Quadcopter stabilization using visual marker-based tracking

Jurij Slabanja (2014) Quadcopter stabilization using visual marker-based tracking. EngD thesis.

Preview

PDF Download (8Mb)

Abstract

In this thesis we implemented a system, which Parrot AR.Drone can use to track a moving target. The system is based on ROS. We used an iRobot Roomba as the target, put a marker on top of it, and drove it around. The drone's goal was to stay as directly as possible above the target. The drone tracked the marker using its front camera. Its movement was controlled by PID controllers. We conducted three experiments to evaluate the system. First we evaluated how accurate our method of determining the drone's height was, compared to its default method. Then how accurately and robustly the drone can hover over a stationary target. Lastly, we evaluated the accuracy and robustness of the drone hovering over a moving target. The results show that our system allows the drone to hover much more reliably over the marker, compared to its default stabilization methods. The system also provides a better means of estimating the distance of the drone from the marker, compared to the built-in sonar.

Actions (login required)