Publication at Faculty of Mathematics and Physics |
2015
Abstract
This paper describes implementation of the system controlling a flying drone to stabilize and hold the drone still regardless of external influences and inaccuracy of sensors. This task is achieved by utilizing visual monocular SLAM (Simultaneous Localization and Mapping) - tracking recognizable points in the camera image while maintaining a 3D map of those points.
The output location is afterwards combined using the Kalman filter with odometry data to predict future location using drone's dynamics model. The resulting location is used afterwards for reactive control of drone's flight.