Abstract :
[en] This dissertation examines integrated service delivery within multi-tier Non-Terrestrial Networks (NTNs) comprising Low Earth Orbit (LEO), Medium Earth Orbit (MEO), and Geostationary Earth Orbit (GEO) satellite constellations. Its objective is to develop models, architectural designs, and resource management algorithms aimed at minimizing ground resource usage while optimizing coverage, beam patterns, power, and bandwidth. The goal is to enable effective resource allocation that meets the dynamic traffic demands and service-level agreement (SLA) requirements of subscribers.
In Chapter 1, an overview of multi-tier NTNs is provided, covering research motivations and methodology. In Chapter 2, the impact of orbital variations on Doppler shift and packet arrival time is evaluated. A resource management technique utilizing dual connectivity (DC) between MEO and GEO satellites is developed to optimize capacity.
Similarly, Chapter 3 presents a study on user equipment (UE) RF design for integrated service delivery in NTNs and examines multi-connectivity (MC) as a resource optimization strategy, where UEs can connect to multiple satellites to achieve higher peak throughput. Originally developed by the 3rd Generation Partnership Project (3GPP) for terrestrial communications in 4G and 5G, MC has shown significant gains in the terrestrial domain; this chapter investigates its potential in satellite communications. Although MC can increase throughput, it is limited by the UE's RF configuration. To address this, a terminal-aware multi-connectivity (TAMC) scheduling algorithm was developed, using available radio resources and propagation data to define an adaptive resource allocation pattern that optimizes uplink data rates while minimizing UE energy consumption. The algorithm operates with a multi-layer NTN resource scheduling architecture, incorporating a softwarized network-layer dispatcher that classifies and differentiates packets based on terminal types, such as IoT and VSAT, and outperforms benchmark algorithms.
Further, Chapter 4 extends this study to uplink transmissions, analyzing UE traffic demands across different traffic classes, including eMBB, URLLC, and mMTC. This chapter also considers user service classifications to enable network dimensioning that meets quality of service (QoS) requirements. A resource management architecture is proposed for multi-tier NTNs, adapted to the 3GPP protocol stack. To enhance energy efficiency in uplink communications, an energy-efficient service-aware multi-connectivity (EE-SAMC) scheduling algorithm was developed. EE-SAMC uses available radio resources and propagation data to intelligently define a dynamic resource allocation pattern that reduces UE energy consumption while maximizing QoS. The algorithm is based on a non-convex combinatorial problem solved through two approaches: an optimization solution and a heuristic approach. In the optimization approach, the problem is divided into two subproblems—(i) joint route and power allocation and (ii) path matching—solved using the interior point algorithm and the Hungarian algorithm, respectively.
Multi-tier NTNs are expected to be a key enabler for 6th-generation (6G) systems, providing ubiquitous coverage. However, effectively managing heterogeneous networks to meet dynamic traffic demands in time-varying environments remains a challenge. Chapter 5 addresses this by exploring dynamic beam and resource allocation techniques to improve the capacity of multi-tier NTNs over stochastic channels in the down-link, meeting the diverse service level agreements (SLAs) of users. Here, a non-convex combinatorial optimization problem with inequality constraints is formulated, which is then separated into two subproblems: (a) dynamic beam allocation and (b) joint power and bandwidth allocation. The dynamic beam allocation subproblem is solved using an iterative algorithm, while joint power and bandwidth allocation employs a multi-agent deep reinforcement learning (MADRL)-aided resource allocation algorithm. This solution leverages MC to maximize capacity and operates within a network architecture with a hybrid gateway station (HGS) that manages satellites and supports various waveforms, including 5G New Radio (NR) and DVB-S2X. The algorithm determines resource allocation patterns based on channel quality indicators (CQI) and traffic classes, such as URLLC, HDTV, and eMBB.
To address the issue on efficient utilization of spectrum resources, operators are in search of innovative ways to dimension network resources and prevent resource under utilization, especially in a mult-tier NTN. Chapter 6 introduces a novel service delivery model where infrastructure providers (InPs) lease NTN resources as slices to mobile virtual service operators (MVSOs). These MVSOs then offer the leased resources to subscribers, facilitating efficient use of NTN resources within the telecommunications ecosystem. The model incorporates an NTN slicing architecture with multi-layer satellites, including LEO, MEO and GEO constellations, featuring a HGS tailored to the virtualization architecture specified by 3GPP. In this context, a multi-objective optimization problem (MOOP) is formulated, comprising two combinatorial objectives for InPs and MVSOs aimed at maximizing revenue. The proposed algorithm addresses joint network slicing and admission control (AC) using techniques such as the non-dominated sorting genetic algorithm II (NSGA-II), multi-objective reinforcement learning (MORL), and a heuristic approach. It further enhances the AC mechanism by leveraging an LSTM-based deep learning model to predict traffic demand for URLLC and eMBB users, preventing SLA violations.
Finally, Chapter 7 outlines conclusions and directions for future research.
