Estimation of uncertainty bounds for robustness analysis

The paper considers the problem of deriving perturbed plant models from noisy input/output data. It is motivated by the use of the former in robust multivariable control analysis/design procedures. The approach taken centres around constructing confidence regions for the plant frequency response at a finite number of frequencies via spectral estimates and on an elementwise basis. The basic building block is the derivation of the distributions required for obtaining the gain and phase confidence intervals at a single frequency for a SISO frequency response.