next up previous contents
Next: Linear Model Varying Systems Up: The Ingredients Previous: The Ingredients

Approximation in SCV

Of particular interest is a theorem due to Jim Agler [A] and extensions of it [AM] [BT]. Let tex2html_wrap_inline813 denote the unit polydisk

displaymath815

in tex2html_wrap_inline817 the space of n complex variables. Let tex2html_wrap_inline821 denote the functions analytic on tex2html_wrap_inline813 and bounded in supremum norm by 1. Also tex2html_wrap_inline825 denotes the tex2html_wrap_inline827 matices. The Nevannlina -Pick type interpolation problem on tex2html_wrap_inline829 ,the bi-disk, is:

Given a finite set of points tex2html_wrap_inline831 in the bi-disk and numbers tex2html_wrap_inline833 in the complex plane C, find a function f in 2 which takes the values tex2html_wrap_inline833 at the points tex2html_wrap_inline831 .
Such problems arise naturally when one is trying to find a function f analytic on the bi-disk which has a supremum norm tex2html_wrap_inline843 and which has the form

  equation89

for some tex2html_wrap_inline845 analytic on the bi-disk. This is because in the typical situation on the bi-disk the zero set tex2html_wrap_inline847 of a function g in 2 is empty or is the union of finitely many one complex dimensional surfaces. Thus tex2html_wrap_inline851 and tex2html_wrap_inline853 have zero sets intersecting (transversally) in only a finite number of points tex2html_wrap_inline855 and at these points f is constrained to take values

  equation94

regardless of what h is. An expert in traditional H tex2html_wrap_inline861 control will recall that this problem on the ordinary disk (as opposed to the bi-disk) will have tex2html_wrap_inline863 and is the model matching problem. Consequently, we shall call problem (1) the bi-disk model matching problem.

   thm101

The proof is in principle constructive, see [AM] [BT], though considerable effort would probably be required to refine it for computer implementation. While Theorem 2.1 is only stated for f a scalar valued function, it holds in a very similar form f is a matrix valued function, see [BT]. However, this article involves only SISO systems, primarily for ease of pedagogy.

In this article we study a control problem which leads to a bidisk model matching problem with tex2html_wrap_inline863 . Thus the interpolating constraints form an infinite set. The implications of this will be discussed later as they arise


next up previous contents
Next: Linear Model Varying Systems Up: The Ingredients Previous: The Ingredients

Bill Helton
Wed Dec 15 17:24:28 PST 1999