Prof. Lenka Pavelková
Institute of Information Theory and Automation
Prague, Czech Republic
Tuesday 2015-11-03 16:00
IST-Seminar-Room V2.268 - Pfaffenwaldring 9 - Campus Stuttgart-Vaihingen
Recursive estimation is an important part of the decision making tasks such as prediction and adaptive control. The system in question is often modelled by an autoregressive model with exogenous inputs (ARX). Uncertainty of this model is mostly assumed to be normal and the estimation of unknown model parameters is based on least squares methods. A state space model is another frequently used model. Considering normal noises in this model, Kalman filtering is standardly used for the states estimation. Nevertheless, the Gaussian models do not respect possible strict boundaries of involved model variables. In such cases, the above mentioned estimation procedures have to be additionally adapted.
In this talk, the mentioned estimation problem with constraints will be addressed straightforwardly within the Bayesian approach. The results of the research of both ARX and state space models with uniformly distributed noises will be presented. An extension towards other bounded distributions will be mentioned. Applications in the transportation area and industry will be shown.
Lenka Pavelková graduated in Technical Cybernetics at the Czech Technical University in Prague, Czech Republic, in 1992, and obtained her Ph.D. in Applied Mathematics in 2009 from the same university. Currently, she works as a research worker at the Department of Adaptive systems, Institute of Information Theory and Automation, Czech Academy of Sciences, Prague. Her research interests are in the field of Bayesian decision making. She focuses mainly on the estimation of constrained systems.