Open access
Date
2023-12Type
- Report
ETH Bibliography
yes
Altmetrics
Abstract
This work proposes a computationally tractable method for the identification of canonical rational transfer function models, using a finite set of input-output measurements. The problem is formulated in frequency-domain as a global optimization problem whose cost function is the sum of weighted squared residuals at each observed frequency datapoint. It is solved by the moment-sum-of-squares hierarchy of semidefinite programs, through a framework for sum-of-rational-functions optimization from Bugarin, Henrion, Lasserre 2016. The generated program contains decomposable term sparsity Wang et al. (2021) which can be exploited for further computational complexity reductions. Convergence of the moment-sum-of-squares program is guaranteed as the bound on the degree of the sum-of-squares polynomials approaches infinity. We discuss extensions of this rational-program method for identification of stable systems, and closed-loop identification. Show more
Permanent link
https://doi.org/10.3929/ethz-b-000646581Publication status
publishedPublisher
ETH ZurichSubject
Frequency domain identification; System analysis and optimization; Linear systems; Large scale optimization problems; Rational optimization; Convex optimizationOrganisational unit
08814 - Smith, Roy (Tit.-Prof.)
Funding
180545 - NCCR Automation (phase I) (SNF)
More
Show all metadata
ETH Bibliography
yes
Altmetrics