Professional Documents
Culture Documents
Non-Fragile Observer-Based Passive Control For Discrete-Time Systems With Repeated Scalar Non-Linearities
Non-Fragile Observer-Based Passive Control For Discrete-Time Systems With Repeated Scalar Non-Linearities
G. Arthi
Nonlinear Dynamics Group, Department of Electrical Engineering, Yeungnam University, 280
In this paper, the non-fragile observer-based passive control problem is discussed for a class of sys-
tems with repeated scalar non-linearities and time-varying delays. The non-linear system is defined by
a discrete-time state equation containing a repeated scalar non-linearity. The system under considera-
tion is modelled by assuming the random imperfect communication links existing between the controller
and observer. The random fluctuations are defined by utilizing the Bernoulli distributed white sequences.
The non-fragile observer-based feedback controller gains are designed to guarantee that the considered
closed-loop control system with repeated scalar non-linearities and time-varying delays is passive. Suf-
ficient conditions are derived for the existence of controller and observer gains by using the Lyapunov
stability theory, passivity theory and linear matrix inequalities. As a final point, a numerical example
by using a marketing-production system is presented to demonstrate the effectiveness of the proposed
theoretical results.
Keywords: discrete-time systems; observer-based passive control; repeated scalar non-linearities; time-
varying delays.
1. Introduction
Passivity is an essential property of many physical systems which may be defined in terms of energy
dissipation and transformation. Passivity theory is intimately related to circuit analysis method, which
plays a crucial role in both electrical networks and control systems. During the last decades, the pas-
sivity theory has been paid much attention from the control community since it has connection with
physical properties and stability of the investigated systems (Fradkov & Hill, 1998; Zhu et al., 2010;
Mathiyalagan et al., 2014; Rakkiyappan et al., 2014). Passive systems have many necessary properties
like the zero dynamics and free dynamics of passive systems are Lyapunov stable. A passive system can
attain stability by means of output feedback and the negative and parallel feedback interconnections of
passive systems (Khalil, 2002; Bao & Lee, 2007). Besides, the passivity-based control is a methodol-
ogy that involves in controlling a system with the intention at making the closed-loop system passive.
Furthermore, the passivity theory provides an energy-based analysis approach (Van der Schaft, 1996),
which is easier to execute when dealing with electrical circuits and mechanical systems.
c The authors 2015. Published by Oxford University Press on behalf of the Institute of Mathematics and its Applications. All rights reserved.
2 of 18 G. ARTHI ET AL.
Discrete-time systems have been an important research topic for its theoretical and practical signif-
icance in the area of stability and control (Elaydi, 2005). In the past few decades, the control problems
for discrete-time systems have attracted considerable research interests and a great number of important
results has been reported (see Park et al., 2009; Huang & Feng, 2010; Hu et al., 2012; Li & Wang,
2013 and the references therein). However, this control is frequently based on the assumption that the
complete state is available, which may not hold in many systems. Therefore, it is more realistic to use
an observer-based feedback controller, which is a dynamic feedback controller that estimates the system
states online. Hence, it is essential to design observers that produce an estimate of the system state. From
with a prescribed weighted 2 − ∞ performance. Since the introduction of the notion of passivity the-
ory, many researchers have discussed the passive control problem. But, passive control problem with
repeated scalar non-linearity has not been studied yet in the past. This fact is the main motivation of
this work. To the best of authors’ knowledge, no results are available for the problem of non-fragile
observer-based passive control for discrete-time systems with repeated scalar non-linearities and time-
varying delay in the literature. The main contributions of this paper are highlighted as follows:
• Non-fragile observer-based passive control is designed for the repeated scalar non-linear systems.
2. Problem description
Notations: The notations used in this paper are quite standard. The superscript ‘’ and ‘(−1)’ stand for
matrix transposition and matrix inverse, respectively. Rn is the n-dimensional Euclidean space, Rn×n
denotes the set of all real matrices of dimension n × n; the notation P > 0 (respectively, P < 0) means
that P is positive definite (respectively, negative definite). I and 0 represent the identity matrix and zero
matrix with compatible dimension, respectively. In symmetric block matrices or long matrix expres-
sions, we use an asterisk (∗) to represent a term that is induced by symmetry. In addition, diag{·} stands
for a block-diagonal matrix, sym(A) is defined as A + A and · refers to the Euclidean vector norm.
