PI observer for robust state estimation of

multiple model systems with delayed

measurements
Rodolfo Orjuela, Benoı̂t Marx, José Ragot
and Didier Maquin

Centre de Recherche en Automatique de Nancy

UMR 7039, Nancy-Université, CNRS
2, Avenue de la Forêt de Haye
54516 Vandœuvre-lès-Nancy, France

Abstract: This paper deals with the problem of state estimation for nonlinear systems modelled
via multiple models and subject to delayed measurements. In contrast to the most popular
results found in the multiple model literature, we consider heterogeneous multiple models, also
known as decoupled multiple model, which allow to use submodels whose state space dimension
can be different. On the basis of this multiple model a proportional-integral observer (PIO) is
designed in order to cope with the robust state estimation with respect to disturbances. Sufficient
delay-independent conditions for ensuring robust performances (attenuation level) and dynamic
performances (exponential convergence) of the estimation error are provided in terms of LMIs
using the Lyapunov-Krasovskii method. The validity of the proposed methodology is illustrated
by an academic example.

Keywords: heterogeneous multiple models; state estimation; proportional-integral observer;

delayed measurements

1. INTRODUCTION admissible trajectories of the system must be taken into

consideration.
Context. Time-delay or dead time is frequently encoun- Tools and proposed method. Nonlinear behaviour of a
tered in several real-world processes, such as chemical system can accurately be represented using the multi-
and thermal processes, electrical and communications net- ple model approach [Murray-Smith and Johansen, 1997,
works, etc. Indeed, any system has an amount of time, Boukhris et al., 1999]. Multiple models have been rec-
more or less important, between its expected response ognized as a popular and a powerful modelling tool of
and its actual response. In many industrial applications complex nonlinear systems. In this approach the operating
time-delays must unavoidably be considered as an inher- space of the nonlinear system under investigation is de-
ent part of the system, and consequently, they must be composed into a number of operating zones, each of them
accounted for in the mathematical model of the system being characterised by a submodel of reduced dimension.
under investigation [Smith and Corripio, 1985]. Therefore, The interpolation of submodels with the help of weighting
considerable attention has been paid in the past decades functions, associated to each operating zone, allows the
to systems with delays and appropriate theoretical tools, global representation of the nonlinear system.
in the frequency and in the time domain, for the analysis,
control and state estimation are well established [Bliman, At this point, it is important to point out that the interpo-
2001, Sename, 2001, Richard, 2003]. Among them, the well lation of submodels can be operated using many schemas
known Lyapunov-Krasovskii functional method is largely [Filev, 1991, Gregorcic and Lightbody, 2008], among these
adopted for stability analysis of linear or nonlinear delayed realisations two schemas can be distinguished: via an
systems. This method provides generally sufficient condi- interpolation of parameters (e.g. Takagi-Sugeno multiple
tions in a LMI form which can be solved in polynomial- model ) or via an interpolation of outputs (e.g. decoupled
time using appropriated standard convex optimization al- multiple model). Many contributions concern the analysis,
gorithms, such as [Boyd et al., 1994, Sturm, 1999]. the control and the state estimation of Takagi-Sugeno
models (with or without delays) [Murray-Smith and Jo-
On the other hand, complex industrial systems generally hansen, 1997, Babuska, 1998, Tanaka and Wang, 2001,
present a nonlinear dynamic behaviour in particular if the Lin et al., 2007], much less are devoted to observer design
whole operating range (global behaviour) is considered using decoupled multiple models. In this work, the latter
instead of a reduced operating range in the neighbourhood realisation will be employed for dealing with the problem of
of an operating point (local behaviour). Hence, nonlinear state estimation for nonlinear systems with delayed mea-
models must be used to deal effectively with the global surements as an extension of the ideas proposed previously
modelling of these systems. This situation is commonly [Orjuela et al., 2007, 2008].
encountered in the state estimation problems because all
Let us notice finally that the basis of the multiple model 2. DECOUPLED MULTIPLE MODEL
approach can be easily related to the hybrid system philos-
ophy in which the interaction of a set of submodels changes The structure of the decoupled multiple model can be
according to a dynamic law. However, in the multiple viewed as a parallel interconnection of several submodels
models smooth transitions between the submodels can via a weighted sum of their outputs [Gatzke and Doyle III,
