Estimation of parameters and unobserved components for nonlinear systems from noisy time series
- We study the problem of simultaneous estimation of parameters and unobserved states from noisy data of nonlinear time-continuous systems, including the case of additive stochastic forcing. We propose a solution by adapting the recently developed statistical method of unscented Kalman filtering to this problem. Due to its recursive and derivative-free structure, this method minimizes the cost function in a computationally efficient and robust way. It is found that parameters as well as unobserved components can be estimated with high accuracy, including confidence bands, from heavily noise-corrupted data.
Author details: | Andre Sitz, Udo Schwarz, Jürgen KurthsORCiDGND, Henning U. Voss |
---|---|
URL: | http://pre.aps.org/ |
Publication type: | Article |
Language: | English |
Year of first publication: | 2002 |
Publication year: | 2002 |
Release date: | 2017/03/24 |
Source: | Physical Review / E. - 66 (2002), S. 016210 |
Organizational units: | Mathematisch-Naturwissenschaftliche Fakultät / Institut für Physik und Astronomie |
Institution name at the time of the publication: | Mathematisch-Naturwissenschaftliche Fakultät / Institut für Physik |