PD observer parametrization design for descriptor systems

https://doi.org/10.1016/j.jfranklin.2005.03.001Get rights and content

Abstract

Based on the design of proportional and derivative (PD) observers, a new parametrization of all observers for descriptor systems is developed. Without using any additional transformations, the present observer parametrization is convenient and efficient to be performed. The connections between the proposed parametrization and the known results are established. For descriptor systems with bounded input disturbances, the estimation error dynamics are parametrized with respect to the disturbance vectors. An example is included to illustrate the parametrization procedures.

Introduction

Descriptor systems, which are also referred to singular systems, differential algebraic systems, generalized dynamical systems, implicit systems, etc., often appear in electronic circuits, aircraft systems, chemical engineering, economic systems, etc. During the past 30 years, many results have been reported for descriptor systems, see e.g. [1], [2]. The factorization approach has proven to be a powerful tool in multivariable control system analysis and synthesis [3]. Recently, the factorization approach has been introduced to investigate descriptor systems. Specifically, the algebraic framework for studying descriptor systems with the factorization approach has been established [4], [5], [6]. The problems of Bezout factorizations and controller parametrizations have been solved for descriptor systems, see [6], [7], [8], [9], [10].

As dual results of controller parametrizations, the observer parametrizations have been developed effectively to solve the robust observer design and the optimal H2/H filtering problems for normal systems [11], [12], [13], [14]. For descriptor systems, the problem of observer parametrization is first studied by Minamide et al. [15], is then discussed by Gao et al. [16] and Gao [17]. Specifically, in the work by Minamide et al. [15], by using descriptor standard form [18], a parametrization of observers is derived. In Ref. [16], a class of reduced-order observers is characterized on the basis of restrictive equivalent transformations. The work in [17] is related to descriptor proportional observers, which is a natural extension of the work [11] for normal systems. However, it is not difficult to find that additional transformations in [15], [16] make the computations involved. Moreover in [15], [17], it is not systematic to find a gain to make the estimation error dynamics both stable and impulse free for descriptor observers.

In this study, a new observer parametrization related to the design of PD descriptor observer is developed. Not any additional transformations are introduced in the derived process. Moreover, the derivative gain and the proportional gain are designed systematically to ensure the estimation error dynamics to be impulse free and stable. This kind of design, therefore, makes the present parameterization convenient and efficient to be performed. The interesting connections between the present parametrization and those given by Minamide et al. [15] and Gao [17] are also established. When disturbances occur in the plant, the estimation error would never be zero. In this case, the estimation error dynamics are also parametrized, which are useful for designing optimal robust observers for descriptor systems. An example is given to illustrate the parametrization procedures.

Section snippets

Preliminaries

Some notations are first given here. R denotes the field of real numbers, Rn denotes the n-dimensional Euclidean space, C+ denotes the closed, complex right half-plane (or the set of all unstable and finite complex numbers), RH denotes the set of proper stable, real-rational functions, RHk×m denotes the class of matrices of size k×m with all elements in RH, AT represents the transpose of the matrix, I denotes the unity matrix, and 0 denotes scalar zero or a null matrix with compatible size.

New observer parametrization design

In this section, a new observer parametrization is proposed and its interesting connections compared to the known results are shown.

Robust observer design

When disturbances occur, the estimation error would never be zero. In this case, it is desired that the estimation error is kept as small as possible. We now characterize the relation between the estimation error and the disturbance vector d(t).

Theorem 3

Given plant (1)–(3) under Assumption 1, the set of all achievable transfer function matrices of the estimation errors w(t)-w^(t) in the presence of the disturbance vector d(t) is parametrized asTde(s)T1(s)-Q(s)T2(s),where Q(s)RHk×p, and

Proof

Without loss

Example

Consider the system given in the form (1)–(3). The corresponding matrices are E=1000,A=0120,B=-11,Bd=-10.1,C=(1,-1),D=0,Dd=1,F=(2,1).

