Mobility Management and Localizability for Cellular Connected UAVs

Sammanfattning: Unmanned Aerial Vehicles (UAVs) connected to cellular networks present novel challenges and opportunities in mobility management and localization, distinct from those faced by terrestrial users. This thesis presents an integrated approach, combining two key aspects essential for the integration of UAVs with cellular networks.Firstly, it introduces the mobility management challenges for cellular-connected UAVs, which differ significantly from terrestrial users. While terrestrial mobility management primarily aims to prevent radio link failures near cell boundaries, aerial users experience fragmented and overlapping coverage with line-of-sight conditions involving multiple ground base stations (BSs). Thus, mobility management for UAVs extends beyond link failure avoidance, aiming to minimize unnecessary handovers while ensuring extended service availability, particularly in up-link communication. Line-of-sight conditions from a UAV to multiple BSs increase the likelihood of frequent handovers, resulting in control packet overheads and communication delays. This thesis proposes two approaches to address these challenges: 1) A model-based service availability-aware Mobility Robustness Optimization (MRO) adapting handover parameters to maintain high service availability with minimal handovers, and 2) A model-free approach using Deep Q-networks to decrease unnecessary handovers while preserving high service availability. Simulation results demonstrate that both the proposed algorithms converge promptly and increase the service availability by more than 40 %  while the number of handovers is reduced by more than 50%  as compared to traditional approaches.Secondly, to assess the ability of a network to support the range-based localization for cellular-connected UAVs, an analytical framework is introduced. The metric B-localizability is defined as the probability of successfully receiving localization signals above a specified Signal-to-Interference plus Noise Ratio (SINR) threshold from at least B ground BSs. The framework, accounting for UAV-related parameters in a three-dimensional environment, provides comprehensive insights into factors influencing localizability, such as distance distributions, path loss, interference, and received SINR. Simulation studies explore the correlation between localizability and the number of participating BSs, SINR requirements, air-to-ground channel characteristics, and network coordination. Additionally, an optimization problem is formulated to maximize localizability, investigating the impact of UAV altitude across different scenarios. Our study reveals that in an urban macro environment, the effectiveness of cellular network-based localization increases with altitude, with localizability reaching 100% above 60 meters. This finding indicates that utilizing cellular networks for UAV localization is a viable option.