Can robots learn through observation? This is the question an interdisciplinary project team at the University of Freiburg is asking itself and researching new training methods for AI-based robots.
Innovative machine learning methods should enable robots to learn tasks demonstrated by humans. Similar to human learning, the robots should be able to acquire the correct behaviour by copying observed behaviour and through interactions. In this way, the robots should be able to act in different environments without having to be reprogrammed for each situation.