Le Laboratoire d’Automatique de l’EPFL a le plaisir de vous inviter aux séminaires selon la liste ci-après. Une mise à jour régulière des informations concernant ces séminaires est disponible à l’adresse sur cette page. En particulier, il est conseillé aux visiteurs externes de vérifier que les séminaires soient dispensés comme prévu ci-dessous.
Where: Salle de séminaire LA-EPFL, ME C2 405 (2è étage), 1015 Lausanne
When: Friday at 10h15
24.03.2006 – Dr. D. Garcia – Laboratoire d’Automatique, EPFL
Les régulateurs de type proportionnel intégral dérivé (PID) sont, de par leurs structures simples, très appréciés et abondamment utilisés par la communauté industrielle. Le nombre restreint de paramètres d’ajustement implique des performances souvent plus modérées que celles pouvant potentiellement être obtenues en utilisant un régulateur d’ordre élevé. En contrepartie, la synthèse d’une commande PID doit pouvoir être réalisée de façon simple et rapide. En pratique, les méthodes ne requérant qu’un minimum d’informations du système à com-mander sont largement favorisées. Différentes approches, ne nécessitant pour la plupart aucun modèle du système à commander, seront présentées dans ce contexte. L’ajustement est considéré dans le domaine fréquentiel, car la stabilité et la robustesse, ainsi que les performances temporelles du système en boucle fermée, peuvent y être représentées. Les indicateurs classiques de robustesse, constitués des marges de phase et de gain, la pulsation de croisement, mais également des paramètres plus avancés, composés des normes infinies des fonctions de sensibilité, peuvent être pris en compte par les procédures présentées. Plu-sieurs exemples et applications seront en outre proposés pour témoigner de la simplicité et de l’efficacité de ces méthodes.
07.04.2006 – Dr. A. Voda – Laboratoire d’Automatique de Grenoble (LAG), INPG, Grenoble, France
Optical disk (such as Compact Disk (CD) and Digital Versatile Disk (DVD)) drives are widely used for the storage of data, music and videos. Even though improvements are observed in obtaining shorter data time access, higher storage capacity and information density on the disk, one of the major obstacles for reli-ability of read-out data is due to the internal and external disturbances affecting the disk. Data are retrieved by focusing a laser beam on the disk surface (data-layer) and then following the track, both with high precision. This is difficult in the presence of disturbances, and control loops are necessary. One of the most important disturbances is the disk eccentricity. A controller is used to deal with this repetitive disturbance, which has a variable frequency (depending on the beam actual position on the track). The resulting controller structure is often quite complex (robust, adaptive).
In the present talk, we present a simple way to find a fixed low-order controller that is able to improve the eccentricity rejection in the focus/tracking control loops for DVD players. It is shown that this can be achieved with pole placement design, followed by controller order reduction. Pole placement method is here adapted to realize repetitive disturbance rejection in a certain bandwidth. The control design method is validated on a real system from STMicroelectronics.
13.04.2006 (Thursday 10:15) – Prof. R. D. Braatz – Department of Chemical and Biomolecular Engineering, University of Illinois at Urbana-Champaign, Illinois, U.S.A.
This presentation describes applications of molecular simulation to microelec-tronics processes and the subsequent development of techniques for multiscale simulation and multiscale systems engineering. The progression of the applica-tions of simulation in the semiconductor industry from macroscopic to molecular to multiscale is reviewed. Multiscale systems theory is presented as an approach that incorporates molecular and multiscale simulation to design processes that control events at the molecular scale while simultaneously optimizing all length scales from the molecular to the macroscopic. It is discussed how design and control problems in microelectronics and nanotechnology, including the targeted design of processes and products at the molecular scale, can be addressed using the multiscale systems tools. This provides a framework for addressing the “grand challenge” of nanotechnology: how to move nanoscale science and tech-nology from art to an engineering discipline.
