Le 3 Juin 2002, 18h, Salle des Conferences (au 46, Rue d'Ulm):
Giuseppe Longo (Informatique, CNRS - ENS)
"Des fondements des mathématiques aux preuves comme programmes : la résistible ascension de la métaphore "le cerveau est un ordinateur digital", quelques alternatives et leur pertinance aux problèmes de l'intelligibilité cognitive et mathématique de l'espace"
Le 11 Juin 2002, 13h, Salle Dussane :
Massimo Isidori (Informatica e Sistemistica, Universita di Roma I)
"Finesse et géométrie: the spirit of nonlinear control" (en collaboration avec le Colloquium ENS)
Le 20 Juin 2002, 17h - 19h 30, Salle des Resistants, deux exposés :
Maria Chiara Carrozza (ARTS Lab, Scuola Superiore Sant'Anna, Pisa)
"Functional replacement and humanoid robotics: the fusion of Natural and Artificial Hardware"
Olivier Coenen (Sony, Paris)
"A Neural Approach to Learning Sensorimotor Coordination in Robots"
--------------
(Resumés)
Maria Chiara Carrozza
Functional Replacement and Humanoids Robotics: the Fusion of Natural and Artificial Hardware
The presentation is aimed at introducing the projects currently ongoing at ARTS Lab in field of biomedical and humanoid robotics, with specific attention to the problem of interfacing artificial hardware directly to the human brain. Recent progresses of microengineering, bioengineering, microelectronics and computer science have provided "enabling technologies" for the development of basic components of a humanoid robot. For example, following an approach of pure functional replacement, we can imagine to fabricate the building blocks of a humanoid robot composed of artificial organs (artificial heart, skin...), and artificial limbs (hands, legs...). Even if there are still several theoretical and technical problems to be solved, related to energy sources, actuators performance and sensors fabrication, there is a positive trend towards the development of human-like components with increasing performance. It is clear that the objective to create a humanoid robot, which is aimed at reproducing human architecture, morphology, and behavior, may open up not only several important technological challenges but also fundamental ethical issues: is it possible to create a humanoid robot? We can answer to this question by going back to the ultimate aim of bioengineering which is devoted to the service of humans. The approach of human robotics is aimed at investigating robotic systems, models and tools to support and assist the human being. The word "human" identifies the common framework for research done at ARTS Lab in the fields of computer assisted surgery, rehabilitation bioengineering, motor control and microengineering. The projects currently ongoing at ARTS Lab are aimed at solving practical problems of disable and elderly people with serious motor or cognitive disabilities in order to restore as much as possible their quality of life. In order to obtain an effective and natural control of the artificial device, it is important to provide the system with a neural interface able to establish the connection to the natural brain. As a case study of human robotics and neural interfaces, the presentation will be concluded with an overview of the CYBERHAND project, which is aimed at investigating the problem of functional replacement of the natural hand.
-------------------
Olivier Coenen.
A Neural Approach to Learning Sensorimotor Coordination in Robots.
Robots are of increasing importance and will become an integral part of society. Despite continuing advances, current robots cannot approach the abilities of even the simplest mammals. Achieving continuous and real-time learning without interference between learnt tasks remains a difficult problem. To attain the learning ability and finesse of movement that animals display, information from a large number of sensorimotor and cognitive signals must be integrated. However, few principles of integration have been proposed. The presentation will introduce the SpikeFORCE project that includes an inter-disciplinary consortium of physicists, neuroscientists and engineers who are investigating such principles of integration and representation in the brain. One objective is to understand the neural principles that give an organism the ability to learn actions tailored to the current context in diverse tasks and with minimal destructive interference between tasks. Another is to investigate how this ability can be recreated in real-time spiking neural networks. The real-time spiking neural networks will be based in part on the cerebellum, a major site of sensorimotor integration and motor learning in the brain. A path from experiments yielding physiological data to the design and construction both in hardware and simulated networks of real-time models of the cerebellum will be described. The potential learning and control abilities of these spiking networks will be introduced and future testing with robots in real and simulated environments will be discussed.
Pour plus d'information, voir http://www.di.ens.fr/users/longo/geocogni.html