Nuclear Norms for System Identification - a direct input-output approach

This paper studies a method for the identification of LTI systems based on the nuclear norm of the Hankel matrix of the model - termed NucID. The nuclear norm has been put forward as a convex proxy to a class of rank-constraints that are hard to work with. The rationale for investigating such approach is that the estimate is more exible/robust in case of low Signal-to-Noise Ratios (SNRs), and other noisy effects in the data. This paper explores the formalisation, gives numerical results and brings up other issues for stimulating the discussion on the use of such approaches. (C) 2015, IFAC (International Federation of Automatic Control) Hosting by Elsevier Ltd. All rights reserved.