2434 H. He et al.
itive systems, switched positive systems (SPSs) have
also attracted more and more attention due to the wide
range of applications, such as communication systems
[17], biological systems [18], and formation flying [19].
With the constraint of positive states, many existing
results cannot be applied to SPSs. Therefore, the co-
positive-type Lyapunov function approach is devel-
oped to analyze SPSs, including stability analysis [20],
control synthesis [21], fault detection [22], positive
observer [23], filtering [24], etc.
As most of the practical systems are not completely
linear, unmodeled nonlinearities widely exist in the
actual system. There are many valid methods to pro-
cess nonlinear systems, such as sliding mode control
(SMC) [25–30], fuzzy control [31,32], adaptive control
[33–35], backstepping control [36], and so on. Due to
the fast response and strong robustness, SMC has been
widely applied to deal with the nonlinear dynamics.
Nowadays, various sliding mode surfaces and reach-
ing laws have been developed. At the s ame time, many
system states are expensive to measure or even are not
available, especially for complex nonlinear hybrid sys-
tems. Therefore, state observer is often applied to esti-
mate states for the subsequent control process [37,38].
Many results on observer-based SMC [39–41]have
been obtained for hybrid systems.
However, so far the stabilization problem of switched
positive nonlinear systems has not yet been fully inves-
tigated and the observer-based SMC for positive sys-
tems has not been studied yet, which remain to be
important and challenging. This motivates our paper.
The differences between the existing relative works
and present work are obvious. First, in references [39–
41], the controlled plants are not switched positive sys-
tems. Second, observer-based SMC law is not applied
in references [14,42,43] because all the system states
are considered to be available. At the same time, the
asynchronous phenomenon and MDADT approach are
considered in this paper. The main contributions of this
paper are listed in three aspects: (i) The sufficient con-
ditions of exponential stability and positivity of the
closed-loop system are proposed; (ii) observer-based
SMC law and MDADT are designed to guarantee the
exponential stability and positivity of switched systems
with L
1
-gain performance; and (iii) an iterative algo-
rithm is proposed to solve the observer gain and con-
troller gain.
The remainder of this paper is organized as follows.
The problem statement and preliminaries are formu-
lated in Sect. 2, including system model, lemmas, and
definitions. The main results are presented in Sect. 3.
Section 4 gives an application example, followed by
conclusion in Sect. 5.
Notation Throughout this paper, symbol R
n
rep-
resents the n-dimensional Euclidean space, the set of
n × m real matrices is written as R
n×m
, and S =
{1, 2,...,N} means a set of positive numbers. The
interval [0, ∞) in R is denoted by R
+
. Symbol ∗
1
means the L
1
-norm of vectors, which is denoted by
x
1
=
n
k=1
| x
k
|, where x
k
is the kth element of x.
P 0( 0) means all the elements of P are positive
(nonnegative). N denotes the set of nonnegative inte-
ger numbers and N
+
= N/{0}. I
n×n
means the n-order
identity matrix. 1
p
n
means that the pth element of the
vector 1
n
is one and the others are zero. λ
min
(χ
ii
,χ
ij
)
denotes the minimum element of vectors χ
ii
and χ
ij
.
λ
max
(χ
ii
,χ
ij
) denotes the maximum element of vec-
tors χ
ii
and χ
ij
.
2 Problem formulation
In this paper, we consider the switched positive non-
linear system described by
˙x(t) = A
σ(t)
x(t ) + A
f
f (x (t)) + B
σ(t)
u(t) + D
σ(t)
w(t),
y(t) = C
σ(t)
x(t ), (1)
where x(t) = (x
1
(t), x
2
(t),...,x
n
(t))
T
∈ R
n
+
is the
state vector; y(t) ∈ R
l
+
is the measurable out vector;
u(t ) ∈ R
m
is the control input; f (x(t)) = ( f
1
(x(t)),
f
2
(x(t)), . . . , f
n
(x(t)))
T
∈ R
n
+
is the nonlinear func-
tion; w(t) ∈ R
p
+
is the external disturbance. The func-
tion σ(t) ∈[0, +∞) → S ={1, 2,...,N } is the
piecewise constant switching signal that may depend
on time t. t
k
means the kth switching of system. τ(t)
is the delay between the switching of system and the
switching of controller, where 0 ≤ τ(t) ≤ τ
s
and
τ
s
< t
k+1
− t
k
. A
σ(t)
, A
f
, B
σ(t)
, C
σ(t)
, D
σ(t)
are
the parameter matrices with appropriate dimensions,
where B
σ(t)
is full column rank, the pairs (A
σ(t)
, B
σ(t)
),
and ( A
σ(t)
, C
σ(t)
) are controllable and observable.
When σ(t) = i , we define A
i
= A
σ(t)
, B
i
= B
σ(t)
,
C
i
= C
σ(t)
, D
i
= D
σ(t)
for the purpose of simplifica-
tion.
123