Future service robots, that shall one day serve in households, need background knowledge from various areas to be able to act efficiently and useful in a human-centered environment. Knowledge about objects that can be manipulated, people that shall be interacted with and knowledge about tasks and their execution is mandatory to realise a useful and autonomous service robot.
The research group "Interactive Learning" investigates ways how this knowledge can be acquired, be learned and be transferred onto a robotic system. The connecting idea is that it won't be feasible to manually model all of the knowledge that real life scenarios require. Instead, the acquisition of the knowledge should be as far as possible conducted autonomously by the system itself by various means of observation. In this approach, the human then naturally acts as a supervisor, guide and archetype for the system.
Sven R. Schmidt-Rohr