varying traffic loads to optimise network performance, reduce energy usage, and provide robust connectivity across diverse environments.