be considered, instead of crisp transitions (switching), for 1999]. By considering a state space representation, this
more flexibility in modelling stage. structure takes the following form [Orjuela et al., 2008]:
Related work. Uppal et al. [2006] have shown a successful
implementation of this structure in order to cope with the ẋi (t) = Ai xi (t) + Bi u(t) + Di w(t) , (1a)
state estimation for fault detection and isolation. How- yi (t) = Ci xi (t) , (1b)
ever, only a note of the convergence estimation error is L
proposed in this work and neither any analytic proof of
X
z(t) = µi (ξ(t))yi (t) , (1c)
the convergence of the estimation error is given, nor the
i=1
time delay is taken into account. On the other hand, the
state estimation in presence of delayed measurements is y(t) = z(t − τ (t)) + W w(t) . (1d)
ni p
investigated in Orjuela et al. [2007]. Sufficient conditions, where xi ∈ R and yi ∈ R are respectively the state
under a LMI form, are given for ensuring the state estima- vector and the output of the ith submodel; u ∈ Rm is the
tion convergence towards zero. However, the robust state multiple model input, z ∈ Rp is the multiple model output
estimation with respect to disturbances (e.g. noise) acting and y ∈ Rp the measured output subject to a variable
on the outputs has not been addressed in this work. time-delay τ (t) and a perturbation w ∈ Rr . The matrices
Ai ∈ Rni ×ni , Bi ∈ Rni ×m , Di ∈ Rni ×r , Ci ∈ Rp×ni and
We introduce a so-called proportional-integral observer W ∈ Rp×r are known and constant.
(PIO) which provides very interesting robustness proper-
ties with respect to perturbations [Beale and Shafai, 1989, Assumption 1. The perturbation is bounded energy sig-
Weinmann, 1991, Busawon and Kabore, 2001]. This ob- nal, i.e. kw(t)k22 < ∞.
server is characterised by the use of two corrective injection
The variable time-delay τ (t) acting on the output is
terms, proportional and integral, instead of the conven-
perfectly modelled by a continuous function that satisfies
tional proportional correction frequently employed in the
the following conditions [Wu et al., 2004]:
Luenberger observer and/or in its classical extensions. The 
PIO has also been successfully employed in the synchro- 0 ≤ τ (t) ≤ τmax
(2)
nization of a chaotic system by Hua and Guan [2005]. τ̇ (t) ≤ β < 1
The extension of the PIO design, based on dissipativity where τmax and β are constant. This time-delay can be
framework, to a particular nonlinear system whose non due for example to the inherent nature of the system, the
linearity is assumed to satisfy a sector bounded constraint, sensor reactivity or the transmission data time-delay.
has been recently proposed in Jung et al. [2007]. However,
the PIO design for standard delayed linear system and for The relative contribution of each submodel to the global
delayed multiple models (Takagi-Sugeno and/or decoupled model according to the current operating point of the
multiple models) seems poorly investigated in the litera- system is quantified by the weighting functions µi (ξ(t)).
ture. They are associated to each operating zone and satisfy
the following convex sum properties:
Contribution and paper organisation. In this communica-
tion, the PIO design for decoupled multiple models with L
X
delayed measurements is investigated. The outline of this µi (ξ(t)) = 1 and 0 ≤ µi (ξ(t)) ≤ 1 , ∀i = 1...L, ∀t . (3)
paper is as follows. A brief description of the decoupled i=1
multiple model is presented in section 2. The state estima- Besides, the current operating point is taken into account
tion problem is presented in section 3 and sufficient con- by means of the so-called decision variable ξ(t) which is
ditions, on the basis of the Lyapunov-Krasovskii method, a characteristic variable of the system (e.g. inputs and/or
are established in order to ensure the stability of the PIO. measured variables).
Finally, in section 4, an academic example illustrates the
state estimation of a decoupled multiple model. In this multiple model the contribution of each submodel
is taken into account via a weighted sum of the submodel
Notations: The following standard notations will be used outputs. Consequently, heterogeneous submodels (i.e. sub-
throughout the paper: P > 0 (P < 0) denotes a positive models of different dimensions) can be considered in the
(negative) definite matrix P ; X T denotes the transpose of modelling stage as a supplementary degree of flexibility
matrix X; I is the identity matrix of appropriate dimension and generality. Therefore, this structure is well adapted
and diag{} is a block diagonal matrix of appropriate for modelling strongly nonlinear systems whose structure
dimension. The L2 −norm of a signal, quantifying its varies with the operating regime, for example when the
R∞
energy is denoted and defined by ke(t)k22 = eT (t)e(t)dt. complexity of the dynamic behaviour is not uniform in
0 the operating range [Gregorcic and Lightbody, 2008].
We shall simply write µi (ξ(t)) = µi (t) and x(t − τ (t)) =
x(∇) where τ (t) > 0 is a variable time-delay. 3. STATE ESTIMATION PROBLEM