The notation E[·] stands for the mathematical expectation operator with respect to the given probability
measure.
Consider the following discrete-time system with repeated scalar non-linearities:
⎫
x(k + 1) = Af (x(k)) + A1 g(x(k − τ (k))) + Bu(k) + Gw(k),⎪
⎪
⎪
⎪
y(k) = Cf (x(k)) + C1 g(x(k − τ (k))) + Dw(k),⎬
(1)
z(k) = Ef (x(k)) + Fw(k),⎪ ⎪
⎪
⎪
⎭
x(k) = ϕ(k), k ∈ [−τM , 0],
where x(k) ∈ Rn denotes the system state vector; u(k) ∈ Rm is the control input; w(k) ∈ Rl is the distur-
bance input which belongs to l2 [0, ∞); z(k) ∈ Rp is the controlled output; y(k) ∈ Rq is the measured out-
put vector; τ (k) denotes the time-varying delay with lower and upper bounds τm τ (k) τM , k ∈ N+ ,
where τm , τM are known positive integers; ϕ(k) is the initial state of the system; A, A1 , B, G, C, C1 , D, E
and F are known real constant matrices with compatible dimensions. The function f (·) is non-linear and
The function g(·) is also non-linear and satisfies the above assumption with f (·) replaced by the non-
linear function g(·).
Remark 2.1 The plant model structure (1) is known as a system with repeated scalar non-linearity (Chu
& Glover, 1999a,b; Chu, 2001). It should be noted that f (·) is odd function (by substituting b = −a)
where x̂(k) ∈ Rn denotes the state estimate vector of the system (1); ŷ(k) ∈ Rq is the output; fˆ (x̂(k)) is
the input of the controller; û(k) ∈ Rm is the output of the controller; K ∈ Rm×n and H ∈ Rn×q are the
gain parameters of the controller and observer to be determined.
In (3), ΔH(k) and ΔK(k) are the perturbed matrices which denotes the fluctuation in observer
and controller gain matrices, respectively. The perturbations are assumed to be satisfied the following:
ΔH(k) = MH FH (k)NH , ΔK(k) = MK FK (k)NK , where MH , NH , MK and NK are known constant matri-
ces of appropriate dimensions, FH (k) and FK (k) are unknown time-varying matrices with Lebesgue
measurable elements bounded by FH (k)FH (k) I and FK (k)FK (k) I.
Based on the existence of packet losses between the controller and observer, we design the packet
losses occurrence by means of a stochastic approach as follows:
where α(k) and β(k) are two mutually independent and Bernoulli distributed white sequences. Here,
α(k) designs the unreliable nature of the connection from the sensor to the controller and β(k) designs
that from the controller to the actuator.
Suppose the communication connection fails (i.e. data are lost), then the value of α(k) is equal to
zero and α(k) = 1 represents the successful transmission. Similar effect occurs for β(k). The sequences
α(k) and β(k) satisfy the following assumption:
u(k) = [K +
K(k)]α(k)β(k)f (x̂(k)).
NON-FRAGILE OBSERVER-BASED PASSIVE CONTROL FOR DISCRETE-TIME SYSTEMS 5 of 18
Let e(k) = x(k) − x̂(k) be the estimation error vector, then we have the following closed-loop system
⎫
x(k + 1) = Af (x(k)) + A1 g(x(k − τ (k))) + B[K +
K(k)]α(k)β(k)f (x̂(k)) + Gw(k),⎪
⎪
⎪
⎪
⎬
e(k + 1) = A[f (x(k)) − f (x̂(k))] + A1 [g(x(k − τ (k))) − g(x̂(k − τ (k)))]
(4)
+ Gw(k) + [H +
H(k)]{C[f (x(k)) − f (x̂(k))] ⎪
⎪
⎪
⎪
⎭
+ C1 [g(x(k − τ (k))) − g(x̂(k − τ (k)))] + Dw(k)}.
and
u(k) = [K +
K(k)]ρ(k)f (x̂(k)).
