The Science
Estimating complex, non-linear model states and parameters from uncertain systems of equations and noisy observation data with current filtering methods is a key challenge in mathematical modeling. To address this challenge, a team from 91, Florida State University, and the University of Kansas developed the first filtering method that directly estimates parameters without solving the estimated model state.
Previously, the most effective filtering methods had to transform parameter estimation into a pseudo parameter process and augment the filtering method to solve the dynamical system and parameter estimation together.
The Impact
The team’s Direct Filtering method produces an observational function to directly utilize measured data. This approach improves efficiency by only estimating the parameter process and improves accuracy and speed by mitigating degeneracy from space state estimation.
PI/Facility Lead: Rick Archibald
ASCR Program/Facility: Applied Mathematics
ASCR PM: Ceren Susut
Funding: SciDAC Institutes
Publication(s) for this work: R. Archibald, F. Bao, and X. Tu, “a direct filter method for parameter estimation.” Journal of Computational Physics, 398, 2019. DOI: 10.1016/j.jcp.2019.108871
