Optimal angle reduction -- a behavioral approach to linear system approximation (Q5954131)

The main aim of the paper under review is to investigate an optimal model approximation problem for the class of linear time-invariant systems in discrete time. For this purpose, the concept of past-future links of a dynamical system is introduced as a generalization of the Hankel singular vectors. This is done in Section 4. Previously in Section 2, the relevant model class, some basic definitions and preliminaries are presented, meanwhile in Section 3, the model approximation problem is stated in terms of a so-called ``optimal angle reduction problem''. Such a problem is solved for reduction of the degree of a given ``to be approximated system'' by one. This is carried out in Section 5, containing the main original contribution of the paper (Theorem 5.1). This reduction technique is described and implemented in Section 6 in terms of canonical isometric state representations. On the other hand, a comparison with other related techniques (global total least squares method, optimal Hankel norm reduction, coprime factorization) is given in Section 7. Finally, in Section 8 a simple illustrative numerical example is provided. This is an extremely self-contained paper with a high tutorial value.