In a previous work [Orjuela et al., 2007], the state esti-

mation of a decoupled multiple model subject to delayed
measurements has been investigated on the basis of a general action, for example, by considering the filtered
proportional observer given by : output signal as follows:

x̂˙ i (t) = Ai x̂i (t) + Bi u(t) + Ki (y(t) − ŷ(t)) , (4a) ẏI (t) = N yI (t) + y(t) (7)
L
X where the matrix N is a fading effect coefficient matrix
ẑ(t) = µi (ξ(t))Ci x̂i (t) , (4b) that regulates the transient response of yI (t) [Jung et al.,
i=1 2007].
ŷ(t) = ẑ(t − τ (t)) , (4c) Let us notice that the equations (5) and (7) can be
ni ×p th
where Ki ∈ R is the gain associated to i observer. gathered as follows:
Sufficient conditions for ensuring the exponential con-  
T T T
vergence towards zero of the estimation errors ei (t) = ẋa (t) = C 1 Ã C 1 + C 2 N C 2 xa (t) + C 2 C̃(∇) C 1 xa (∇)
xi (t) − x̂i (t) have been established. Note however that  
disturbances acting on the systems are not taken into + C 1 B̃u(t) + C 2 W + C 1 D̃ w(t) , (8a)
consideration. Consequently, the sensitivity problem of the
T
state estimation with respect to disturbances has not been y(t) = C̃(∇)C 1 xa (∇) + W w(t) , (8b)
previously addressed in this work. On the other hand, the T
proportional observer offers only one degree of freedom yI (t) = C 2 xa (t) (8c)
Ki for reducing the influence of the noise and providing at where the augmented vector xa ∈ R (n+p)
and the constant
the same time good dynamic performances (two antagonist bloc matrices C 1 and C 2 of appropriate dimensions are
design goals). respectively given by:
In this paper, the proportional observer (4) is replaced by      
x(t) I 0
a more general observer, known as proportional-integral xa (t) = , C 1 = (n×n) , C 2 = (n×p) . (9)
observer (PIO), which provides very interesting robustness yI (t) 0(p×n) I(p×p)
properties with respect to perturbations. In particular,
good trade-off between dynamic and robust performances The structure of the proposed PIO can be obtained on the
of the observer can be obtained with the PIO because two basis of the augmented form (8):
degrees of freedom are available for the observer design.  T T
 T
x̂˙ a (t) = C 1 Ã C 1 + C 2 N C 2 x̂a (t) + C 2 C̃(∇) C 1 x̂a (∇)
3.1 Preliminaries
+K̃P (y(t) − ŷ(t)) + K̃I (yI (t) − ŷI (t)) + C 1 B̃u(t) , (10a)
T
For the simplicity of manipulations, the decoupled multi- ŷ(t) = C̃(∇)C 1 x̂a (∇) , (10b)
ple model (1) is rewritten in the following compact form T
[Orjuela et al., 2007, 2008]: ŷI (t) = C 2 x̂a (t) . (10c)
where the gains K̃P and K̃I must be designed and the
ẋ(t) = Ãx(t) + B̃u(t) + D̃w(t) , (5a) matrices Ã, B̃ and C̃(t) have been previously defined
z(t) = C̃(t)x(t) , (5b) in (6). Let us notice that an additional injection term,
given by the filtered (or integral) output estimation error
y(t) = z(∇) + W w(t) , (5c) yI (t) − ŷI (t), is included in the dynamic equation of the
where: observer. Hence some degrees of freedom are in this way
introduced for observer design. The use of the two gains
L
T X K̃P (proportional action) and K̃I (integral action) is at
x(t) = xT1 (t) · · · xTi (t) · · · xTL (t) ∈ Rn , n =

ni , (6a) the origin of the name Proportional-Integral Observer.
i=1
(n×n) 3.3 Problem formulation
à = diag {A1 · · · Ai · · · AL } ∈ R , (6b)
T T (n×m)
 T T

