| Time: | December 6, 2016 |
|---|---|
| Download as iCal: |
|
Prof. Hakan Hjalmarsson
Dept. of Automatic Control
KTH Royal Institute of Technology, Stockholm, Sweden
Tuesday, 2016-12-06 16:00
IST-Seminar-Room V9.22 - Pfaffenwaldring 9 - Campus Stuttgart-Vaihingen
Abstract
Departing from a waterbed effect for the modeling accuracy, inherent in all full scale model estimation, and the Cramér-Rao bound, that limits performance of data driven methods, we discuss how to do data driven control design for complex systems. We compare direct design, when data is used to directly tune a controller, with indirect design, where a model is used as a proxy for the information contained in the data. We argue that broad band excitation in experiments for complex systems may lead to poor performance and discuss application oriented experiment design as a way to cope with this. We show that this technique leads to intuitive appealing designs which alleviate the choice of model structure, the hardest problem in system identification.
Biographical Information
Håkan Hjalmarsson received a M.Sc. in Electrical Engineering in 1988 and a Ph.D. in 1993, both from Linköping University, Sweden. He has served as an Associate Editor for Automatica and IEEE Transactions on Automatic Control and been Guest Editor for European Journal of Control and Control Engineering Practice. He is Professor at the School of Electrical Engineering, KTH, Stockholm, Sweden. He is an IEEE Fellow and Chair of the IFAC Coordinating Committee on Systems and Signals. He is co-recipient of a European Research Council Advanced Grant. His main research area is system identification.