From the above equality and (4), we can now obtained the closed-loop system as follows:
⎫
x(k + 1) = Af (x(k)) + A1 g(x(k − τ (k))) + B[K +
K(k)]ρ̄f (x(k)) ⎪
⎪
⎪
⎪
+ (ρ(k) − ρ̄)B[K +
K(k)]f (x(k)) − B[K +
K(k)]ρ̄ f (e(k))⎪
¯ ⎪
⎪
⎪
⎬
− (ρ(k) − ρ̄)B[K +
K(k)]f¯ (e(k)) + Gw(k), (5)
⎪
⎪
⎪
⎪
e(k + 1) = Af¯ (e(k)) + A1 ḡ(e(k − τ (k))) + Gw(k) + [H +
H(k)] ⎪
⎪
⎪
⎪
⎭
× {C f¯ (e(k)) + C1 ḡ(e(k − τ (k))) + Dw(k)}
where
ξ(k) = [x (k) e (k)] , f˜ (ξ(k)) = [f (x(k)) f¯ (e(k))] , f¯ (e(k)) = f (x(k)) − f (x̂(k)),
g̃(ξ(k − τ (k))) = [g (x(k − τ (k))) ḡ (e(k − τ (k)))] ,
ḡ(e(k − τ (k))) = g(x(k − τ (k))) − g(x̂(k − τ (k))),
A + B[K +
K(k)]ρ̄ −B[K +
K(k)]ρ̄
φ1 = ,
0 A + [H +
H(k)]C
B[K +
K(k)](ρ(k) − ρ̄) −B[K +
K(k)](ρ(k) − ρ̄)
φ2 = ,
0 0
A1 0 G
φ3 = , φ4 = .
0 A1 + [H +
H(k)]C1 G + [H +
H(k)]D
Before finishing this section, we recall the following definition and lemmas which will be used in the
derivation of our results.
6 of 18 G. ARTHI ET AL.
The main objective of this paper is to design the passivity-based control design for system (6). It
should be noted that the problem of passivity-based controller can guarantee the system is strictly-output
passive with the energy function J(z(k), w(k)) = 2z (k)w(k) + γ w (k)w(k). In order to deal this, we
present the following definition.
Definition 2.2 The discrete-time system with repeated scalar non-linearities (6) is called passive if
there exists a scalar γ > 0 and for all n ∈ N such that
Lemma 2.3 (Boyd et al., 1994) Given constant matrices Ω1 , Ω2 and Ω3 with
appropriate
dimensions,
−1 Ω1 Ω3
where Ω1 = Ω1 and Ω2 = Ω2 , then Ω1 + Ω3 Ω2 Ω3 < 0, if and only if, ∗ −Ω < 0.
2
Lemma 2.4 (Xie, 1996) Assume that Ω, M and N are real matrices with appropriate dimensions and
F(k) is a matrix function satisfying F (k)F(k) I. Then, Ω + MF(k)N + [MF(k)N] < 0 holds if and
only if there exist a scalar
> 0 satisfying Ω +
−1 MM +
N N < 0.
Lemma 2.5 (Li et al., 2012) For any matrix Π ∈ Rq×n with q < n and full row rank (rank (Π ) = q), there
exists an SVD of Π as follows: Π = U[S 0]V , where S ∈ Rq×q is a diagonal matrix with non-negative
diagonal elements in decreasing order, U ∈ Rq×q , V ∈ Rn×n are the unitary matrices.
Lemma 2.6 (Li et al., 2012) Given matrix Π ∈ Rq×n with q < n and full row rank (rank (Π ) = q),
assume that Y ∈ Rn×n is a symmetric matrix, Y1 then
there exists a matrix Ȳ ∈ Rq×q satisfying Π Y = Ȳ Π
if and only if Y can be described as: Y = V 0 Y2 V , where Y1 ∈ Rq×q , Y2 ∈ R(n−q)×(n−q) and V ∈ Rn×n
0
3. Main results
In this section, we first investigate the non-fragile observer-based design to guarantee that the resulting
closed-loop discrete-time system with repeated scalar non-linearities (6) is passive.