04.05.2006 (Thursday 15:15) – Prof. M. Campi – Department of Electrical Engineering for Automation, University of Brescia, Italy
Many worst-case robust control problems cannot be solved due to computational intractability. In this talk, a new probabilistic solution framework is proposed for robust control analysis and synthesis problems that can be expressed in the form of robust convex optimization. This includes for instance the wide class of NP-hard control problems that can represented by means of parameter-dependent linear matrix inequalities (LMIs). By appropriate sampling of the constraints, one obtains a convex optimization problem (the scenario problem) that can be easily solved through standard optimization techniques. The solution of the scenario problem is approximately feasible for the original (usually infinite) set of constraints, i.e. the measure of the set of original constraints that are violated by the scenario solution rapidly decreases to zero as the number of samples is in-creased. Explicit and efficient bounds on the number of samples required to at-tain a priori specified levels of probabilistic guarantee of robustness are given. A rich family of control problems which are in general hard to solve in a deterministically robust sense is therefore amenable to polynomial-time solution if robustness is intended in the proposed risk-adjusted sense. The same approach can be applied to a number of other contexts beyond robust control.
19.05.2006 (Room ME B10, map )– Prof. P. Scott – Centre for New Media, Knowledge Media Institute, The Open University, Milton Keynes, UK
New technologies for working live in an online community can transform work and learning. Under the aegis of the Prolearn European Network of Excellence we have conducted a wide range of global experiments in live cooperative learn-ing (see http://prolearn.tv/). Of the various models we have deployed into this network and beyond, the single most successful experiment has been the FlashMeeting video conferencing service. Over the last 2 years some 2000 live, naturalistic and real meetings have been recorded on just one of our re-search servers. Meetings range from 2 to 30 participants, and are conducted in a web browser using a simple audio-video applet. All meetings are recorded and instantly available to participants. Some groups annotate, and syndicate meeting records beyond the meeting community. Popular meeting models have included online seminars, lectures, and project work, but some of the most compelling on-line events are innovative student peer-tutoring groups and global communities of practice. Extensions of this work now support e-twinning of schools across the world and a range of other novel learning models.
09.06.2006 – Dr. B. Mahn – BASF AG, Ludwigshafen, Germany
It is widely recognized in industry that advanced process control techniques have the potential to improve the performance of operations. In the chemical and pet-rochemical industry, linear MPC is the most commonly used and established ad-vanced control strategy. Due to the nonlinearity of polymer processes, nonlinear model predictive control is more suitable as this technology has the ability to cope with process nonlinearities and constraints. Nonlinear model predictive control is an optimization-based multivariable con-trol technique that uses a nonlinear dynamic process model for the prediction of the process behavior. Recent developments in optimization algorithms have led to NMPC methods that are very reliable and robust. The presentation will elabo-rate on the advantages and disadvantages of model predictive control (N)MPC for the setpoint tracking control of an industrial batch polymerization reactor and of an industrial divided wall column. The performance of the (N)MPC imple-mentations is assessed via simulation and experimental results.
16.06.2006 – Prof. F. E. Cellier – Institut für Computational Science, ETHZ
Fuzzy Inductive Reasoning (FIR) is a non-parametric model synthesis technique designed for accurately modeling complex systems with continuously changing states from observations of their input/output behavior. FIR offers an attractive alternative to Artificial Neural Networks (ANN), as its model synthesis proce-dure converges much more rapidly than the training mechanisms used by differ-ent ANN approaches. Also, since FIR is a non-parametric technique, it does not fall prey to making meaningless predictions in situations where the testing data are incommensurate to the training data used for tuning the ANN parameter set. Finally, FIR contains an intrinsic error model, i.e., FIR does not only make pre-dictions about an output variable, but also about the error of its own prediction. In this respect, FIR resembles more statistical approaches to inductive modeling, and yet, FIR is much easier to use than elaborate nonlinear statistical techniques. FIR has proven successful in many application areas, such as biomedical (car-diovascular systems modeling, control of anesthetic agents during surgery), eco-logical (control of shrimp growth in shrimp farming), and technological (fault recognition and isolation in complex power plants, fuzzy controller design for tanker ship steering).