Information Technology Reference
In-Depth Information
Autonomous Landing of AR.Drone
Roman Barták, Andrej Hraško, and David Obdržálek
Charles University in Prague, Faculty of Mathematics and Physics, Praha,
Czech Republic
{roman.bartak,david.obdrzalek}@mff.cuni.cz
,
andrej@hrasko.eu
Abstract.
Inexpensive robotic toys are becoming widely available and though
they are called toys, they provide gradually better sensing features and rising
computational power. Adding new autonomous functions, such as autonomous
landing for flying drones, helps the users with routine maneuvers and so allows
using them with lower risk of breaks caused by loss of control in critical opera-
tional phases. In this paper we present software for autonomous landing of an
AR.Drone - an affordable robotic quadricopter. This software uses an easy-to-
customize landing pattern and exploits only the sensors available on the drone.
The whole landing process is simplified to a “push-button” approach for the end
user, allowing for safer overall operation.
1
Introduction
AR.Drone by Parrot Inc. is a high-tech yet affordable flying toy - a quadricopter with
several sensors including two cameras, accelerometer, gyroscope, sonar, and a built-in
controller for basic stabilization. Pre-processed information from sensors can also be
received via Wi-Fi so the computer can serve as a more powerful external brain for
the drone. For controlling the drone from a computer and for visualizing information
from sensors, several software applications exist. We use
Control Tower
[6] as the
basic computer interface to AR.Drone and we extended its functionally with auto-
nomous landing to a user-customizable pattern. The idea is that the user flies the
drone above the pattern and by pushing a button in the software GUI the drone auto-
nomously lands on that pattern. Such functionality is great for beginner users who can
control the drone in free space but are not able to precisely land or are afraid of con-
trolling it low above the ground where mistakes in control or turbulences caused by
airflow could easily cause crashing and then damaging of the drone.
2
Landing Description
For autonomous landing, the drone must reliably detect the landing point. There exist
methods for visual navigation (e.g. using an H-shaped landing pattern [5] or a specific
circular pattern [3]), however they require better visual sensors (cameras) and they
usually do not support horizontal orientation.