Theorem 3.1 Let the positive scalars τm , τM , ρ̄ and gain matrices K, H be given. The discrete-time
system with repeated scalar non-linearities (6) is passive, if there exist positive definite matrices P1 , P2 ,
Q1 , Q2 , R1 , R2 , S1 , S2 and positive constants
1 ,
2 , ρ1 , ρ2 , γ such that the following LMI holds:
⎡ ⎤
Φ1 M̄K
1 N̄K M̄H
2 N̄H
⎢ ⎥
⎢∗ −
1 I 0 0 0 ⎥
⎢ ⎥
Θ̂ = ⎢
⎢∗ ∗ −
1 I 0 0 ⎥ < 0,
⎥ (7)
⎢ ⎥
⎣∗ ∗ ∗ −
2 I 0 ⎦
∗ ∗ ∗ ∗ −
2 I
NON-FRAGILE OBSERVER-BASED PASSIVE CONTROL FOR DISCRETE-TIME SYSTEMS 7 of 18
where
⎡ ⎤
Φ0 0 0 0 0 0 0 0 0 0
⎢∗ −ρ1 I −E η1 η2 0 ⎥
⎢ 0 0 0 0 ⎥
⎢ ⎥
⎢∗ ∗ −ρ1 I 0 0 0 η3 η4 η5 0 ⎥
⎢ ⎥
⎢∗ ∗ ∗ −ρ2 I
0 ⎥
⎢ 0 0 A1 P1 0 0 ⎥
⎢ ⎥
⎢∗ ∗ ∗ ∗ −ρ2 I 0 0 η6 0 0 ⎥
Φ1 = ⎢ ⎥,
Proof. In order to establish our results, we consider the following Lyapunov–Krasovskii functional for
the system (6):
3
V (ξ(k), k) = Vs (ξ(k), k), (8)
s=1
where
V1 (ξ(k), k) = ξ (k)Pξ(k),
k−1 −τm
k−1
V2 (ξ(k), k) = ξ (s)Qξ(s) + ξ (s)Qξ(s),
s=k−τ (k) i=−τM +1 s=k+i
k−1
k−1
V3 (ξ(k), k) = ξ (s)Rξ(s) + ξ (s)Sξ(s),
s=k−τm s=k−τM
E[
V1 (k)] = E[ξ (k + 1)Pξ(k + 1) − ξ (k)Pξ(k)]
= E[[φ1 f˜ (ξ(k)) + φ2 (ρ(k) − ρ̄)f˜ (ξ(k)) + φ3 g̃(ξ(k − τ (k))) + φ4 w(k)] P
[φ1 f˜ (ξ(k)) + φ2 (ρ(k) − ρ̄)f˜ (ξ(k)) + φ3 g̃(ξ(k − τ (k))) + φ4 w(k)] − ξ (k)Pξ(k)]
E[
V (k)] E[[φ1 f˜ (ξ(k)) + φ3 g̃(ξ(k − τ (k)))] P[φ1 f˜ (ξ(k)) + φ3 g̃(ξ(k − τ (k)))]
− ξ (k)Pξ(k) + ρ̄(1 − ρ̄)[φ2 f˜ (ξ(k))] P[φ2 f˜ (ξ(k))] + ξ (k)Qξ(k)
+ (τM − τm )ξ (k)Qξ(k) − ξ (k − τ (k))Qξ(k − τ (k)) + ξ (k)Rξ(k)
+ ξ (k)Sξ(k) − ξ (k − τm )Rξ(k − τm ) − ξ (k − τM )Sξ(k − τM )]. (12)
E[
V (k)] E[η (k)[Π + Π1 PΠ1 + Π2 PΠ2 ]η(k)], (13)
where
⎡ ⎤
Π (1, 1) 0 0 0 0 0
⎢ ∗ −R 0 ⎥
⎢ 0 0 0 ⎥
⎢ ⎥
⎢ ∗ ∗ −Q + ρ2 I 0 0 0 ⎥
Π =⎢
⎢ ∗
⎥,
⎢ ∗ ∗ −S 0 0 ⎥
⎥
⎢ ⎥
⎣ ∗ ∗ ∗ ∗ −ρ1 I 0 ⎦
∗ ∗ ∗ ∗ ∗ −ρ2 I
Now, to discuss the passive-based control problem for system (6), we consider the performance index
J given by
Then
E[
V (k)] E[[φ1 f˜ (ξ(k)) + φ3 g̃(ξ(k − τ (k))) + φ4 w(k)] P[φ1 f˜ (ξ(k)) + φ3 g̃(ξ(k − τ (k)))
+ φ4 w(k)] − ξ (k)Pξ(k) + ρ̄(1 − ρ̄)[φ2 f˜ (ξ(k))] P[φ2 f˜ (ξ(k))]
+ (τM − τm + 1)ξ (k)Qξ(k) − ξ (k − τ (k))Qξ(k − τ (k)) + ξ (k)[R + S]ξ(k)
− ξ (k − τm )Rξ(k − τm ) − ξ (k − τM )Sξ(k − τM )],
which implies
E[
V (k)] E[ζ (k)[Φ + Φ1 PΦ1 + Φ2 PΦ2 ]ζ (k)], (15)
where
It is noted that the matrix inequality (15) is non-linear and unable to directly solve them using the
existing numerical toolboxes. Now, in order to make this matrix inequalities solvable, by applying the
10 of 18 G. ARTHI ET AL.
