Conférences plénières > Mercredi 16 novembre
Mercredi 16 novembre
8h15-9h15
Antoine CHAILLET
L2S-Univ. Paris Sud-CentraleSupélec
|
 |
Antoine Chaillet was born in Douai in 1979. In 2002, he received his B.Sc. degree from ESIEE Amiens, and his M.Sc. degree in Control Engineering from Univ. Paris Sud in 2003. In July 2006, he received his Ph.D. degree in Control Theory from Univ. Paris Sud. In 2006-2007, he was a post-doctoral fellow at Univ. degli Studi di Pisa, Italy. Since Sept. 2007, he has been serving as an associate professor at L2S-Univ. Paris Sud-CentraleSupélec. He is a junior member of Institut Universitaire de France. His research interests include stability and robustness analysis of nonlinear systems and control theory for neuroscience. He coordinates the ANR project SynchNeuro on the modeling and control of parkinsonian brain oscillations.
https://sites.google.com/site/antoinechaillet/home
|
|
| Title: Closed-loop attenuation of pathological brain oscillations in a spatio-temporal model
Abstract
Several disorders are related to pathological brain oscillations. In the case of Parkinson’s disease, sustained low-frequency oscillations (especially in the beta-band, 13–30Hz) correlate with motor symptoms. Among the hypotheses to explain the generation of such pathological oscillations, one is the possible pacemaker role played by two feedback-interconnected neuronal populations, one of which is excitatory whereas the other one is inhibitory. In this scenario, the abnormal increase of synaptic weights between these two populations, combined with the inherent transmission delays, gives rise to some instability which translates into sustained oscillations.
In this talk, we rely on a spatiotemporal model of the neuronal populations involved to show that a simple proportional feedback on the excitatory population is enough to attenuate pathological oscillations, provided that the internal synaptic weights within the inhibitory population are sufficiently low. The model used is a delayed nonlinear integro-differential equation know as delayed neural field. Delays are allowed to be position-dependent in order to model longer transmission delays between more distant neurons.
In order to analyze such neuronal populations networks, we extend stability and robustness tools to spatiotemporal delayed dynamics. We provide conditions under which each population is input-to-state stable (ISS) with respect to the inputs coming from the other population. Extending small-gain theorems to spatio-temporal delayed dynamics, we show that proportional feedback successfully suppresses pathological oscillations in the network, provided that the internal coupling inside the inhibitory population is not too strong. The use of ISS allows in turn to evaluate the robustness of the proposed feedback with respect to imperfect actuation and control delays. |
|
Mercredi 16 novembre
11h45-12h45
Denis Efimov a reçu son PhD en Automatique de l'Université d'Etat de Génie Electrique (Saint-Pétersbourg, Russie) en 2001, et le Doctorat es Science en Automatique de l'Institut de Génie Mécanique (Saint-Pétersbourg, Russie) en 2006. Entre 2006 et 2011, il a travaillé au LSS (Supelec, France), à l'Institut Montefiore (Université de Liège, Belgique) et au laboratoire IMS (Université de Bordeaux I, France). Depuis 2011, il a rejoint l'équipe Inria Non-A (Lille). Il est membre de plusieurs comités IFAC et membre sénior IEEE. Ses principaux intérêts de recherche portent sur l'analyse des oscillations, l'observation, le contrôle et la stabilité des systèmes non linéaires.
|
|
|
Titre: Systèmes homogènes : une théorie et un outil pour le contrôle et l'estimation
Résumé
Une brève introduction à la théorie des systèmes dynamiques homogènes est donnée. Tout d'abord, les propriétés de base et l'analyse de la stabilité seront discutées pour ces systèmes non linéaires particuliers. Puis nous montrerons comment utiliser ces modèles homogènes pour garantir de meilleures performances (taux de convergence, robustesse, insensibilité aux retards...). Enfin, des résultats de contrôle et d'estimation seront présentés pour illustrer les avantages des algorithmes homogènes.
|
|
|
