From motor to interaction learning in robots (IROS 2008 workshop)

From motor to interaction learning in robots

An IROS 2008 workshop

September 26, 2008

Acropolis Convention Center, Nice, FRANCE

Organizers:

Olivier Sigaud, ISIR, UPMC-Paris 6, Paris, France
Jan Peters, Max Planck Institute for Biological Cybernetics, Tubingen, Germany
Sethu Vijayakumar, University of Edinburgh, Scotland

Objectives and topics:

From an engineering standpoint, in front of the increasing complexity of robotic systems and the increasing demand for more autonomous and more adaptive robots, the design of on-line, general purpose learning capabilities for their control becomes more and more mandatory. Furthermore, beyond autonomy and adaptation, in the case of humanoid robots in particular, the need for "natural" interaction capabilities with the human user(s) raises new challenges at the interface between interaction models and learning techniques. From neurophysiological studies of human subjects, it is more and more clear that there is much in common between motor learning processes and more cognitive learning and developmental processes, particularly dedicated to interaction. After the much celebrated discovery of the so called "mirror neurons" relating motor learning to imitation and language acquisition, several neurophysiological studies have revealed that brain areas generally considered as motor, such as the cerebellum, or dedicated to action selection, like the basal ganglia, are in fact implied in more general cognitive functions learning such as tool use, imitation, language and so forth. Taken together, these facts advocate for a hierarchical understanding of the brain architecture where motor learning and interaction learning are tightly coupled processes at the root of cognition. The complexity of the computational models resulting from this line of thinking raises the problem of their validation. Here, robotics plays a prominent role as a tool to evaluate the capability of these wide-scope models to give account of the phenomena they address.

We particularly welcome papers on the following topics:

robotic implementation of computational or neuromimetic models of motor learning
robotic implementation of computational or neuromimetic models of interaction learning
robotic controllers combining motor learning and interaction learning
hierarchical models of mammal motor systems
computational principles underlying motor learning and/or interaction learning

Intended audience:

The goal of this workshop is to bring together researchers in "engineering sciences" (roboticists, machine learning experts) interested in motor learning and interaction learning to design better robots, and researchers in "life science" (computational neurosciences researchers, developmental psychologists) interested in robotic implementation of models of these processes in animals and humans, to get a more accurate picture of what both learning processes share, how they interact and how they can eventually be combined in the design of robot controllers.

Selected contributions

Andry, P., Gaussier, P., Lagarde, M., Boucenna, S. and Hafemeister, L.: Proprioception and imitation: on the road to agent individuation	talk	slides
Billard, A. : Learning the underlying dynamics of motion from imitating humans	talk	slides (300Mo)
Borji, A., Ahmadabadi, M. N. and Araabi, B. N.: Interactive Learning of Top-down Attention Control and Motor Actions	poster	slides
Chaminade, T.: Applying motor resonance to humanoid robots	poster	slides
Giovannangeli , C., Boucenna, S. and Gaussier, P.: About the constuctivist role of self-evaluation for interactive learnings and self-development	poster
Grenet, G., Alexandre, F.: Behavior learning using emotional conditioning	poster	slides
Herbort, O., Pedersen, G., Butz, M. V.: A Neural Network Model to Learn and Flexibly Control a Redundant, Dynamic Arm	talk	slides
Hörnstein, J., Gustavsson, L., SantosVictor, J. and Lacerda, F.: Modeling Speech Imitation	poster	slides
Howard, M., Klanke, S. and Vijayakumar, S. : Learning Nullspace Potentials from Constrained Motion for Apprenticeship Learning	talk	slides
Kober, J. and Peters, J.: Learning Ball-In-The-Cup	poster	slides
Lopes, M., Melo, F. S., Kenward, B. and Santos-Victor, J.: A Computational Model for Social Learning Mechanisms	talk	slides
Metta, G., Sandini, G., Vernon, D., Natale, L. and Nori, F.: The RobotCub Approach to the Development of Cognition	talk	slides
Nguyen-Tuong, D. and Peters, J.: Local Gaussian Processes Regression for Real-time Model-based Robot Control	poster	slides
Oudeyer, P.-Y. and Baranès, A.: Developmental active learning with intrinsic motivation	talk	slides
Peters, J.: Motor Skill Learning for Robotics	talk	slides
Quinton, J.-C. and Buisson, J.-C.: Interactivist sensorimotor learning: computational implementation and parallel optimization	poster	slides
Robertson, P.: Memory-based Simultaneous Learning of Motor, Perceptual and Navigation Skills	poster	slides
Saegusa, R., Sakka, S., Metta, G. and Sandini, G.: Autonomous Learning Evaluation toward Active Motor Babbling	poster	slides
Salotti, J.-M. and Lepretre, F.: Classical and Operant Conditioning as Roots of Interaction for Robots	poster	slides
Shen, Q., Saunders, J., Kose-Bagci, H. and Dautenhahn, K.: Acting and Interacting Like Me? A Method for Identifying Similarity and Synchronous Behavior between a Human and a Robot	poster	slides
Sigaud, O., Padois, V., Salaün, C. and Truchet, A.:Model-Based Actor-Critic and Operational Space Control in the context of Motor Learning	talk	slides
Toussaint, M.: Probabilistic inference methods in robotics	talk	slides

Final schedule: download

We plan to get post-workshop proceedings published as a book.

Important dates:

Workshop	September 26
Post-proceedings full paper submission	February 2009

Program committee:

Philippe Gaussier	University of Cergy-Pontoise, France
Giorgio Metta	University of Genova, Italy
Pierre-Yves Oudeyer	INRIA, Bordeaux Sud-Ouest, France
Jan Peters	Max Planck Institute, Tubingen, Germany
Olivier Sigaud	UPMC-Paris 6, Paris, France
Marc Toussaint	University of Berlin, Germany
Sethu Vijayakumar	University of Edinburgh, Scotland