National Technical University of Athens (NTUA)
Αναδρομική εκτίμηση κατάστασης μη γραμμικών στοχαστικών δυναμικών συστημάτων: νέα μη γραμμικά φίλτρα Kalman, βελτιωμένα φίλτρα σωματιδίων και εφαρμογή σε συστήματα συνεχούς καταγραφής της γλυκόζης
Abstract
dc:descriptionThe first part of this thesis deals with the recursive state estimation of discrete time nonlinear stochastic systems. The formulation of the problem and some basic results are presented in the introduction, with emphasis on the Kalman Filter and its nonlinear variants (EKF, UKF, GHKF), while the Particle Filters are also mentioned. Then, some heuristic modifications of nonlinear Kalman Filters are presented: inversion of the output function, estimation of a specific state function, use of an alternative sigma-point set for UKF and use of the mode. In the following, systems that consist of linear dynamical systems interconnected through static nonlinear characteristics are studied. Initially, the case of single-input nonlinearities is considered and it is proved that in this case it is possible to reduce the computation of all the expected values involved to one and two-dimensional integration. This way, more accurate calculations can be made than with the direct integration on the state space, which may be of high dimension. Then the case of multi-input characteristics is treated, and it is shown that again it is possible to reduce the integration order. Furthermore, it is shown that the results of the proposed techniques are invariant to similarity transformations of the state space. A quadrature technique, alternative to the Gauss-Hermite quadrature, specially designed for nonlinear filters using norm minimization concepts is proposed and its convergence is proved. Finally, improvements for the Auxiliary Particle Filter and the Unscented Particle Filter are presented. All the proposed techniques are compared with the standard ones in suitable examples. The results show that in some cases the improvement is drastic. The second part deals with the use of filters in data from a Continuous Glucose Monitoring System (CGMS). The importance of the CGMS in the construction of an Artificial Pancreas is explained and the problems of CGMSs are described. Filters based on the Kalman Filter and on the Particle Filter are designed using simple models of the systems dynamics. The filters are applied to experimental data from ICU patients, and the results show that the use of filters leads to significant reduction of the glucose estimation error.
Degree
thesis:*- Grantor dc:publisher
- National Technical University of Athens (NTUA)
- Year dc:date
- 2011
Author and committee
dc:creator, dc:contributor.*- Authors dc:creator
-
- Charalampidis, Alexandros
- Χαραλαμπίδης, Αλέξανδρος
Subjects
dc:subject × 20- Μη γραμμικά συστήματα
- Στοχαστικά συστήματα
- Αναδρομική εκτίμηση κατάστασης
- Νόμος Bayes
- Φίλτρο Καλμαν
- Φίλτρα σωματιδίων
- Τεχνητό πάγκρεας
- Συστήματα συνεχούς καταγραφής της γλυκόζης
- Nonlinear systems
- Stochastic systems
- Recursive state estimation
- Bayes' rule
- Kalman filter
- Particle filters
- Artificial pancreas
- Continous glucose monitoring systems
- Επιστήμες Μηχανικού και Τεχνολογία
- Επιστήμη Ηλεκτρολόγου Μηχανικού, Ηλεκτρονικού Μηχανικού, Μηχανικού Η/Υ
- Engineering and Technology
- Electrical Engineering, Electronic Engineering, Information Engineering
Rights
- Language dc:language
- gre
Identifiers
dc:identifier.*- Identifier
- 10.12681/eadd/24438