Robots worldwide will learn from each other

Published on: 13 January, 2010

Eindhoven University of Technology, together with six other European research institutes, is to develop a world wide web for robots. Through this database, robots in both healthcare and industry will be able to learn from each others’ capabilities, and will master new movements much faster.

“Right now robots still have to reinvent the wheel each time they are performing a simple task,” says René van de Molengraft, associate professor of mechanical engineering at the Eindhoven University of Technology (TU/e). “In Project RoboEarth we are developing a kind of world wide web for robots. This enables them to store every action they have learned. Thanks to this, other robots in the world don’t have to learn these actions themselves anymore. This will make it much faster for robots to master new actions.” The project is scheduled to start this week, with a budget of almost six million euros.

One of the greatest challenges of the 21st century is developing robots that can move around in the human world, performing tasks that are beneficial to society. While people may think that today’s robots are pretty autonomous, the reality is that they are not as independent as we would hope. Every single task a robot performs has been programmed into its memory beforehand. And while performing, the robot can only rely on its own observations in order to do the task as perfectly as possible. The downside is, when performing the exact same task at a different location, the robot has to start all over again: making an image of its surroundings, then coming up with a plan on how to finish the task to satisfaction. There is no worldwide, shared memory that robots can draw from.

This is about to change with the RoboEarth research project. Led by TU/e, six European research institutes are developing a system that will enable robots in healthcare and industry to execute tasks that were not anticipated at the time of robots’ construction. Robots will take this new knowledge from a worldwide database, also named RoboEarth, that contains other robots’ experiences.

Robots will start to learn from each other and will be able to adapt much faster to new environments, and to execute complex tasks there. The robots communicate with the database through a wireless internet connection. This means enormous quantities of useful information will be opened up for all robots and that they will not have to reinvent their skills again and again. Industrial companies will be able to expand the knowledge in RoboEarth or even use it for totally new applications. RoboEarth is expected to accelerate innovation and will hopefully lead to a higher quality in robotic applications.

The other participants in the project are the Universität Stuttgart, the ETH Zürich, the Universidad de Zaragoza, the Technische Universität München and Philips Applied Technologies. The team is planning to make six robots, to demonstrate the use of RoboEarth.