Date of Award


Document Type


Degree Name

Master of Science in Electrical and Computer Engineering (MSECE)

First Advisor

Dr. Abdollah Homaifar


An American scientific writer/critic of the mid-20th century, Joseph Wood Krutch, stated, “Technology made large populations possible: large populations now make technology indispensable.” Unmanned aerial vehicles (UAVs) have become a vital part of society, evolving from solely military use to commercial and even personal day-to-day use. UAV application has grown for mobility on demand, increasing the number of UAVs in the sky. Consequently, more UAVs increases the possibility of aerial collisions, challenging the safety of passengers flying and bystanders on the ground. Understanding the behavior of UAVs and unmanned aerial systems (UAS) in general, therefore, will decrease the possibility of aerial collisions. This thesis focuses on perception of UAVs; comparing how well machine-learning (ML) algorithms can analyze and predict their current states through supervised learning approaches. In this thesis, a data-driven software tool was developed to create a fundamental approach for UAV performance inspection. The developed algorithm was used for UAV behavioral analysis and the results provided better accuracy for predicting UAVs current state. This procedure includes using a multi-class classification for a single testing scenario, running the simulation environment with one Ar.Drones quadcopter for data gathering, hardware implementation for real-world implementation, and using Robot Operating System (ROS) as a middleware. All software development and implementation were conducted in Python programming language due to its high compatibility and robustness within a ROS development environment. The scenario stores the velocity readings on x, y, z directions, the altitude, and the corresponding state labels in matrix form. The procedure breaks the samples into training and testing for application of proposed supervised learning algorithms to predict output states of the system. Furthermore, these predictions are evaluated and analyzed, where results were compared within different ML approaches.