B̃ = B1 · · · Bi · · · BL ∈R , (6c) The dynamics of the estimation error
T
D̃ = D1 · · · Di T · · · DL T ∈ R(n×r) ,
 T 
(6d)
(p×n)
ea (t) = xa (t) − x̂a (t) (11)
C̃(t) = [µ1 (t)C1 · · · µi (t)Ci · · · µL (t)CL ] ∈ R . (6e)
can be easily established by considering the gathered forms
The matrix C̃(t) is time-varying because the weighting of the decoupled multiple model (8) and the PIO (10):
functions are taken into consideration in this matrix.
ėa (t) = Λ1 ea (t) + Λ2 (∇)ea (∇) + Λ3 w(t) , (12)
3.2 PIO structure where
In a PIO, the classic used proportional action given by the T T T
Λ1 = C 1 Ã C 1 + C 2 N C 2 − KI C 2 , (13a)
gains Ki in (4) is replaced by the use of two correction
T
actions: proportional and integral. For this purpose, a Λ2 (∇) = C 2 − KP C̃(∇) C 1 , (13b)
Rt

supplementary integral variable yI (t) = y(η)dη must Λ3 = C 2 − KP W + C 1 D̃ . (13c)
0
be taken into consideration in the PIO architecture. Note Let us remark, from equations (13a) and (13c), that
however that this integration can be replaced by a more the proportional gain KP can be used to modulate the
influence of the disturbances on the estimation error ea (t) min γ subject to
and the observer dynamics can be improved with the help  T T
H H + Λ1 P + P Λ 1 + Q Λ̃2,i Λ3

of the integral gain KI .  Λ̃T2,i −(1 − β)e−2ατmax Q 0  < 0
Remark 1. Note, from equation (13c), that the distur- Λ3 T
0 −γI
bance w(t) can be totally decoupled (i.e. its influence is
for i = 1 · · · L where
null) if and only if KP W = C 2 W + C 1 D̃. Notice that
T T T
    Λ1 = C 1 Ã C 1 + C 2 N C 2 + αI − GI C2 ,
D̃ KP 1
= KP W = W (14)
T
Λ̃2,i = P C 2 − GP C̃i C 1 ,
W I


and consequently, KP 1 = D̃W T (W W T )−1 . So, this condi- Λ3 = P C 2 − GP W + P C 1 D̃ ,
tion cannot be easily satisfied in practice because W is not C̃i = [0 · · · Ci · · · 0] .
necessary of complete rank. Besides, the total disturbance for a given decay rate α and for prescribed matrices H
decoupling case reduce inevitably the freedom degrees to and N . The observer gains are given by KP = P −1 GP
adjust the dynamic performances. and KI √= P −1 GI ; the L2 gain from w(t) to ν(t) is given
by γ = γ.
The robust PIO design problem can thus be formulated as Proof. The exponential convergence of the estimation
finding the matrices K̃P and K̃I such that the influence error (11) is investigated via a Lyapunov-Krasovskii func-
of w(t) on the estimation error (11) is attenuated and the tional proposed by Mondié and Kharitonov [2005]:
state estimation error remains globally bounded for any
blend between the submodels. To this end, the following Z0
objective signal ν(t) which only depends on the estimation V (t) = eTa (t)P ea (t) + eTa (t + θ)e2αθ Qea (t + θ)dθ
error ea (t) is introduced:
−τ (t)

ν(t) = Hea (t) , (15) (17)

where H is a matrix of appropriate dimension chosen where P and Q are symmetric, positive definite matrices.
by the designer. The weighting matrix H in (15) can be Mondié and Kharitonov [2005] propose this functional
selected for totally or partially taking into consideration in order to provide sufficient delay-independent condi-
the transfer from w(t) to ea (t). tions for the exponential stability of linear systems with
constant time-delay. We use the same functional but by
Finally, the expected performances of the PIO can be considering a time-varying delay. On the other hand, the
formulated by the following robust performances: Lyapunov-Krasovskii functional (17) can be rewritten as
follows:
lim ea (t) = 0 for w(t) = 0 , (16a)  T   
t→∞ ea (t) P 0 ea (t)
V (t) = (18)
kν(t)k22 ≤ γ 2 kw(t)k22 for w(t) 6= 0 and ν(0) = 0 , (16b) ea (∇) 0 0 ea (∇)
where γ is the L2 gain from w(t) to ν(t) to be minimized. Z0
Notice that condition (16a) ensures convergence towards + eTa (t + θ)e2αθ Qea (t + θ)dθ .
zero of the estimation error in the perturbation free case.
On the other hand, condition (16b) guarantees attenuation −τ (t)