According to the well-known criteria [1], one can show easily that this plant comprises impulsive modes, but satisfies Assumption 1. Choose L1=01such that (E+L1C) is non-singular, and then select L2=-1-3such that detΔ(s)=det[s(E+L1C)-(A+L2C)] is a stable polynomial with zeros {-1,-1}. The observer parametrization in the form (16) becomesw^(s)=[Γ(s)-Q(s)N˜(s)]u(s)+

Conclusions

A new parametrization of all causal observers for descriptor systems has been obtained on the basis of the PD observer design. The proposed parametrization is more convenient and efficient to be performed compared with the previous work. The estimation error dynamics of observers for descriptor systems with respect to unknown input disturbances have also been parametrized. An example is included to illustrate the parametrization procedures. The proposed observer parametrization provides useful

Acknowledgments

This work is supported partially by the Alexander Humboldt Foundation (IV-CHM/1117303 STP). The author would like to thank the referees for their helpful comments.

References (20)

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

Cited by (35)

  • A fixed-time observer for discrete-time singular systems with unknown inputs

    2019, Applied Mathematics and Computation
    Citation Excerpt :

    The singular system, also named descriptor system, differential-algebraic system or generalized state-space system, is a type of special dynamic system which has the superiority of describing systems with algebraic constraints such as circuit system, multi-body mechanical system, economic system, and so on [17–20]. So far, a great amount of achievements on the singular system observer designs have been made in the literature [17,21–27]. Note that for the singular system observer designs, most of the results were focused on the continuous-time system, and only a little work for discrete-time ones has been reported [28–33].

  • New insight into reachable set estimation for uncertain singular time-delay systems

    2018, Applied Mathematics and Computation
    Citation Excerpt :

    In this paper, we present a new insight into the reachable set estimation for a class of uncertain singular time-delay systems, where the considered singular system does not need to be decomposed and a proportional-derivative state feedback controller (PDSFC) is designed to guarantee all the reachable states of the considered closed-loop singular system contained by an ellipsoid. It is worth mentioning that though PDSFCs have been widely used in singular systems; see, e.g. [25–30], we first use them to study the problem of reachable set estimation for uncertain singular systems with time-varying delays. The main contributions are stated as follows.

  • Observer design for one-sided Lipschitz descriptor systems

    2016, Applied Mathematical Modelling
    Citation Excerpt :

    Descriptor systems are also called singular systems, for which the initial condition always satisfies some algebraic constraints in contrast to the traditional ODE-based models. Observer design for these descriptor systems to estimate their state vectors has vast applications in controller synthesis, filtering, and fault diagnosis [2,3,6,10–13]; however, extra efforts are required to apply the generalized Lyapunov theory based on LaSalle's principle to ensure convergence of the estimation error. Considerable research activities concerning observer design for linear (see for example, [1–3,10,12,14]) as well as nonlinear (see, for example, [3,6,11,13]) descriptor systems have been carried out in the past decade.

  • Full- and reduced-order observer design for rectangular descriptor systems with unknown inputs

    2015, Journal of the Franklin Institute
    Citation Excerpt :

    An observer is a mathematical realization which uses the input and output informations of a given system and its output asymptotically approaches the true state values of the given system. Observer design problem for normal systems has received a great attention in the literature [8–14] and the techniques used for them have been extended successfully to descriptor linear [15–22] and nonlinear [23–26] systems. There are many practical situations where control systems arise with noise or disturbances, so the problem of observer design for systems with unknown inputs is of great importance.

  • Observer design for a class of nonlinear descriptor systems

    2013, Journal of the Franklin Institute
    Citation Excerpt :

    Based on the complete parametric solution to the generalized Sylvester matrix equations, a parametric design approach for the proposed GPI observers was established. In [17,18], proportional derivative (PD) observers were constructed by introducing derivative terms. Furthermore, the so-called generalized proportional integral derivative (GPID) observers were proposed for linear descriptor systems by introducing a derivative term to the GPI observers [19].

View all citing articles on Scopus
View full text