SCO Seminar - Marieke Heidema, University of Groningen

Title: System Identification using Energy-Bounded Noise Models: A Full Characterization of Chebyshev Centers and Radii

Abstract:

This presentation covers our recent research results on the identification of linear time-invariant systems using noisy input-output data in a set-membership framework. It is assumed that the noise sequence belongs to a bounded set, induced by a quadratic matrix inequality, that allows for capturing various energy-bounded noise models. In this framework, the identification problem is to find an estimation for the true system so that the worst-case estimation error with respect to a given norm is minimized. For such a problem, we present closed-form solutions with respect to arbitrary unitarily invariant matrix norms. Examples of these norms include the Frobenius norm, spectral norm, Schatten p-norms, and Ky Fan k-norms.