Platzhalter für Bild, Bild auf Titelfolie hinter das Logo einsetzen Oliver Pajonk, Bojana Rosic, Alexander Litvinenko, Hermann G. Matthies ISUME 2011, Prag, A Deterministic Filter for non-Gaussian State Estimation Institute of Scientific Computing Picture: smokeonit (via Flickr.com)
3rd May 2011 | Oliver Pajonk, ISUME 2011 | A Filter for non-Gaussian State Estimation Motivation / Problem Statement State inference for dynamic system from measurements Proposed Solution Hilbert space of random variables (RVs) + representation of RVs by PCE a recursive, PCE-based, minimum variance estimator Examples Method applied to: a bi-modal truth; the Lorenz-96 model Conclusions Outline
3rd May 2011 | Oliver Pajonk, ISUME 2011 | A Filter for non-Gaussian State Estimation Motivation Estimate state of a dynamic system from measurements Lots of uncertainties and errors Bayesian approach: Model “state of knowledge” by probabilities New data should change/improve “state of knowledge” Methods: Bayes’ formula (expensive) or simplifications (approximations) Common: Gaussianity, linearity Kalman-filter-like methods KF, EKF, UKF, Gaussian-Mixture, … popular: EnKF All: Minimum variance estimates in Hilbert space Question: What if we “go back there”? [Tarantola, 2004] [Evensen, 2009]
3rd May 2011 | Oliver Pajonk, ISUME 2011 | A Filter for non-Gaussian State Estimation Motivation / Problem Statement State inference for dynamic system from measurements Proposed Solution Hilbert space of random variables (RVs) + representation of RVs by PCE a recursive, PCE-based, minimum variance estimator Examples Method applied to: a bi-modal truth; the Lorenz-96 model Conclusions Outline
Tool 1: Hilbert Space of Random Variables 3rd May 2011 | Oliver Pajonk, ISUME 2011 | A Filter for non-Gaussian State Estimation *Under usual assumptions of uncorrelated errors! [Luenberger, 1969]
Tool 2: Representation of RVs by Polynomial Chaos Expansion (1/2) 3rd May 2011 | Oliver Pajonk, ISUME 2011 | A Filter for non-Gaussian State Estimation * Of course, there are still more representations – we skip them for brevity. [e.g. Holden, 1996]
Tool 2: Representation of RVs by Polynomial Chaos Expansion (2/2) 3rd May 2011 | Oliver Pajonk, ISUME 2011 | A Filter for non-Gaussian State Estimation [Pajonk et al, 2011] “min-var-update”:
3rd May 2011 | Oliver Pajonk, ISUME 2011 | A Filter for non-Gaussian State Estimation Motivation / Problem Statement State inference for dynamic system from measurements Proposed Solution Hilbert space of random variables (RVs) + representation of RVs by PCE a recursive, PCE-based, minimum variance estimator Examples Method applied to: a bi-modal truth; the Lorenz-96 model Conclusions Outline
Example 1: Bi-modal Identification 3rd May 2011 | Oliver Pajonk, ISUME 2011 | A Filter for non-Gaussian State Estimation 12 … 10
Example 2: Lorenz-84 Model 3rd May 2011 | Oliver Pajonk, ISUME 2011 | A Filter for non-Gaussian State Estimation [Lorenz, 1984]
Example 2: Lorenz-84 – Application of PCE-based updating 3rd May 2011 | Oliver Pajonk, ISUME 2011 | A Filter for non-Gaussian State Estimation PCE “Proper” uncertainty quantification Updates Variance reduction and shift of mean at update points Skewed structure clearly visible, preserved by updates
Example 2: Lorenz-84 – Comparison with EnKF 3rd May 2011 | Oliver Pajonk, ISUME 2011 | A Filter for non-Gaussian State Estimation
Example 2: Lorenz-84 – Variance Estimates – PCE-based upd. 3rd May 2011 | Oliver Pajonk, ISUME 2011 | A Filter for non-Gaussian State Estimation
Example 2: Lorenz-84 – Variance Estimates – EnKF 3rd May 2011 | Oliver Pajonk, ISUME 2011 | A Filter for non-Gaussian State Estimation
Example 2: Lorenz-84 – Non-Gaussian Identification 3rd May 2011 | Oliver Pajonk, ISUME 2011 | A Filter for non-Gaussian State Estimation (a) PCE-based (b) EnKF
Conclusions & Outlook 3rd May 2011 | Oliver Pajonk, ISUME 2011 | A Filter for non-Gaussian State Estimation Recursive, deterministic, non-Gaussian minimum variance estimation method Skewed & bi-modal identification possible Appealing mathematical properties: Rich mathematical structure of Hilbert spaces available No closure assumptions besides truncation of PCE Direct computation of update from PCE efficient Fully deterministic: Possible applications with security & real time requirements Future: Scale it to more complex systems, e.g. geophysical applications “Curse of dimensionality” (adaptivity, model reduction,…) Development of algebra (numerical & mathematical)
References & Acknowledgements 3rd May 2011 | Oliver Pajonk, ISUME 2011 | A Filter for non-Gaussian State Estimation Pajonk, O.; Rosic, B. V.; Litvinenko, A. & Matthies, H. G., A Deterministic Filter for Non-Gaussian Bayesian Estimation, Physica D: Nonlinear Phenomena, 2011, Submitted for publication Preprint: The authors acknowledge the financial support from SPT Group for a research position at the Institute of Scientific Computing at the TU Braunschweig. Lorenz, E. N., Irregularity: a fundamental property of the atmosphere, Tellus A, Blackwell Publishing Ltd, 1984, 36, Evensen, G., The ensemble Kalman filter for combined state and parameter estimation, IEEE Control Systems Magazine, 2009, 29, Tarantola, A., Inverse Problem Theory and Methods for Model Parameter Estimation, SIAM, Philadelphia, 2004 Luenberger, D. G., Optimization by Vector Space Methods, John Wiley & Sons, 1969 Holden, H.; Øksendal, B.; Ubøe, J. & Zhang, T.-S., Stochastic Partial Differential Equations, Birkhäuser Verlag, 1996