Date of Defense
5-12-2024 11:00 AM
Location
F1 - 1077
Document Type
Thesis Defense
Degree Name
Master of Science in Electrical Engineering (MSEE)
College
College of Engineering
Department
Electrical Engineering
First Advisor
Prof. Atef Abdrabou
Keywords
Throughput, 5G technology, network slicing, dynamic resource allocation, UAV, reinforcement learning, numerology.
Abstract
In response to the surging demand for cellular services, driven partly by the post-COVID-19 era, network operators are grappling with the challenge of enhancing network capacity and throughput. Fifth-generation mobile technology and beyond (6G) with techniques like network slicing are emerging as promising solutions for faster data speeds and improved network performance across various devices. In network slicing, distributing radio resources between different virtual operators (slices) is often performed in a static manner. Therefore, due to the dynamic user demand, the users of one virtual operator could suffer from a lack of resources, whereas another virtual operator with overlapping cells is underutilized. The research of this thesis addresses the dynamic resource allocation between the two virtual operators (slices) with the aid of aerial base stations (UAVs). The proposed algorithm harnesses reinforcement learning to find the optimal placement of the aerial base station to maximize the network users’ throughput. This enables UAVs to dynamically manage radio resources by serving the users of the underutilized slice while offloading some of the users of the heavily utilized slice to the ground cells of the underutilized slice. Comprehensive tests, using both single-UAV and dual-agent UAVs on a grid of overlapping cells, evaluated average user throughput, the percentage of satisfied users (with improved throughput), and throughput enhancement for satisfied users. Different offloading strategies are examined based on the number of offloaded users, their locations, and the capability of their user equipment (numerology). One of the main findings of this research is that offloading a certain percentage of the users in a heavily utilized slice cell based on their locations leads to a significantly improved user throughput.
Included in
DYNAMIC RESOURCE ALLOCATION FOR BEYOND 5G NETWORKS THROUGH UAV-ASSISTED COMMUNICATION
F1 - 1077
In response to the surging demand for cellular services, driven partly by the post-COVID-19 era, network operators are grappling with the challenge of enhancing network capacity and throughput. Fifth-generation mobile technology and beyond (6G) with techniques like network slicing are emerging as promising solutions for faster data speeds and improved network performance across various devices. In network slicing, distributing radio resources between different virtual operators (slices) is often performed in a static manner. Therefore, due to the dynamic user demand, the users of one virtual operator could suffer from a lack of resources, whereas another virtual operator with overlapping cells is underutilized. The research of this thesis addresses the dynamic resource allocation between the two virtual operators (slices) with the aid of aerial base stations (UAVs). The proposed algorithm harnesses reinforcement learning to find the optimal placement of the aerial base station to maximize the network users’ throughput. This enables UAVs to dynamically manage radio resources by serving the users of the underutilized slice while offloading some of the users of the heavily utilized slice to the ground cells of the underutilized slice. Comprehensive tests, using both single-UAV and dual-agent UAVs on a grid of overlapping cells, evaluated average user throughput, the percentage of satisfied users (with improved throughput), and throughput enhancement for satisfied users. Different offloading strategies are examined based on the number of offloaded users, their locations, and the capability of their user equipment (numerology). One of the main findings of this research is that offloading a certain percentage of the users in a heavily utilized slice cell based on their locations leads to a significantly improved user throughput.