It is assumed that the observer-based feedback gains in system (6) have the fluctuations. By noticing
this point and using the assumption on uncertain gain matrices, we obtain
where Φ1 , M̄K , N̄K , M̄H and N̄H are given in the statement of Theorem 3.1.
Then by utilizing the Lemma 2.4, for positive constants
1 and
2 , we obtain
Φ̂ = Φ1 + 1−1 M̄K M̄K + 1 N̄K N̄K + 2−1 M̄H M̄H + 2 N̄H N̄H . (17)
In view of Lemma 2.3, it is easy to see that (7) follows from (17). If LMI (7) holds, then
E[
V (k) − 2[Ēf˜ (ξ(k)) + Fw(k)] w(k) − γ w (k)w(k)] 0,
NON-FRAGILE OBSERVER-BASED PASSIVE CONTROL FOR DISCRETE-TIME SYSTEMS 11 of 18
which implies
n
n
2E [z (k)w(k)] −γ w (k)w(k).
k=0 k=0
Hence, the considered discrete-time system (6) with repeated scalar non-linearities is passive by means
of Definition 2.2. Hence, the proof is completed.
Now, we are in a position to determine the controller gains based on the passive control of discrete-
where
⎡ ⎤
Φ0 0 0 0 0 0 0 0 0 0
⎢∗ −ρ1 I −E η̂1 η̂2 ⎥
⎢ 0 0 0 0 0 ⎥
⎢ ⎥
⎢∗ ∗ −ρ1 I 0 0 0 η̂3 η̂4 η̂5 0 ⎥
⎢ ⎥
⎢∗ ∗ ∗ −ρ2 I A ⎥
⎢ 0 0 1 P1 0 0 0 ⎥
⎢ ⎥
⎢∗ ∗ ∗ ∗ −ρ2 I 0 0 η̂6 0 0 ⎥
Φ̃1 = ⎢
⎢∗
⎥,
⎥
⎢ ∗ ∗ ∗ ∗ −2F − γ I G P2 η̂7 0 0 ⎥
⎢ ⎥
⎢∗ ∗ ∗ ∗ ∗ ∗ −P1 0 0 0 ⎥
⎢ ⎥
⎢∗ ∗ ∗ ∗ ∗ ∗ ∗ −P2 ⎥
⎢ 0 0 ⎥
⎢ ⎥
⎣∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ −P1 0 ⎦
∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ −P2
η̂1 = A P1 + ρ̄M B ,
η̂2 = ρ̄(1 − ρ̄)M B , η̂3 = −ρ̄M B , η̂4 = A P2 + C N ,
η̂5 = − ρ̄(1 − ρ̄)M B , η̂6 = A
1 P2 + C1 N , η̂7 = G P2 + D N ,
N̄K = [0 0 0 0 0 0 0 0 NK NK 0 0 0 0 0 0 0],
M̄K = [0 0 0 0 0 0 0 0 0 0 0 0 0 ρ̄MK B P1 0 ρ̄(1 − ρ̄)MK B P1 0] ,
N̄H = [0 0 0 0 0 0 0 0 0 NH C 0 NH C1 NH D 0 0 0 0],
P11 0
M̄H = [0 0 0 0 0 0 0 0 0 0 0 0 0 0 MH P2 0 0] , P1 = V V .
