AI-driven 5G networks for autonomous positioning system platform

Abstract

Unmanned Aerial Vehicles (UAVs) are becoming essential for various urban applications, such as surveillance, delivery, logistics, disaster management, and traffic monitoring. However, their positioning performance in urban environments can be limited due to challenges such as non-line-of-sight (NLOS) propagation, multipath interference, and signal blockage caused by tall buildings, trees, and other obstacles. These factors lead to reduced positioning accuracy and unreliable communication. To address these issues, this thesis introduces three key and novel contributions. First, it presents one of the first real-world evaluations of the 5G network performance for UAV operations at altitudes between 50 and 110 meters, using XCAL-based field trials. This provides new insights into the altitude-dependent Quality of Service (QoS) parameters such as latency, throughput, and handover (HO) efficiency and provides practical recommendations for UAV-specific connectivity protocols. Second, a novel hybrid positioning framework is proposed that integrates the observed time difference of arrival (OTDOA) of the new 5G radio (NR) with the fusion of sensor and barometric pressure sensor through an Extended Kalman Filter (EK). This combination significantly improves positioning accuracy (2.8–7 m) in GNSS GNSS-challenged urban environment, which has not been demonstrated in prior UAV studies. Third, the thesis introduces a lightweight feedforward neural network (FNN) for mitigating NLOS errors in 5G-based UAV positioning. Trained on simulated MATLAB data, the model corrects time-of-arrival (TOA) measurements in real time, reducing positioning error to 1.3 m in LOS and 1.7 m in NLOS, outperforming conventional methods. Unlike existing solutions, this model is designed for real-time deployment on UAV platforms with limited resources. Overall, this research strengthens UAV navigation and connectivity in urban airspace by combining 5G advancements, sensor fusion, and AI-powered error correction. The novelty lies in the integration of real-world 5G performance analysis, a hybrid OTDOA sensor fusion framework, and an AI-based NLOS correction model into a unified solution for reliable, accurate, and scalable Urban Air Mobility (UAM), opening the door to future improvements in AI-driven 5G networks for autonomous system platforms.