Professional Documents
Culture Documents
Controllability and Observability For A Class of Time-Varying Impulsive Systems On Time Scales
Controllability and Observability For A Class of Time-Varying Impulsive Systems On Time Scales
Abstract. The aim of this paper is to study the controllability and ob-
servability for a class of linear time-varying impulsive control systems on time
scales. Sufficient and necessary conditions for state controllability and state
observability of such systems are established. The corresponding criteria for
time-invariant impulsive control systems on time scales are also obtained.
Keywords: time scale, linear impulsive control system, controllability,
observability.
AMS Subject Classification: 93B05, 34A37, 34H05, 93B07.
1 Introduction
Differential equations with impulses have a considerable importance in varied
applications as physics, engineering, biology, medicine, economics, neuronal
networks, social sciences, and so on. Many investigations have been car-
ried out concerning the existence, uniqueness, and asymptotic properties of
solutions. We refer to the monographs [7, 11, 29, 40] and the references
therein. It is well known that the study of controllability plays an important
role in the control theory. In recent years, some research dealing with the
study of controllability for impulsive systems [10, 16, 23, 32, 34, 41, 44, 47].
The most dynamical systems are analyzed in either the continuous or dis-
crete time domain. The population dynamical models in continuous time
are usually appropriate for organism that have overlapping generations. On
other hand, many biological populations are more accurately described by
non-overlapping generations. The dynamics of these populations often are
1
Corresponding author
2 Preliminaries
Let Rn be the space of n-dimensional column vectors x = col(x1 , x2 , ...xn )
with a norm || · ||. A time scale T is a nonempty closed subset of R. The
notations [a, b], [a, b), and so on, will denote time scales intervals such as
[a, b] := {t ∈ T; a ≤ t ≤ b}, where a, b ∈ T. The set of all rd-continuous
functions f : T → Rn will be denoted by Crd (T, Rn ). A function f : T → Rn
is piecewise rd-continuous (we write f ∈ Cprd (T, Rn )) if it is regulated and if
it is rd-continuous at all, except possibly at finitely many, right-dense points
t ∈ T.
1
We denote by Crd (T, Rn ) the set of all functions f : T → Rn that are
differentiable on T and its delta-derivative f ∆ (t) ∈ Crd (T, Rn ). The set
of rd-continuous (respectively rd-continuous and regressive) matrix-valued
functions A : T → Mn (R) is denoted by Crd (T, Mn (R)) (respectively by
Crd R(T, Mn (R))). We recall that a matrix-valued function A is said to be
The following theorem shows that we can express the matrix exponential
as a finite sum of powers of the matrix A with infinitely rd-continuous delta
differentiable functions as coefficients.
Proposition 2 ([19, Theorem 5.1]). For the system (3) with A ∈ Mn (R)
∞
constant, there exist scalar functions γ0 (t, τ ), ..., γn−1 (t, τ ) ∈ Crd (T+ , R) such
that the unique solution has representation
n−1
X
eA (t, τ ) = γi(t, τ )Ai . (4)
i=0
Lemma 1. For any t ∈ (tk−1 , tk ], k = 1, 2, ..., the solution of the initial value
problem( 1) is given by
k−1
Y 1
Y
x(t) = ΦAk (t, tk−1 ) (1 + ci ) ΦAi (ti , ti−1 )x0
i=1 i=k−1
k−1 k−1
Rt X Y R ti
+ Φ (t, σ(τ ))Bk (τ )u(τ )∆τ +
tk−1 Ak
[ (1 + cj ) ti−1 ΦAk (t, tk−1 ) (5)
i=1 j=i
i+1
Y
× ΦAr (tr , tr−1 )ΦAi (ti , σ(τ ))Bi (τ )u(τ )∆τ ].
r=k−1
Since
x(t+
1 ) = (1 + c1 )x(t1 ) =
R t1
= (1 + c1 )ΦA1 (t1 , t0 )x0 + (1 + c1 ) t0
ΦA1 (t1 , σ(τ ))B1 (τ )u(τ )∆τ
it follows that
x(t) = ΦA2 (t, t1 )(1 + c1 )ΦA1 (t1 , t0 )x0 +
Z t1
+(1 + c1 ) ΦA2 (t, t1 )ΦA1 (t1 , σ(τ ))B1 (τ )u(τ )∆τ
t0
Rt
+ t1
ΦA2 (t, σ(τ ))B2 (τ )u(τ )∆τ
and so, (5) is true for k = 2. Next, suppose that (5) is true for k = p, that
is, for t ∈ (tp−1 , tp ], we have
p−1
Y 1
Y
x(t) = ΦAp (t, tp−1 ) (1 + ci ) ΦAi (ti , ti−1 )x0
i=1 i=p−1
p−1 p−1
Rt X Y R ti
+ Φ (t, σ(τ ))Bp (τ )u(τ )∆τ +
tp−1 Ap
[ (1 + cj ) ti−1 ΦAp (t, tp−1 )
i=1 j=i
i+1
Y
× ΦAr (tr , tr−1 )ΦAi (ti , σ(τ ))Bi (τ )u(τ )∆τ ].
