{"id":10610,"date":"2024-12-16T17:31:09","date_gmt":"2024-12-16T16:31:09","guid":{"rendered":"https:\/\/www.uni.lu\/research-en\/?post_type=events&p=10610"},"modified":"2024-12-16T17:31:10","modified_gmt":"2024-12-16T16:31:10","slug":"doctoral-defence-michael-ninma-dazhi","status":"publish","type":"events","link":"https:\/\/www.uni.lu\/research-en\/events\/doctoral-defence-michael-ninma-dazhi\/","title":{"rendered":"Doctoral Defence: Michael Ninma DAZHI"},"content":{"rendered":"\n
\n

The Doctoral School in Science and Engineering<\/a> is happy to invite you to Michael Ninma DAZHI\u2019s defence entitled<\/p>\n\n\n\n

\nModeling, optimization and reconfiguration of hybrid satellite networks for integrated service delivery<\/h5>\n\n\n\n

Supervisor: Assist. Prof Bhavani Shankar MYSORE RAMA RAO<\/a><\/p>\n\n\n\n

Multi-orbital non-terrestrial network (NTN) comprising Low Earth Orbit (LEO), Medium Earth Orbit (MEO), and Geostationary Earth Orbit (GEO) constellations is emerging as a critical technology to enhance capacity and meet the growing demand for heterogeneous traffic across diverse use cases. This hybrid architecture will play a pivotal role in 5G-Advanced and 6G by enabling enhanced ubiquitous coverage. This thesis presents innovative architectures, models, algorithms, and strategies to design and implement such technology.<\/p>\n\n\n\n

First, we investigate the impact of orbital dynamics on resource allocation optimization, focusing on Doppler shifts and packet arrival time variations in MEO and GEO orbits. Specifically, we propose a dual connectivity (DC) algorithm to optimize both delay and throughput. Next, we examine the radio frequency (RF) design of user equipment (UE) in a multi-orbital NTN, presenting a link budget analysis alongside an algorithm that introduces multi-connectivity (MC) across orbits while accounting for the RF constraints of internet of things (IoT) and very small aperture terminal (VSAT) devices. Additionally, we evaluate resource management techniques within a novel multi-orbital NTN architecture aligned with the 3rd Generation Partnership Project (3GPP) 5G protocol stack. A non-convex combinatorial problem is formulated to maximize energy efficiency for UEs. To address this, we present a scheduling algorithm for uplink transmission from UEs handling different traffic types, including Ultra-Reliable Low-Latency Communication (URLLC), Massive Machine-Type Communication (mMTC), and Enhanced Mobile Broadband (eMBB). This algorithm leverages interior-point methods and the Hungarian algorithm for optimization.<\/p>\n\n\n\n

Another key contribution involves solving the resource allocation problem in a large-scale, dynamic multi-orbital NTN with time-varying characteristics. We introduce artificial intelligence (AI) into radio access network (RAN) resource management within an open RAN architecture using a radio intelligent controller (RIC). This system supports two waveforms: 5G NR and DVB-S2X. A non-convex combinatorial problem is formulated to maximize capacity in a stochastic Rician channel. We propose a novel algorithm that dynamically allocates beams, power, and bandwidth using iterative methods and a multi-agent deep reinforcement learning (MADRL) approach. The algorithm also incorporates a newly derived channel quality index (CQI) for DVB-S2X links, tailored for different user states.<\/p>\n\n\n\n

Finally, we explore the efficient utilization of NTN resources by developing a business model in which infrastructure providers (InPs) lease resources as slices to mobile virtual service operators (MVSOs). These MVSOs, in turn, offer the resources to subscribers, enabling more efficient use within the telecommunications ecosystem. The model leverages the network virtualization architecture defined by 3GPP. A multi-objective optimization problem (MOOP) is formulated with two combinatorial objectives aimed at maximizing revenue for both InPs and MVSOs. The proposed solution integrates joint network slicing and admission control (AC) techniques, utilizing the non-dominated sorting genetic algorithm II (NSGA-II), multi-objective reinforcement learning (MORL), and a heuristic approach. Additionally, a long short-term memory (LSTM)-based deep learning model is used to predict traffic demand for URLLC and eMBB users, helping to prevent SLA violations and improve the AC mechanism.<\/p>\n\n\n\n