Modeling and control of distributed systems: system-identification based on data.
|Project members:||Dr.ir. P. Massioni (Paolo), prof.dr.ir. M. Verhaegen (Michel)|
|Keywords:||Distributed estimation and control, Identification and estimation, Distributed and large-scale systems|
With few exceptions control problems with spatially distributed sensing and actuation up to now have not been thoroughly studied in control theory due to the perception of their technological infeasibility. Recently, however, technological progress is bringing dramatic changes to this picture. In particular, advances in micro-electro-mechanical systems (MEMS) make feasible the idea of microscopic devices with actuating, sensing, computing and telecommunications capabilities. Distributing a large array of such devices in a spatial configuration gives unprecedented capabilities for control; examples already include distributed flow control, "smart" mechanical structures, and Cross Directional control in the chemical process industry. For all these applications the control variables can be conveniently and appropriately thought to be distributed in space, in addition to the internal states. Important questions that arise are (i) how to design control algorithms for these systems with regard to global objectives; and (ii) how can these control algorithms be implemented in a distributed array, e.g. in a "localized" way. Needless to say that these questions can be studied from many points of view, and internationally there is a strong emerging activity devoted to exploring different lines of research.
This project aims to investigate control methodologies based on the identification. Such methods spring from the control methodology of the Delft Center for Systems and Control, and will lead to development of low order models for high performance feedback control from experimental data. The first innovative research question is to develop identification methods for identifying models to approximate the dynamic behaviour of spatially distributed dynamic systems. The second is to integrate this identification procedure with a robust control method for the class of spatially distributed models.