Date of Award

January 2023

Document Type

Thesis

Degree Name

Master of Science (MS)

Department

Computer Science

First Advisor

Prakash Rangathan

Abstract

Global Positioning Systems (GPS) are critical for accurate navigation in aviation networks.GPS information could be subjected to several interferences due to challenging terrain, rural areas, high-rise buildings, and intentional or unintentional jamming, spoofing, or unknown causes. Recently, the frequency of GPS/Automatic Dependant Surveillance-Broadcast (ADS-B)-related interferences in commercial airlines has increased. This thesis focuses on enhancing civil aviation safety using machine learning (ML) algorithms, particularly focusing on vulnerabilities in the ADS-B and its role in improving the National Air Space (NAS) surveillance and safety. Through scholarly article surveys, foundational issues in the ADS-B are identified. A study centered around a GPS interference event at Dallas Fort Worth (DFW) Airport, from which, key indicators like Navigation Integrity Criteria (NIC) were evaluated to detect GPS interferences. Several ML models, including neural networks, random forest, logistic regression, decision trees, Naïve Bayesian, support vector machines, and stochastic gradient descent classifiers, using a simulated dataset of 180,000 points, with 25,792 modeled as GPS interference instances are trained and tested in a rigorous training methodology. This model accurately captured anomalies, with the multi-layer perceptron (MLP) achieving a True Positive Rate (TPR) of 99.8\%. Additionally, the thesis presents data-driven approaches for advancing the state of the art in trajectory prediction. Employing the K-nearest Neighbors Regressor (kNN), a simple, non-parametric, and scale-independent model, led to a 40\% improvement in trajectory prediction quality and attained 94m accuracy for 2D position estimation. Such improvements in accuracy are critical for optimal trajectory estimations in GPS-denied environments.

Download

Share

COinS