Our mission is to make robots learn in a developmental fashion - similar to children. Why should that be useful? Already now robots are present in many areas of our life, taking over tasks that are too dangerous to repetitive, or require to high precision for humans. However, their area of application is bound because preprogrammed robots are not able to successfully interact with our complex and constantly changing world.
In living beings, this problem is solved by interplay of learning, self-organization and innate
information. Self-organization—ubiquitous in nature—offers promising perspectives for practical appli-
cations because it is based on local interactions and typically scales well to large fault tolerant systems.
In our research we were building the theoretical basis for self-organized robot control. We study how dy-
namical systems and information theory can be used to generate sensory-motor coordination in
In order to make robots successful they have not only to efficiently explore but also to learn from expe-
rience. In order to make this learning useful for unknown future tasks it is best to learns a model of the
body and its environment. Classical regression models are not good at predictions for new situations. I
investigated how a model can identify the underlying relationships in data and thus obtaining a model
that can extrapolate well . While diving into deep learning we realized a potential for speeding up
stochastic gradient descent.
Google scholar: scholar.google.de/citations?user=b-JF-UIAAAAJ