level on the estimation error with respect to perturbations. On the other hand, the robust state estimation objectives
In the following, we will investigate the exponential con- (16) are guaranteed if there exists a Lyapunov-Krasovskii
vergence of the estimation error. Indeed, the exponential functional (17) such that [Boyd et al., 1994]:
convergence is a strong form of convergence that guar-
antees dynamic performances of the estimation error, in
V̇ (t) + 2αV (t) < γ 2 wT (t)w(t) − ν T (t)ν(t) (19)
particular a convergence velocity via a decay rate.
where α is the so-called decay rate for convergence velocity
and γ is the attenuation level from w(t) to ν(t) for robust
3.4 Computation of the gains K̃P and K̃I estimation. So, inequality (19) must be ensured in order to
provide a solution of the robust state estimation problem.
This section focuses on the computation of the PIO gains
K̃P and K̃I . For this purpose, the Lyapunov-Krasovskii Let us consider initially the time-derivative of the func-
functional method is employed in order to provide delay- tional (17) evaluated via Leibniz-Newton formula:
independent sufficient conditions, in LMIs terms [Boyd
et al., 1994], for ensuring the robust performances (16). V̇ (t) = ėTa (t)P ea (t) + eTa (t)P ėa (t) + eTa (t)Qea (t)
− (1 − τ̇ (t))e−2ατ (t) eTa (∇)Qea (∇) (20)
Theorem 1. Consider the decoupled multiple model (1) Z0
with time-delay (2) and assumption 1. There exists a − 2α eTa (t + θ)e2αθ Qea (t + θ)dθ
PIO (10) ensuring the objectives (16) if there exists
two symmetric, positive definite matrices P and Q, two −τ (t)
matrices GP and GI and a positive scalar γ solution of that can be upper bounded, using the a priori knowledge
the constrained optimisation problem: (2) of the time-delay as follows:
 V̇ (t) ≤ ėTa (t)P ea (t) + eTa (t)P ėa (t) + eTa (t)Qea (t) L
" #
X 2αP +H T H+Λ̃T 1 P +P Λ̃1 +Q P Λ̃2,i P Λ3
µi (∇) Λ̃T
2,i P −(1−β)e−2ατmax Q 0 < 0.
− (1 − β)e−2ατmax eTa (∇)Qea (∇) (21) ΛT 0 −γ 2 I
i=1 3 P
Z0 (28)
− 2α eTa (t + θ)e2αθ Qea (t + θ)dθ ,
where the constant matrices Λ̃2,i are given by:
−τ (t)
T
which becomes finally by considering the dynamics of the


Λ̃2,i = C 2 − KP C̃i C 1 . (29)
estimation error (12):
According to the properties (3) of the weighting functions,
the above matrix inequality (28) can be satisfied by con-
Z0
sidering the simultaneous solution of the set of inequalities
V̇ (t) ≤ −2α eT (t + θ)e2αθ Qe(t + θ)dθ expressed for each upper bound of µi (∇). On the other
−τ (t) hand, the linearisation of the matrix inequality (28) can be
+wT (t)ΛT3 P ea (t) + eTa (t)P Λ3 w(t) (22) achieved by the standard change of variables GP = P K̃P ,
h iT h T ih i GI = P K̃I and γ 2 = γ. Hence, sufficient conditions for
ea (t) Λ1 P +P Λ1 +Q P Λ2 (∇) ea (t)
+ ea (∇) ea (∇) . ensuring the robust performances (16) are obtained under
ΛT
2 (∇)P −(1−β)e−2ατmax Q
a LMI form which can be solved via classical LMI tools.
Hence, the left-hand side of the inequality (19) is ensured The proof of theorem 1 is completed. ✷
by the following inequality:
4. SIMULATION EXAMPLE
#T 
ea (t)
" 
2αP 0 P Λ3 Let us consider the state estimation of a decoupled multi-
V̇ + 2αV ≤ ea (∇) 0 0 0 (23) ple model with L = 2 submodels. The parameters of the
ΛT3 P 0 0
w(t) submodels are:
 T  " ea (t) # "
−2 0.5 0.6
#
Λ1 P +P Λ1 +Q P Λ2 (∇) 0
h i
−0.2 −0.6
+ ΛT
2 (∇)P −(1−β)e−2ατmax Q 0 ea (∇) . A1 = −0.3 −0.9 0 , A2 =
0.3 −1
,
0 0 0 w(t) −1.3 0.6 −0.8
 T  T
Now, by considering (23) then the inequality (19) can be B1 = 1 0.2 0.5 , B2 = −0.5 0.8 ,
guaranteed if the following inequality holds: 
D1 = 0.1 0.2 −0.3
T
,

