Lilian de Greef

Accessible Aerial Autonomy

Nick Berezny, Lilian de Greef, Bradley Jensen, Kimberly Sheely, Malen Sok, David Lingenbrink, Zachary Dodds


This work presents a combination of software and hardware that makes aerial autonomy substantially more accessible both in terms of programmatic complexity and in terms of cost. We use the ARDrone quadrotor helicopter and Willow Garage's Robot Operating System software infrastructure to demonstrate several autonomous tasks. Using vision as the sole aerial sensor, we demonstrate point-to-point navigation, aerial support of a ground robots, and robot localization within image-based maps. In contrasting several variations of SURF- feature matching, we demonstrate that low-cost aerial platforms can support robust, landmark-free visual spatial reasoning. This evaluation shows that aerial platforms can be practical, cost- and time-effective components of task-performing systems. We argue that aerial autonomy should be considered a broadly accessible resource, within reach of any investigator or educator of AI robotics.