0 P22
Proof. As stated by Lemmas 2.5 and 2.6, for the transpose input matrix B ∈ Rq×n , there exists a matrix
P̄1 ∈ Rq×q satisfying B P1 = P̄1 B if and only if P1 ∈ Rn×n can be expressed as
P 0
P1 = V 11 V ,
0 P22
where P11 ∈ Rq×q , P22 ∈ R(n−q)×(n−q) and V ∈ Rn×n is the unitary matrix of SVD of B . Therefore,
P̄1 = USP11 S −1 U −1 , P̄1−1 = USP11
−1 −1 −1
S U and P1 B = BP̄1 .
Remark 3.3 It should be mentioned that the repeated scalar non-linearities on x(k) and x(k − τ (k)) in
(1) are not inevitably going to be the similar, which improves the flexibility of this model (for instance,
there might be non-linearities on x(k) but no non-linearities on x(k − τ (k))). In case, if we assume
that the non-linear functions f = g, i.e. the identical non-linearities appear on x(k) and x(k − τ (k)),
then some constricted stability conditions can be found, which deserves various exertion in imminent
research.
Remark 3.4 It should be noted that, if A1 = 0 and C1 = 0, then (6) turns to be a delay free system. The
proposed passivity control design can be easily deduced for the delay free discrete-time system with
repeated scalar non-linearity, which is presented through the following corollary.
Corollary 3.5 For given positive scalar ρ̄, the discrete-time system (6) with A1 = 0 and C1 = 0 is
passive, if there exist positive definite matrices P11 ∈ Rq×q , P22 ∈ Rn−q×n−q , P2 and positive constants
ˆ1 ,
ˆ2 , ρ1 , γ such that the following LMI holds:
⎡ ⎤
Φ̂1 M̂K
ˆ1 N̂K M̂H
ˆ2 N̂H
⎢ ⎥
⎢∗ −ˆ
1 I 0 0 0 ⎥
⎢ ⎥
⎢∗ ∗ −ˆ
1 I 0 0 ⎥ < 0, (19)
⎢ ⎥
⎢ ⎥
⎣∗ ∗ ∗ −ˆ
2 I 0 ⎦
∗ ∗ ∗ ∗ −ˆ
2 I
where
⎡ ⎤
−P1 + ρ1 I 0 0 0 0 0 0 0 0
⎢ ⎥
⎢ ∗ −P2 + ρ1 I 0 0 0 0 0 0 0 ⎥
⎢ ⎥
⎢ ∗ ∗ −ρ1 I 0 −E η̂1 0 η̂2 0 ⎥
⎢ ⎥
⎢ ⎥
⎢ ∗ ∗ ∗ −ρ1 I 0 η̂3 η̂4 η̂5 0 ⎥
⎢ ⎥
⎢
Φ̂1 = ⎢ ∗ ∗ ∗ ∗ −2F − γ I G P2 η̂7 0 0 ⎥,
⎥
⎢ ⎥
⎢ ∗ ∗ ∗ ∗ ∗ −P1 0 0 0 ⎥
⎢ ⎥
⎢ ∗ ∗ ∗ ∗ ∗ ∗ −P2 0 0 ⎥
⎢ ⎥
⎢ ⎥
⎣ ∗ ∗ ∗ ∗ ∗ ∗ ∗ −P1 0 ⎦
∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗ −P2
NON-FRAGILE OBSERVER-BASED PASSIVE CONTROL FOR DISCRETE-TIME SYSTEMS 13 of 18
N̂K = [0 0 NK NK 0 0 0 0 0],
M̂K = [0 0 0 0 0 ρ̄MK B P1 0 ρ̄(1 − ρ̄)MK B P1 0] ,
N̂H = [0 0 0 NH C NH D 0 0 0 0], M̂H = [0 0 0 0 0 0 MH P2 0 0] ,
4. Numerical example
where j = 1, 2. Let
⎡ ⎤ ⎡ ⎤
p1 (k + 1) s1 (k)
⎢p (k + 1)⎥ ⎢ s (k) ⎥
⎢ 2 ⎥ ⎢ 2 ⎥
x(k) := ⎢ ⎥, u(k) := ⎢ ⎥. (20)
⎣ i1 (k) ⎦ ⎣a1 (k)⎦
i2 (k) a2 (k)
Let the effect of the advertisement to the sales in the marketing procedure and the relationship between
amount of sales and production in the production process (assuming one-period gestation lag) be written
by a model of the form
