Today’s robots have difficulty understanding unstructured, human-oriented environments. Most current robotic systems rely heavily on the explicit specification of every eventuality they might encounter while executing their tasks. Each response to a contingency has to be programmed in advance.

The worst thing about all this, is that the next time another robot has to perform a similar task in a similar environment, the same programming steps have to be repeated all over again. The lack of memorization and learning prevents robots from improving efficiently upon their sensing and action capabilities over time. This approach faces serious limitations, because the real world is generally too nuanced, too complicated and too unpredictable to be summarized within a limited set of specifications. There will inevitably be novel situations and robotic systems will always have gaps, conflicts or ambiguities in their knowledge and task representations. A solution to the above problems is to store a robot’s knowledge of the environment, and the descriptions needed to perform its task in a global world-wide web-style accessible database. If thatparticular robot or any other robot is to perform a similar action in a more or less similar environment, they can access the stored data and immediately commence with their task.

This is exactly what the RoboEarth project is about: sharing knowledge between robots all over the world by building up huge knowledge bases on objects, maps, task descriptions, domain ontologies and any form of information a robot can use while executing its task.

This reusable knowledge will provide a powerful feed forward to any robot’s 3D sensing, acting and learning capabilities, hereby increasing efficiency, inter-operability and conceptual understanding by robots of our complex, daily human environment.