Orientation estimation fusing a downward looking camera and inertial sensors for a hovering UAV

Abstract

Nowadays, a particular kind of UAVs (Unmanned Aerial Vehicle) known as quadcopters or quadrotors have become very popular. This is mainly due to their reduced size and high maneuverability which allow them to operate in indoor environments. The requirement for orientation estimation in these vehicles is twofold: for low level stability control, and for high level navigation and motion planning. Orientation estimation is usually carried out fusing measurements of different sensors including inertial sensor, magnetic compass, sonar, GPS, camera, etc. As is known, GPS signal is not available in indoor environments and the Earth's magnetic field is highly disturbed by ferromagnetic structures. In the present work we describe a new approach for quadrotor orientation estimation fusing inertial measurements with a downward looking camera. Inertial sensor are used for orientation estimation based on the gravity vector, and the camera provides information related to the heading direction. The camera heading or yaw angle estimation is based on spectral features extracted from the floor plane. Experimental results show the performance of the presented approach applied to a hovering UAV.