r=p−1
p−1 p
R tp X Y R ti
+(1 + cp ) tp−1 ΦAp (tp , σ(τ ))Bp (τ )u(τ )∆τ + [ (1 + cj ) ti−1 ΦAp (tp , tp−1)
i=1 j=i
i+1
Y
× ΦAr (tr , tr−1 )ΦAi (ti , σ(τ ))Bi (τ )u(τ )∆τ ].
r=p−1
It follows that
p 1
Y Y
x(t) = ΦAp+1 (t, tp ) (1 + ci ) ΦAi (ti , ti−1 )x0
i=1 i=p
p p
Rt X Y R ti
+ Φ
tp Ap+1
(t, σ(τ ))Bp+1 (τ )u(τ )∆τ + [ (1 + cj ) ti−1 ΦAp+1 (t, tp )
i=1 j=i
i+1
Y
× ΦAr (tr , tr−1 )ΦAi (ti , σ(τ ))Bi (τ )u(τ )∆τ ],
r=p
3 Controllability
Definition 1. The impulsive system (1) is called controllable on [t0 , tf ],
with tf > t0 , if given any initial state x0 ∈ Rn there exists a piecewise
rd-continuous input signal u(·) : [t0 , tf ] → Rm such that the corresponding
solution of (1) satisfies x(tf ) = 0.
where Ψ1 (τ ) := Ψ1 (t0 , σ(τ ) = ΦA1 (t0 , σ(τ )), for τ ∈ (t0 , t1 ], and
i−1
Y
Ψi (τ ) := Ψi (t0 , σ(τ )) = ΦAj (tj−1 , tj )ΦAi (ti−1 , σ(τ )), τ ∈ (ti−1 , ti ], (8)
j=1
for i = 2, 3, ..., k.
If Ak (t) = Ak and Bk (t) = Bk are constant matrices then
Z ti
Gi := G(t0 , ti−1 , ti ) = Ψi (t0 , σ(τ ))Bi BiT ΨTi (t0 , σ(τ ))∆τ , (9)
ti−1
where Ψ1 (τ ) := Ψ1 (t0 , σ(τ ) = eA1 (t0 , σ(τ )), for τ ∈ (t0 , t1 ], and
i−1
Y
Ψi (τ ) := Ψi (t0 , σ(τ )) = eAj (tj−1 , tj )eAi (ti−1 , σ(τ )), τ ∈ (ti−1 , ti ], (11)
j=1
for i = 2, 3, ..., k.
The Gramian matrix in the case of time scales was defined in [21]. The
above definition is adopted from [21] for impulsive case. Now we are formu-
lating the results for controllability.
Theorem 1. (i) If there exists at least l ∈ {1, 2, ..., k} such that rank(Gl ) =
n, then the impulsive system (1) is controllable on [t0 , tf ] (tf ∈ (tk−1 , tk ]).
(ii) Assume that ci 6= −1, i = 1, 2, ..., k − 1. If the impulsive system (1)
is controllable on [t0 , tf ] (tf ∈ (tk−1 , tk ]), then
rank(G0 G1 ... Gk ) = n. (12)
Proof. (i) Let l ∈ {1, 2, ..., k} be such that rank(Gl ) = n, that is, G(t0 , tl−1 , tl )
is invertible. Then for a given x0 ∈ Rn , choose
al BlT (t)ΨTl G−1
l x0 if t ∈ (tl−1 , tl ]
u(t) = (13)
0 if t ∈ [t0 , tf ] r (tl−1 , tl ],
l+1
R tl Y
tl−1
ΦAk (tf , tk−1) ΦAr (tr , tr−1 )ΦAl (tl , σ(τ ))Bl (τ )BlT (τ )ΨTl (τ )G−1
l ∆τ ]x0 .
