A Vision-based Navigation Algorithm for a VTOL-UAV

Abstract

This paper addresses the implementation and evaluation of a vision-based navigation filter for the Unmanned Aerial Vehicle ARTIS (Autonomous Rotorcraft Testbed for Intelligent Systems). The navigation filter uses vision data from ground feature tracking based on a Lucas-Kanade algorithm in order to compensate GPS-failure. An Extended Kalman Filter (EKF) is used for sensor data fusion. The developed algorithm handles data synchronisation and latency compensation. The filter is evaluated in Software in the Loop (SITL) simulation, in Hardware in the Loop (HITL) simulation and in flight test. Despite the inherent error accumulation the relative navigation approach allows the helicopter a notable area of operation. ARTIS performs stable flight in hover domain using the vision-based navigation approach. The results show the capability of the navigation algorithm to converge and to compensate GPS-failure.