where f (x(k)) and g(x(k − τ (k))) are saturation non-linearity function. It is readily seen that the above
model for combined marketing and production control problem fits well into the format of (1). In the
sequel, we consider a particular example (20) where
⎡ ⎤ ⎡ ⎤
0.3 0 0 0 −0.2 0 0 0
⎢ 0 0.5 0 0⎥ ⎢ −0.1 0 ⎥
A=⎢ ⎥ , A1 = ⎢ 0 0 ⎥,
⎣−0.7 0 0.3 0 ⎦ ⎣−0.05 0 0.1 0 ⎦
0 −0.5 0 0.3 0 −0.01 0 −0.05
0.5 0 −0.1 0.3
B = [1 0 − 0.7 0] , G = [0.2 0 0.5 1] , C = ,
0 0.2 0 0
0 0.2 0 −0.5
C1 = , D = [0.3 0] , E = [0 0.5 0.1 0], F = 0.1,
0.1 0 −0.3 0
14 of 18 G. ARTHI ET AL.
(a) (b)
0.1 0.8
x (k)
0.7 2
0 Estimation of x (k)
2
0.6
−0.1
x1(k) & its estimation
−0.2 0.4
0.3
0.2
−0.4
x (k)
0.1
1
−0.5 Estimation of x1(k) 0
−0.6 −0.1
0 20 40 60 80 100 0 20 40 60 80 100
Time k Time k
(c) (d)
1.2 0.4
x3(k)
1 Estimation of x3(k) 0.2
0.8 0
x3(k) & its estimation
0.6 −0.2
0.4 −0.4
0.2 −0.6
x (k)
4
0 −0.8 Estimation of x4(k)
−0.2 −1
0 20 40 60 80 100 0 20 40 60 80 100
Time k Time k
Fig. 1. State responses of the closed-loop system. (a) Responses of state x1 (k) and its estimation. (b) Responses of state x2 (k) and
its estimation. (c) Responses of state x3 (k) and its estimation. (d) Responses of state x4 (k) and its estimation.
By utilizing the Matlab LMI Toolbox and according to Theorem 3.2, we can obtain the following feasi-
ble solution for the considered model to be passive
⎡ ⎤
−0.0587 0.1910
⎢ 0.0589 −0.1105⎥
K = [−0.0696 − 0.0167 0.0540 0.0005], ⎢
H =⎣ ⎥
0.1334 0.3376 ⎦
−0.1164 0.1124
2
2
1
1
e (k)
e2(k)
0 0
1
−1 −1
0 50 100 0 50 100
Time k Time k
2
2
1 1
e (k)
e (k)
0
3
4
0
−1
−1
−2
−2
0 50 100 0 50 100
Time k Time k
0.6
0.5
0.4
0.3
z(k)
0.2
0.1
−0.1
0 20 40 60 80 100
Time k
following: ⎧
⎪
⎨x(k), |x(k)| 1,
f (x(k) = 1, x(k) > 1, and g(x(k − τ (k))) = x(k) sin(x(k)).
⎪
⎩
−1, x(k) < −1.
The state response of the closed-loop system is presented in Figs 1 and 2 which show the error
responses for 10 random initial conditions. The output is depicted in Fig. 3. Therefore, all the simulation
results, which confirm our theoretical analysis for the problem of non-fragile observer-based passive
5. Conclusion
In this paper, a non-fragile observer-based passive control of discrete-time systems with repeated scalar
non-linearities and time-varying delays has been studied. To design the state observer and non-fragile
controller, which ensure the passivity of the resulting closed-loop system, has been proposed based
on the LMI approach. Finally, a practical simulation example is presented to show the validity and
advantages of the proposed results.
