Elsevier

Systems & Control Letters

Volume 59, Issues 3–4, March–April 2010, Pages 167-172
Systems & Control Letters

On the well-posedness of multivariate spectrum approximation and convergence of high-resolution spectral estimators

https://doi.org/10.1016/j.sysconle.2010.01.001Get rights and content

Abstract

In this paper, we establish the well-posedness of the generalized moment problems recently studied by Byrnes–Georgiou–Lindquist and coworkers, and by Ferrante–Pavon–Ramponi. We then apply these continuity results to prove the almost sure convergence of a sequence of high-resolution spectral estimators indexed by the sample size.

Introduction

Consider a linear, time invariant system x(t+1)=Ax(t)+By(t),ACn×n,BCn×m, with the transfer function G(z)=(zIA)1B, where A is a stability matrix, B is full column rank, and (A,B) is a reachable pair. Suppose that the system is fed with a m-dimensional, zero-mean, wide-sense stationary process y having a spectrum Φ. The asymptotic state covariance Σ of the system (1) satisfies: Σ=GΦG. Here and in the following, G(z)=G(z1), and integration takes place over the unit circle with respect to the normalized Lebesgue measure dϑ/2π. Let S+m×m(T) be the family of bounded, coercive, Cm×m-valued spectral density functions on the unit circle. Hence, ΦS+m×m(T) if and only if Φ1S+m×m(T). Given a Hermitian and positive-definite n×n matrix Σ, consider the problem of finding ΦS+m×m(T) that satisfies (3), i.e., that is compatible with Σ. This is a particular case of a moment problem. In the last ten years, much research has been produced, mainly by the Byrnes–Georgiou–Lindquist school, on generalized moment problems [1], [2], [3], [4], [5], and analytic interpolation with complexity constraint [6], and their applications to spectral estimation [7], [8], [9] and robust control [10]. It is worth recalling that two fundamental problems of control theory, namely the covariance extension problem and the Nevanlinna–Pick interpolation problem of robust control, can be recast in this form [5].

Eq. (3), where the unknown is Φ, is also a typical example of an inverse problem. Recall that a problem is said to be well posed, in the sense of Hadamard, if it admits a solution, such a solution is unique, and the solution depends continuously on the data. Inverse problems are typically not well posed. In our case, there may well be no solution Φ, and when a solution exists, there may be (infinitely) many. It was shown in [11], that the set of solutions is nonempty if and only if there exists HCm×n such thatΣAΣA=BH+HB. When (4) is feasible with Σ>0, and there are infinitely many solutions Φ to (3). To select a particular solution it is natural to introduce an optimality criterion. For control applications, however, it is desirable that such a solution be of limited complexity. It should namely be rational and with an a priori bound on its MacMillan degree. One of the great accomplishments of the Byrnes–Georgiou–Lindquist approach is having shown that the minimization of certain entropy-like functionals leads to solutions that satisfy this requirement. In [11], Georgiou provided an explicit expression for the spectrum Φˆ that exhibits a maximum entropy rate among the solutions of (3).

Suppose now that some a priori information about Φ is available in the form of a spectrum ΨS+m×m(T). Given G, Σ, and Ψ, we now seek a spectrum Φ, which is the closest to Ψ in a certain metric, among the solutions of (3). Paper [5] deals with such an optimization problem in the case when y is a scalar process. The criterion there is the Kullback–Leibler pseudo-distance from Ψ to Φ. A drawback of this approach is that it does not seem to generalize to the multivariable case. This motivated us to provide a suitable extension of the so-called Hellinger distance with respect to which the multivariable version of the problem is solvable (see [12], [9]).

The main result of this paper is contained in Section 3. We show there that, under the feasibility assumption, the solution to the spectrum approximation problem with respect to both the scalar Kullback–Leibler pseudo-distance and the multivariable Hellinger distance depends continuously on Σ, thereby proving that these problems are well-posed. In Section 4 we deal with the case when only an estimate Σˆ of Σ is available. By applying the continuity results of Section 3, we prove a consistency result for the solutions to both approximation problems.

Section snippets

Spectrum approximation problems

In this section, we collect some background material on the spectrum approximation problems. The reader is referred to [11], [5], [12], [9] for a more detailed treatment.

Well-posedness of the approximation problems

In this section, we show that both the dual problems (12), (23) are well-posed, since their unique solution is continuous with respect to a small perturbation of Σ. The well-posedness of the respective primal problem then easily follows. All these continuity properties rely on the following basic result.

Theorem 3.1

Let A be an open and convex subset of a finite-dimensional Euclidean space V . Let f:AR be a strictly convex function, and suppose that a minimum point x̄ of f exists. Then, for all ε>0 ,

Consistency

So far we have shown that both the approximation problems admit a unique solution for all ΣPΓ, and that the solution is continuous with respect to the variations δΣRangeΓ. The necessity of a restriction to RangeΓ becomes crucial in the case when we only have an estimate Σˆ of Σ.

In line with the Byrnes–Georgiou–Lindquist theory, and following an estimation procedure we have sketched in [9], we want to use the above theory to provide an estimate Φˆ of the true spectrum of the process y.

Let G(z)

Conclusion and future work

In this paper, we have considered the constrained spectrum approximation problems with respect to both the Kullback–Leibler pseudo-distance (scalar case) and the Hellinger distance (multivariable case). The range of the operator Γ:ΦGΦG is the subspace of the Hermitian matrices that conveys all the structure that is needed from a positive-definite matrix in order to be an asymptotic covariance matrix of the system with transfer function G(z). As such, it is also a natural subspace to which

References (16)

  • C.I. Byrnes et al.

    A convex optimization approach to the rational covariance extension problem

    SIAM J. Control and Optim.

    (1999)
  • T. Georgiou

    Spectral analysis based on the state covariance: The maximum entropy spectrum and linear fractional parameterization

    IEEE Trans. Automat. Control

    (2002)
  • C.I. Byrnes et al.

    The generalized moment problem with complexity constraint

    Integral Equations Operator Theory

    (2006)
  • T. Georgiou

    Relative entropy and the multivariable multidimensional moment problem

    IEEE Trans. Inform. Theory

    (2006)
  • T. Georgiou et al.

    Kullback–Leibler approximation of spectral density functions

    IEEE Trans. Inform. Theory

    (2003)
  • A. Blomqvist et al.

    Matrix-valued Nevanlinna–Pick interpolation with complexity constraint: An optimization approach

    IEEE Trans. Automat. Control

    (2003)
  • C.I. Byrnes et al.

    A new approach to spectral estimation: A tunable high-resolution spectral estimator

    IEEE Trans. Signal Process.

    (2000)
  • T. Georgiou et al.

    A convex optimization approach to ARMA modeling

    IEEE Trans. Automat. Control

    (2008)
There are more references available in the full text version of this article.

Cited by (16)

View all citing articles on Scopus

Partially supported by the Ministry of Education, University, and Research of Italy (MIUR), under Project 2006094843: New techniques and applications of identification and adaptive control.

View full text