Dr.-Ing. John Nassour
Fakultät für Informatik
Professur Künstliche Intelligenz
"Biologically inspired action representation on humanoid robots"
14:00 Uhr, Straße der Nationen 62, Böttcher-Bau, 1/336
Alle interessierten Personen sind herzlich eingeladen!
A major problem in robotics is the generalization of the motor action with respect to different environmental conditions. How could robots learn to perform motor tasks in a specific condition and apply what they have learned in a new condition? How do robots learn new motor tasks based on previously acquired motor skills? This raises the question about the components of the motor program in robotics that provides the possibility to generate adaptive and transferable behaviors. In this presentation, I present the benefits of Neuroscience studies in building a unified framework for robot action generation. Inspired by the hierarchical neural control of movement from cortex to spinal cord, I propose a model that generates a concrete action representation in robotics. The action program is composed of four basic modules: 1) spatial coordination, 2) temporal coordination, 3) sensory motor adaptation, and 4) pattern selection. The latter two are considered for behavior initiation. The model hosts rhythmic and non-rhythmic patterns, accordingly, the robot is able to generate different tasks such as locomotion and manipulation. My results demonstrate that mimicking human intelligence in robotics leads us in the long run to a new generation of robots that learn how to adapt and transfer their behaviors.