Asea Brown Boveri Ltd. (ABB) Award 2008 - Calinon Sylvain

"For the thorough and exemplary work in his Ph.D. thesis and important scientific contributions resolving two major issues in robot programming by demonstration."

Continuous extraction of task constraints in a robot programming by demonstration framework.

Robot Programming by Demonstration (RbD) explores user-friendly means of teaching a robot new skills. Recent advances in RbD have identified a number of key-issues for ensuring a generic approach to the transfer of skills across various agents and situations. This thesis focuses on the two generic questions of what-to-imitate and how-to-imitate, which are respectively concerned with the problem of extracting the essential features of a task and determining a way to reproduce these features in different situations. The perspective adopted in this work is that the robot may infer what the key aspects of a skill are by observing multiple demonstrations and assuming that the key components are invariant across the demonstrations. We show that statistical methods based on Gaussian Mixture Model (GMM) and Gaussian Mixture Regression (GMR) can be used efficiently for incremental and active teaching scenarios where the user only provides a small number of demonstrations, either by wearing a set of motions sensors, or by helping the robot refine its skill by kinesthetic teaching, that is, by embodying the robot and putting it through the motion. These techniques are then applied for enabling a humanoid robot to extract automatically different constraints by observing several manipulation tasks and to reproduce the learned skills in new situations.

Distingué également par la mention EPFL-PRESS remise par les PPUR.