Electrical and Computer Engineering ETDs

Publication Date

Fall 11-15-2019


The goal of this thesis is the detection and tracking of a ground vehicle, in particular a car-like robot, by a quadrotor. The first challenge to address in any pursuit or tracking scenario is the detection and unique identification of the target. From this first challenge, comes the need to precisely localize the target in a coordinate system that is common to the tracking and tracked vehicles. In most real-life scenarios, the tracked vehicle does not directly communicate information such as its position to the tracking one. From this fact, arises a non-cooperative constraint problem. The autonomous tracking aspect of the mission requires, for both the aerial and ground vehicles, robust pose estimation during the mission. The primary and crucial functions to achieve autonomous behaviors are control and navigation. The principal-agent being the quadrotor, this thesis explains in detail the derivation and analysis of the equations of motion that govern its natural behavior along with the control methods that permit to achieve desired performances. The analysis of these equations reveals a naturally unstable system, subject to non-linearities. Therefore, we explored three different control methods capable of guaranteeing stability while mitigating non-linearities. The first two control methods operate in the linear region and consist of the intuitive Proportional Integrate Derivative controller (PID). The second linear control strategy is represented by an optimal controller that is the Linear Quadratic Regulator controller (LQR). The last and final control method is a nonlinear controller designed from the Sliding Mode Control Theory. In addition to the in-depth analysis, we provide assets and limitations of each control method. In order to achieve the tracking mission, we address the detection and localization problems using respectively visual servoing and frame transform techniques. The pose estimation challenge for the aerial robot is cleared up using Kalman Filtering estimation methods that are also explored in depth. The same estimation method is used to mitigate the ground vehicle’s real-time pose estimation and tracking problem. Analysis results are illustrated using Matlab. A simulation and a real implementation using the Robot Operating System are used to support the obtained results.


Quadrotor, matlab, ROS, Kalman filter, Autonomous, Control

Document Type




Degree Name

Electrical Engineering

Level of Degree


Department Name

Electrical and Computer Engineering

First Committee Member (Chair)

Rafael Fierro

Second Committee Member

Balu Santhanam

Third Committee Member

Fernando Moreu

Fourth Committee Member

Christopher D.Petersen