Boosting Development of Service-Robots:

RoboEarth, a World Wide Web for Robots

RoboEarth enables robots to learn from each other

06.04.2010, News

A world wide web for robots – this is the aim of a joint initiative of six European research institutes, among them Technische Universitaet Muenchen (TUM). The worldwide database, called RoboEarth, will enable robots in health care and industry to learn from each other’s capabilities, thus speeding up the process of adopting new operations. The project funded by the European Commission will take 4 years and has a €5.6m budget.

One of the challenges of the 21st century is to develop robots capable of moving around in the human world, carrying out socially useful tasks. But although robots are already famous for their skills in performing tasks fairly independently, this independence is often still disappointing. Each task performed by the robot has been programmed in advance by humans, but while executing it, the robot relies entirely on its own monitoring of the process to optimize the task. Worse still, whenever the robot has to perform the same task again at a different location, it has to start all over again: it has to visualize the surroundings and decide how to complete the task successfully. There is no collective worldwide memory for robots to draw from.

This situation is about to change, with the RoboEarth research project. Six European research institutes, among them the Technische Universitaet Muenchen, will be developing a system allowing robots to perform tasks in, for example, homecare or healthcare situations that had not been planned when the robot was constructed. Robots will be able to retrieve this new knowledge from a worldwide database dubbed RoboEarth, in which the experiences of other robots are stored. This will allow robots to learn from each other, and adapt much faster to new surroundings. Companies will be able to extend the knowledge stored in RoboEarth or use it for completely new applications. Hence, RoboEarth will speed up the innovation process, and this is expected to result in higher quality robot applications.

The project will be led by Eindhoven University of Technology (TU/e). Other participants besides TU Muenchen will be Universitaet Stuttgart, ETH Zürich, Universidad de Zaragoza, and Philips Applied Technologies. The team will be making six demo versions that will demonstrate the use of RoboEarth, including a robot capable of offering beverages to hospital patients, and a system that will show how the knowledge acquired by robot A improves the performance of robot B.

Contact

Dr. Reinhard Lafrenz
Technische Universitaet Muenchen
Institut für Informatik
Boltzmannstraße 3, 85748 Garching, Germany
Tel.: +49 89 289-18136
E-Mail

Kontakt: presse@tum.de

More Information