Within the RobVision project a climbing robot has been built that is navigated into a ship structure. The robot is equipped with a head for monocular and stereo vision. From the CAD-model of the ship view points are selected such that the head can look at directions with sufficient features. Features for the views are extracted automatically. The position of the robot is estimated from the features detected by two vision approaches, one searching in the full stereo images for junctions and measuring their 3D position, the second tracking monocular 2D edge features. The multitude of features renders position estimation reliable for locating the robot within a few centimetres. Gyroscopes are used to stabilise the head while the robot moves. The system developed in the RobVision project is able to navigate a climbing robot into ship sections using visual feedback. The approach is flexible, because it can operate in any environment that has distinct features and that has a CAD model. The robot is able to carry a work package of 50 kg into the cell. Navigation is fully autonomous and 3D. The accuracy of the pose estimation of the demonstrator has been 5 to 35 mm. The accuracy does not depend on the path length, but on the distance to features at the final location. Image processing is executed in 120 ms (and can be improved to 40 ms), which allows a robot speed of 25 cm/s. This technique opens up other potential applications. The model-based approach enables to measure any modelled object and to feed back the measurement data directly into the CAD system. The impact of this capability is manifest. Ship builders and the construction industry can supervise the production quality and consequently reduce production time.
A system to navigate a robot into a ship structure
Navigationssystem eines Kletterroboters in einem Schiffskörper
2001
8 Seiten, 9 Bilder, 9 Quellen
Aufsatz (Konferenz)
Englisch
A system to navigate a robot into a ship structure
British Library Online Contents | 2003
|SLUB | 1953
|