Discovering art in robotic motion: From imitation to innovation via interactive evolution

Abstract

The proposed research deals with a usage of artificial intelligent techniques in humanoid robotics. The objective is to design an interactive robotic behavior. The focus is on social robotics and how to improve the interaction between human and robot. In the first part user interactively cooperates with the System and this is the way how the robot acquires knowledge. Our approach is based on real manipulation where the robot freely interacts with humans and objects, and learns during continuous interaction with the environment. The experimental setup is the humanoid robot NAO. In the second part of the system we use the Interactive Evolutionary Computation - a technique which optimizes systems based on subjective human evaluation. So far, the algorithm is applied to a system of evolution of pleasant motions or dances. A user can build a robot behavior without any programming knowledge and then the evolutionary algorithm helps user to discover new possibilities of motions. © 2011 Springer-Verlag.