Topics on Fast NMPC : The control updating period issue
Event details
| Date | 12.10.2012 |
| Hour | 10:15 › 11:00 |
| Speaker | Mazen Alamir (Directeur de recherche CNRS, University of Grenoble) |
| Location |
ME C2 405
|
| Category | Conferences - Seminars |
In this talk, the need for special formulations when implementing Nonlinear Model Predictive Control to systems showing fast dynamic is recalled. In particular, the now widely admitted idea of distributing the optimization of the cost function over the system life-time is discussed. More precisely, this means that only a finite number of iterations of some descent method are applied before the control sequence is updated leading hopefully to recover the ideal standard NMPC asymptotically. Nevertheless, the way this finite number of iterations is determined is still to be investigated. The talk then focuses on a proposition for a rational adaptation of this number using cheap computations that are based on the on-line algorithm behaviour. While the proposed framework is general, some specific investigations are proposed when the underlying descent method is the fast gradient that is currently under intense investigation by the fast NMPC community. An illustrative example is given in order to assess the efficiency of the proposed framework. It is shown in particular that regardless the initial choice of the control updating period, the scheme converges to a somehow optimal behaviour.
Biographic sketch
Mazen Alamir graduated in Mechanics (Grenoble, 1990) and Aeronautics (Toulouse 1992). He received his PhD in Nonlinear Model Predictive Control in 1995. Since 1996, he has been a CNRS research associate (National French Research Centre) in the Control Systems Department of the University of Grenoble and becomes Director of Research in 2008. His main research topics are NMPC, Moving-Horizon Observers, Nonlinear Hybrid Systems, Signature Based Diagnosis, Optimal cancer treatment as well as industrial applications (Smart Buildings, Stirling Engines, Cryogenic plants, polymerization and chromatographic processes). He is member of the IFAC technical committees on “Nonlinear Systems” and “Optimal Control” and is member of the IEEE Conference Editorial Board. He served as head of the “Nonlinear Systems and Complexity Group” of the control system department. He was recipient of five funding grants from the French National Research Agency and managed over ten R&D industrial contracts. He organized the first IFAC workshop on NMPC for fast systems in Grenoble, 2006 and served as guest editor of a special Fast- NMPC issue of the international Journal of Robust and Nonlinear Control.
Biographic sketch
Mazen Alamir graduated in Mechanics (Grenoble, 1990) and Aeronautics (Toulouse 1992). He received his PhD in Nonlinear Model Predictive Control in 1995. Since 1996, he has been a CNRS research associate (National French Research Centre) in the Control Systems Department of the University of Grenoble and becomes Director of Research in 2008. His main research topics are NMPC, Moving-Horizon Observers, Nonlinear Hybrid Systems, Signature Based Diagnosis, Optimal cancer treatment as well as industrial applications (Smart Buildings, Stirling Engines, Cryogenic plants, polymerization and chromatographic processes). He is member of the IFAC technical committees on “Nonlinear Systems” and “Optimal Control” and is member of the IEEE Conference Editorial Board. He served as head of the “Nonlinear Systems and Complexity Group” of the control system department. He was recipient of five funding grants from the French National Research Agency and managed over ten R&D industrial contracts. He organized the first IFAC workshop on NMPC for fast systems in Grenoble, 2006 and served as guest editor of a special Fast- NMPC issue of the international Journal of Robust and Nonlinear Control.
Practical information
- General public
- Free
Organizer
- Colin Jones
Contact
- Colin Jones ([email protected])