Modelling of UAV communications in integrated terrestrial and non-terrestrial networks

Abstract

This diploma thesis investigates enhancing Quality of Experience (QoE), especially in areas where terrestrial networks cannot meet demands. The focus is on a hybrid architecture integrating Non-Terrestrial Networks (NTNs) with standard terrestrial networks. A vital aspect of this architecture is using Unmanned Aerial Vehicles (UAVs) to provide dynamic service in underserved areas. A machine learning-based methodology uses a hybrid clustering algorithm (DBSCAN + Mean Shift) to group users based on location and network conditions. This clustering facilitates the strategic allocation of users to the most appropriate network type, guided by signal quality considerations. The study further incorporates constrained K-means for initial UAV positioning and a Genetic Algorithm for optimizing these placements, aiming to boost network efficiency and reduce interference. Simulations using Python tools indicate that this hybrid approach improves network QoE compared to terrestrial network systems. The results from this study contribute to the development of more resilient and extensive network infrastructures, demonstrating the effectiveness of combining terrestrial and non-terrestrial network technologies.