D2 = −0.1 0.1
T
,
h i h i
#T  1 0.8 0.5 0.7 0.3
ea (t) C1 = , C2 = ,
"
ΛT

1 P +P Λ1 +Q P Λ2 (∇) 0 0.2 0.3 −0.5 0.2 −0.5
ea (∇) T
Λ2 (∇)P −(1−β)e −2ατmax
Q0  T
w(t) 0 0 0 W = 0.2 −0.3 .
  "
ea (t)
# The time-delay function appearing in the output of the
2αP +H T H 0 P Λ3
+ 0 0 0 ea (∇) < 0 . (24) systems is given by τ (t) = 0.5 + 0.45 sin(0.5t) with τmax =
ΛT3 P 0 −γ 2 I
w(t) 0.95. The upper bound of its derivative is β = 0.225. Here,
the decision variable ξ(t) is the input signal u(t) ∈ [0, 1].
Let us notice that the above inequality is a quadratic form The weighting functions are obtained from normalised
T
in [ ea (t) ea (∇) w(t) ] . Consequently, the inequality (19) for Gaussian functions:
ensuring robust objectives (16) is finally guaranteed if the
following inequality holds: XL

" # µi (ξ(t)) = ωi (ξ(t))/ ωj (ξ(t)) , (30)

T T
2αP +H H+Λ1 P +P Λ1 +Q P Λ2 (∇) P Λ3 j=1
ΛT
2 (∇)P −(1−β)e−2ατmax Q 0 <0 . 
2

ΛT3 P 0 −γ 2 I ωi (ξ(t)) = exp −(ξ(t) − ci ) /σ 2 , (31)
(25) with σ = 0.4, c1 = 0.3 and c2 = 0.7.
Note that a time-varying matrix Λ2 (∇), given by (13b),
A solution that satisfies the theorem 1, is given by:
appears in this inequality and its resolution cannot be
computed using standard LMI algorithms. 0.075 0.144 −0.220 −0.074 0.075 0.999 0.002 T
 
K̃P = −0.237 −0.444 0.678 0.225 −0.238 0.002 0.992 ,
However, the time-delayed varying matrix C̃(∇) in Λ2 (∇)  −0.431 0.139 −0.068 −0.054 0.046 2.269 −0.045 T
can be rewritten as a weighted sum of constant matrices K̃I = −0.141 −0.434 −0.114 −0.022 0.011 0.042 2.252 ,
as follows: with a decay rate α = 0.2 and an attenuation level γ = 0.5 2

