
Received: 27 January 2017 Revised: 12 September 2017 Accepted: 6 November 2017
DOI: 10.1002/acs.2848
RESEARCH ARTICLE
Homogenous polynomial H∞filtering for uncertain
discrete-time systems: A descriptor approach
T. Zoulagh1B. El Haiek1A. Hmamed1A. El Hajjaji2
1Laboratoire Signaux Systémes et
Informatique, Department of Physics,
Faculty of Sciences Dhar El Mahraz,
Universite Sidi Mohamed Ben Abdellah,
BP 1796 Atlas, 30000 Fes, Morocco
2Modeling Information System
Laboratory, UFR of Sciences, University of
Picardie Jules Verne, 33 Rue St Leu, 80000
Amiens, France
Correspondence
T. Zoulagh, Laboratoire Signaux Systémes
et Informatique, Department of Physics,
Faculty of Sciences Dhar El Mehraz,
Universite Sidi Mohamed Ben Abdellah,
BP 1796 Atlas, 30000 Fes, Morocco.
Summary
This paper focuses on the robust full-order H∞filter design for linear
discrete-time systems with polytopic uncertainties. Less conservative robust H∞
filter design procedures are given in terms of linear matrix inequality con-
straints. By using a descriptor approach, 2 sufficient conditions for the H∞filter
analysis and design are proposed via linear matrix inequalities. The homogenous
polynomial parameter-dependent Lyapunov functions are used for the asymp-
totic stability analysis of the system error. Finally, to demonstrate the efficiency
of the proposed approach, simulation results with comparative studies are used.
KEYWORDS
descriptor representation, H∞filtering, homogenous polynomial approach, linear matrix inequalities
(LMIs), polytopic uncertainty
1INTRODUCTION
Filtering or estimating signals that are perturbed by noise have been a great field of research in control and signal the-
ory since the pioneering work of Kalman, the so-called Kalman filter.1Many methods have been used in filtering and
control problems, some of them and the most used in recent years are H2,H∞, and the energy-to-peak filtering.2-9The
advantage of H∞filtering in comparison with the traditional Kalman filtering methods is that no statistical assumptions
on the exogenous signals are needed. These approaches are used mainly when dealing with linear systems subject to
uncertainties.
Thus, the filter design problem for discrete time has been tackled in other works for linear systems,10-12 systems with
stochastic incomplete measurement and mixed delays,13 fuzzy systems,14 and fuzzy neural network systems with stochas-
tic jumps and time delays.15 Furthermore, the robust filtering design can be set as a convex optimization formulation
and may be solved via the linear matrix inequality (LMI) tools, which are often considered as powerful tools in analysis,
control, and filter design, as the LMI control Toolbox,16,17 YALMIP,18 RoLMIP (Robust LMI Parser),19 and some effective
semidefinite programming solvers such as SeDuMi20 and MOSEK.21
In order to reduce the conservatism of LMI conditions for uncertain linear systems, many works have been proposed in
the literature, including quadratic stability22-25; parameter-dependent Lyapunov functions were tackled;12,24,26 and lately,
several results on a robust filter design through homogenous polynomial parameter-dependent Lyapunov matrices of
arbitrary degrees have been studied.10-12,27-29 For instance, sufficient filter design conditions for time-delay systems were
proposed in the work of Gao et al.30 However, most of these results may be conservative. Therefore, the main contribution
is the combination, for the first time, between the homogenous polynomial approach, discrete systems, and descriptor
representation to get less conservative results. The motivation of this work concerns the development of the less conser-
vative conditions with a smaller bound for H∞performance criteria, ie, bounds that are aftermost to the worst-case norms
of the uncertain systems with the robust filter for polytopic uncertain discrete-time systems.
378 Copyright © 2017 John Wiley & Sons, Ltd. wileyonlinelibrary.com/journal/acs Int J Adapt Control Signal Process. 2018;32:378–389.