IEEE TRANSACTIONS ON AUTOMATIC CONTROL, VOL. 56, NO. 4, APRIL 2011 895
as stable by our algorithm. But again, we see that the roots of the poly-
nomials detected as stable come close to the stability boundary, thus we
conclude again that the constraints (19) are not too conservative. The
roots of all considered stable polynomials are shown in Fig. 3.
VI. C
ONCLUSION
In this technical note, the problem of characterization of stability
domain in coefficients space of a polynomial has been considered. The
proposed characterization generalizes the existing results both for con-
nected and disjoint stability regions. A new LMI formulation of sta-
bility domain in coefficients space is proposed for a very general sta-
bility region. The applicability of the method is demonstrated on two
problems.
REFERENCES
[1] O. Bachelier, D. Henrion, B. Pradin, and D. Mehdi, “Robust root-clus-
tering of a matrix in intersections or unions of subregions,” SIAM J.
Control and Optimiz., vol. 43, pp. 1078–1093, 2004.
[2] S. P. Bhattacharyya, H. Chapellat, and L. H. Keel, Robust Control: The
Parametric Approach. Upper Saddle River: Prentice-Hall, 1995.
[3] N. Cohen, A. Levkovich, E. Zeheb, and P. de Oliveira, “Frequency re-
sponse envelopes of a family of uncertain discrete-time systems,” Cir-
cuits, Syst., Signal Process., vol. 22, pp. 19–41, 2003.
[4] Y. Dolgin and E. Zeheb, “LMI stability constraints for disjoint stability
regions using Rouche theorem,” in Proc. IEEE CDC & CCC, Shanghai,
Dec. 16–18, 2009.
[5] D. Henrion, D. Peaucelle, D. Arzelier, and M. Sebek, “Ellipsoidal
approximation of the stability domain of polynomial,” IEEE Trans.
Autom. Control, vol. 48, no. 12, pp. 2255–2259, Dec. 2003.
[6] D. Henrion, M. Sebek, and V. Kucera, “Positive polynomials and ro-
bust stabilization with fixed-order controllers,” IEEE Trans. Autom.
Control, vol. 48, no. 7, pp. 1178–1186, Jul. 2003.
[7] O. Herrmann and W. Schuessler, “Design of nonrecursive digital filters
with minimum phase,” Electron. Lett., vol. 6, no. 11, pp. 329–330, May
1970.
[8] J. Hoagg and D. Bernstein, “Robust stabilization of discrete-time sys-
tems,” in Proc. IEEE CDC Conf., 2004, vol. 3, pp. 2346–2351.
[9] T. Iwasaki and S. Hara, “Generalized KYP lemma: Unified frequency
domain inequalities with design applications,” IEEE Trans. Autom.
Control, vol. 50, no. 1, pp. 41–59, Jan. 2005.
[10] L. Keel and S. Bhattacharyya, “Robust, fragile, or optimal,” IEEE
Trans. Autom. Control, vol. 42, pp. 1098–1105, 1997.
[11] P. Khargonekar, I. Petersen, and K. Zhou, “Robust stabilization of
uncertain linear systems: Quadratic stabilizability and
H
control
theory,” IEEE Trans. Autom. Control , vol. 35, pp. 356–361, 1990.
[12] R. Moses and D. Liu, “Determining the closest stable polynomial to an
unstable one,” IEEE Trans. Signal Process., vol. 39, no. 4, pp. 901–906,
Apr. 1991.
Recursive Identification for MIMO Hammerstein Systems
Xing-Min Chen and Han-Fu Chen, Fellow, IEEE
Abstract—This technical note considers the recursive identification for
the multi-input and multi-output (MIMO) Hammerstein system with
internal noise and observation noise and with linear part being an ARX
system. With the help of the generalized Yule-Walker equation and the
correlations of system signals, the recursive algorithms are proposed for
estimating unknown coefficients of the linear subsystem, while the system
nonlinearity is recursively estimated by using the multivariable kernel
functions. Strong consistency of the estimates is proved under reasonable
conditions, and a simulation example is provided.
Index Terms—Kernel estimate, MIMO Hammerstein system, recursive
identification, stochastic approximation.
I. I
NTRODUCTION
The Hammerstein system is composed of a static nonlinearity fol-
lowed by a linear subsystem. This type of models is widely used in
modeling practical systems, for example, the PH neutralization pro-
cesses, the control of distillation columns etc. For recent decades much
attention has been paid to identification of Hammerstein systems. How-
ever, most of existing results are on single-input single-output (SISO)
systems, not too many papers on identification of MIMO Hammerstein
systems.
When the parametric approach [1], [2] is applied, the nonlinearity of
the system is often written as a linear combination of basis functions
with unknown coefficients. Then, the problem of identification reduces
to parameter estimation. On the basis of this representation, identifica-
tion of MIMO Hammerstein and Wiener systems expressed in the state
space form is considered in [3] by using the subspace method. Identifi-
cation algorithms for MIMO Hammerstein systems with nonlinearities
modeled in different ways are given in [4] by applying the multivariable
output error state space method. Identification of MIMO Hammerstein
systems in the state-space representation with nonlinear feedback con-
trol is discussed in [5] with the help of a subspace method. In [6], iden-
tification of the MIMO Hammerstein system is considered for linear
subsystems being of FIR and in the state space form, respectively, but
consistency is still not well established. The parametric approach to
identifying MIMO systems usually requires a large number of basis
functions be involved, and they are not easy to be adequately selected.
The nonparametric approach [7], [8] to identification of Hammer-
stein systems requires not much a priori information concerning the
nonlinearity of the system. Since the curse of dimensionality may occur
when identifying high-dimensional systems [8], [9], the approxima-
tion by additive functions may be applied and the resulting system
is the multichannel Hammerstein system which is widely applied in
multisensor systems, power systems, physiology and nervous systems
[8]. An early work [10] considered the nonparametric identification for
two-channel Hammerstein systems. A nonparametric method is pro-
posed in [11] to identify multi-input Hammerstein systems, while for
Manuscript received April 09, 2010; revised April 20, 2010 and September
09, 2010; accepted November 29, 2010. Date of publication December 23,
2010; date of current version April 06, 2011. The work was supported by the
NSFC under Grants 60821091 and 60874001. Recommended by Associate
Editor E. Weyer.
The authors are with the Key Laboratory of Systems and Control of CAS,
Institute of Systems Science, AMSS, Chinese Academy of Sciences, Beijing
100190, China (e-mail: xmchen@amss.ac.cn; hfchen@iss.ac.cn).
Digital Object Identifier 10.1109/TAC.2010.2101691
0018-9286/$26.00 © 2010 IEEE