L
and for H = I and N = −0.3.
X
C̃(∇) = µi (∇)C̃i , (26) In the simulation the disturbance acting on the output is a
i=1 normally distributed random signal. The state estimation
where C̃i is a constant block matrix given by: error is plotted in figure 1 where ei is the ith component
of the estimation error vector (11). The measured and the
estimated outputs are shown in figure 2. In both cases,
C̃i = [0 · · · Ci · · · 0] (27)
the error around the origin time is due to the differences
such that the term Ci is found on the ith block column of between initial conditions of the multiple model and the
C̃i . By considering (26) then the inequality (25) becomes observer.
 1
SIAM Studies in Applied Mathematics 15, Philadelphia,
0.8
e1 (t) 1994.
e2 (t) K.K. Busawon and P. Kabore. Disturbance attenuation
0.6 e3 (t)
0.4 e4 (t) using proportional integral observers. International
0.2 e5 (t) Journal of Control, 74(74):618–627, 2001.
0 D. Filev. Fuzzy modeling of complex systems. Interna-
−0.2 tional Journal of Approximate Reasoning, 5(3):281–290,
−0.4 1991.
−0.6 E. P. Gatzke and F. J. Doyle III. Multiple model approach
−0.8
0 5 10 15 20
for CSTR control. In 14th IFAC World Congress,
time (s) Beijing, China, 1999.
G. Gregorcic and G. Lightbody. Nonlinear system iden-
Figure 1. State estimation error tification: From multiple-model networks to gaussian
processes. Engineering Applications of Artificial Intelli-
1.5
y(t) gence, 21(7):1035–1055, 2008.
1
ŷ(t) C. Hua and X. Guan. Synchronization of chaotic systems
based on PI observer design. Physics Letters A, 334
0.5 (5-6):382–389, 2005.
J. Jung, K. Huh, and T. Shim. Dissipative proportional
0
integral observer for a class of uncertain nonlinear
systems. In American Control Conference, ACC, New
−0.5
York City, USA, 2007.
−1 C. Lin, Q.G. Wang, T.H. Lee, and Y. He. LMI Approach to
0 5 10 15 20
time (s) Analysis and Control of Takagi-Sugeno Fuzzy Systems
with Time Delay. Springer-Verlag Berlin Heidelberg,
Figure 2. Actual output and its estimated 2007.
S. Mondié and V.L. Kharitonov. Exponential estimates
5. CONCLUSION
for retarded time-delay systems: An LMI approach.
IEEE Transactions on Automatic Control, 50(2):268–
A proportional-integral observer (PIO), on the basis of a 273, 2005.
decoupled multiple model, is proposed in this contribution R. Murray-Smith and T.A. Johansen. Multiple model
in order to cope with the state estimation problem of a Approaches to Modelling and Control. Taylor & Francis,
nonlinear system in presence of delayed and perturbed London, 1997.
measurements. The decoupled multiple model is an in- R. Orjuela, B. Marx, D. Maquin, and J. Ragot. A de-
teresting structure for modelling nonlinear systems with coupled multiple model approach for state estimation of
variable structure because the dimensions of the employed nonlinear systems subject to delayed measurements. In
submodels can be different. On the other hand, the PIO 3rd IFAC Advanced Fuzzy and Neural Network Work-
offers more degree of freedom for robust state estimation shop, AFNC, Valenciennes, France, 2007.
with respect to the classic proportional observer previously R. Orjuela, B. Marx, D. Maquin, and J. Ragot. State
proposed. The robust stability problem of the estimation estimation for nonlinear systems using a decoupled
error is investigated using the Lyapunov-Krasovskii func- multiple mode. International Journal of Modelling
tional method and delay-independent sufficient conditions, Identification and Control, 4(1):59–67, 2008.
under a LMI form, are established. J.P. Richard. Time-delay systems: an overview of some
Further research will be to investigate the state estimation recent advances and open problems. Automatica, 39(10):
by considering distributed delays in the submodel dynamic 1667–1694, 2003.
equations. O. Sename. New trends in design of observers for time-
delay systems. Kybernetika, 37(4):427–458, 2001.
REFERENCES C.A. Smith and A.B. Corripio. Principles and practice of
automatic process control. John Wiley & Sons, 1985.
R. Babuska. Fuzzy Modeling for Control. Kluwer Aca- J. F. Sturm. Using SeDuMi 1.02, A Matlab toolbox
demic Publishers, London, 1998. for optimization over symmetric cones. Optimization
S. Beale and B. Shafai. Robust control system design with Methods and Software, 11(1):625–653, 1999.
a proportional integral observer. International Journal K. Tanaka and H. Wang. Fuzzy Control Systems design
of Control, 50(1):97–111, 1989. and analysis. John Wiley & Sons, New York, 2001.
P. Bliman. LMI characterization of the strong delay- F.J. Uppal, R.J. Patton, and M. Witczak. A neuro-fuzzy
independent stability of linear delay systems via multiple-model observer approach to robust fault diag-
quadratic Lyapunov-Krasovskii functionals. Systems nosis based on the DAMADICS benchmark problem.
and Control Letters, 43(4):263–274, 2001. Control Engineering Practice, 14(6):699–717, 2006.
A. Boukhris, G. Mourot, and J. Ragot. Non-linear A. Weinmann. Uncertain Models and Robust Control.
dynamic system identification: a multiple-model ap- Springer-Verlag Wien New York, Vienne, 1991.
proach. International Journal of Control, 72(7/8):591– M. Wu, Y. He, J.H. She, and G.P. Liu. Delay-dependent
604, 1999. criteria for robust stability of time-varying delay sys-
S. Boyd, L. El Ghaoui, E. Feron, and V. Balakrishnan. Lin- tems. Automatica, 40(8):1435–1439, 2004.
ear Matrix Inequalities in System and Control Theory.