Rational basis functions for robust identification from frequency and time-domain measurements (Q1295162)

The paper provides an analysis of the use of rational model structures in a robust estimation context. The key result is the provision of necessary and sufficient conditions on the poles of the rational model structures. The paper shows how robust estimation algorithms using both time and frequency domain data could be constructed together with explicit error bounds on the estimation accuracy. These results have implications, as discussed, for mixed parametric/non-parametric estimation, model reduction and may be extended to the multivariable setting.