Vision-based Hover Stabilization for a Micro Helicopter

Abstract

Penetrating urban regions or buildings with a micro helicopter is likely to fail because of missing satellite based navigation (e. g. GPS). At the DLR Institute of Flight Systems a vision system is developed for use on a micro helicopter. It could augment or even replace satellite based navigation to a certain extent. To estimate the motion and relative position of the helicopter a model is used comprised out of the projection of three dimensional features of the real world onto the image plane and the integration of accelerations. The necessary features are provided by a Lucas-Kanade tracker or through the recognition of an artificial pattern with known geometry. The framework of an extended Kalmanfilter is used to actually calculate the estimation of the model states. This paper describes the design of the vision system and the integration in the available Autonomous Rotorcraft Testbed for Intelligent Systems (ARTIS) operated by the DLR.