Acknowledgements
The authors would like to thank the chief editor, associate editor and anonymous reviewers for their
constructive comments and suggestion for improving the quality of the paper.
Funding
This work was supported by Basic Science Research Program through the National Research Founda-
tion of Korea (NRF) funded by the Ministry of Education (2013R1A1A2A10005201).
References
Bao, J. & Lee, P. L. (2007) Process Control: The Passive Systems Approach. New York: Springer.
Boyd, S., Ghoui, L. E., Feron, E. & Balakrishnan, V. (1994) Linear Matrix Inequalities in System and Control
Theory. Philadelphia, PA: SIAM.
Chen, X. & Hao, F. (2013) Observer-based event-triggered control for certain and uncertain linear systems. IMA J.
Math. Control Inform., 30, 527–542.
Chen, J. D., Yang, C. D., Lien, C. H. & Horng, J. H. (2008) New delay-dependent non-fragile H∞ observer-
based control for continuous time-delay systems. Inform. Sci., 178, 4699–4706.
Chu, Y. C. (2001) Further results for systems with repeated scalar nonlinearities. IEEE Trans. Automat. Control,
46, 2031–2035.
Chu, Y. C. & Glover, K. (1999a) Bounds of the induced norm and model reduction errors for systems with
repeated scalar nonlinearities. IEEE Trans. Automat. Control, 44, 471–483.
Chu, Y. C. & Glover, K. (1999b) Stabilization and performance synthesis for systems with repeated scalar non-
linearities. IEEE Trans. Automat. Control, 44, 484–496.
Dong, H., Wang, Z. & Gao, H. (2009a) H∞ filtering for systems with repeated scalar nonlinearities under unreli-
able communication links. Signal Process., 89, 1567–1575.
Dong, H., Wang, Z. & Gao, H. (2009b) H∞ fuzzy control for systems with repeated scalar nonlinearities and
random packet losses. IEEE Trans. Fuzzy Syst., 17, 440–450.
Dong, H., Wang, Z. & Gao, H. (2010) Observer-based H∞ control for systems with repeated scalar nonlinearities
and multiple packet losses. Internat. J. Robust Nonlinear Control, 20, 1363–1378.
NON-FRAGILE OBSERVER-BASED PASSIVE CONTROL FOR DISCRETE-TIME SYSTEMS 17 of 18
Dong, H., Wang, Z., Ho, D. W. C. & Gao, H. (2011) Robust H∞ filtering for Markovian jump systems with
randomly occurring nonlinearities and sensor saturation: the finite-horizon case. IEEE Trans. Signal Process.,
59, 3048–3057.
Elaydi, S. (2005) An Introduction to Difference Equations. New York: Springer.
Fradkov, A. L. & Hill, D. J. (1998) Exponential feedback passivity and stabilizability of nonlinear systems.
Automatica, 34, 697–703.
Gao, H., Lam, J. & Wang, C. (2005) Induced l2 and generalized H2 filtering for systems with repeated scalar
nonlinearities. IEEE Trans. Signal Process., 53, 4215–4226.
Hu, J., Wang, Z., Gao, H. & Stergioulas, L. K. (2012) Robust H∞ sliding mode control for discrete time-delay
Yang, H., Xia, Y. & Shi, P. (2010) Observer-based sliding mode control for a class of discrete systems via delta
operator approach. J. Franklin Inst., 347, 1199–1213.
Zhang, Y., Shi, P., Nguang, S. K. & Karimi, H. R. (2014) Observer-based finite-time fuzzy H∞ control for
discrete-time systems with stochastic jumps and time-delays. Signal Process., 97, 252–261.
Zheng, Z., Ma, Y., Li, F. & Wu, L. (2014) Filter design for Markovian jump repeated scalar nonlinear systems.
J. Franklin Inst., 351, 4287–4301.
Zhu, S., Shen, Y. & Chen, C. (2010) Exponential passivity of neural networks with time-varying delay and
uncertainty. Phys. Lett. A, 375, 136–142.