r=k−1
Since
l+1
Y
ΦAr (tr , tr−1 )ΦAl (tl , σ(τ ))Ψ−1
l (τ ) = ΦAk−1 (tk−1 , tk−2 )...ΦAl (tl , σ(τ ))
r=k−1
1
Y
×ΦAl (σ(τ ), tl−1 )ΦAl−1 (tl−1 , tl−2 )...ΦA1 (t1 , t0 ) = ΦAi (ti , ti−1 ),
i=k−1
it follows that
k−1
Y 1
Y
x(tf ) = ΦAk (tf , tk−1 ) (1 + ci ) ΦAi (ti , ti−1 )x0
i=1 i=k−1
k−1 l+1
Y R tl Y
+[ (1 + cj )al Φ (t , t )
tl−1 Ak f k−1
ΦAr (tr , tr−1 )ΦAl (tl , σ(τ ))Ψ−1
l (τ )
j=l r=k−1
Therefore, we obtain
"k−1 k−1
# 1
Y Y Y
x(tf ) = (1 + ci ) + (1 + cj )al ΦAk (tf , tk−1 ) ΦAi (ti , ti−1 ) = 0,
i=1 j=l i=k−1
k−1 k−1
R tf X Y
+ Φ (t , σ(τ ))Bk (τ )u(τ )∆τ +
tk−1 Ak f
[ (1 + cj )× (15)
i=1 j=i
i+1
R ti Y
ti−1
ΦAk (tf , tk−1 ) ΦAr (tr , tr−1 )ΦAi (ti , σ(τ ))Bi (τ )u(τ )∆τ ].
r=k−1
Z ti
× ΦAi (ti , σ(τ ))Bi (τ )u(τ )∆τ ]
ti−1
Z tf
−ΦA1 (t0 , t1 )ΦA2 (t1 , t2 )...ΦAk (tk−1 , tf ) ΦAk (tf , σ(τ ))Bk (τ )u(τ )∆τ.
tk−1
Z tf
− xTα Ψk (t0 , σ(τ ))Bk (τ )u(τ )∆τ = 0.
tk−1
k
(1 + cj ) 6= 0, it follows that xα xTα = 0. This contradicts xα 6= 0 and
Q
Since
j=1
so we conclude that rank(G0 G1 ... Gk ) = n.
where Wi = Λi (Bi Ai Bi ... Ain−1 Bi ) for i = 1, 2, ..., k−1, Wk = Λk−1 eAk (tk−1 , tf )
(Bk Ak Bk ... Akn−1 Bk ), and Λi = eA1 (t0 , t1 )eA2 (t1 , t2 )...eAi (ti−1 , ti ).
Proof. Suppose that the impulsive system (1) is controllable on [t0 , tf ]. If the
rank condition (16) does not hold, then there exists nonzero xα ∈ Rn such
that
xTα Λi Aji Bi = 0,
Z ti
= xTα Λi−1 eAi (ti−1 , σ(τ ))Bi BiT Ψi (τ )∆τ
ti−1
Z ti
= xTα Λi−1 eAi (ti−1 , ti )eAi (ti , σ(τ ))Bi BiT Ψi (τ )∆τ
ti−1
Z ti
= xTα Λi eAi (ti , σ(τ ))Bi BiT Ψi (τ )∆τ
ti−1
" n−1 #
Z ti X
= γij (ti , σ(τ ))xTα Λi Aji Bi BiT Ψi (τ )∆τ = 0
ti−1 j=0
which implies that the rank condition (16) fails. This contradiction proves
that the impulsive system (1) is controllable on [t0 , tf ] (tf ∈ (tk−1 , tk ]).
where
Ψ1 (0, σ(t)) = eA1 (0, σ(t)) t ∈ (0, t1 ],
and
i−1
Y
Ψi (0, σ(t)) = eAj (tj−1 , tj )eAi (ti−1, σ(t)) t ∈ (ti−1 , ti ], i = 2, 3.
j=1
1 6
e − 41 e−2 e − 41 e−5/2 − 14 e11/2
1 3/2
G2 = 4 2 ,
1 3/2
2
e − 14 e−5/2 − 14 e11/2 11 −3
4
e − e + 14 e5
and
a b
G3 = ,
b c
where
2
a = (8e−6 − 2e−16 − e−26 )
125
2
b = (2e−21/2 − 8e−13/2 + 2e−33/2 − e−41/2 + e−53/2 − e−61/2 )
125
2
c = (8e−7 − 4e−11 + e−15 − 2e−17 + 2e−21 − e−27 + 2e−31 − e−35 ).
125
Then we obtain
and
a b
, t ∈ ( 25 , 3]
b c
Ψ3 (0, t)B3 (t)B3T (t)ΨT3 (0, t) = (29)
d e
, t ∈ [4, 29 ]
e f
where
1 21−10t 1
a = (e − e17/2−5t + e−4 )
25 4
1 1 6−5t 1 1
b = ( e + e8−5t − e37/2−10t − e41/2−10t − e−9/2 )
25 4 2 4
1 1 16−10t 1 1
c = ( e + e18−10t + e20−10t − e11/2−5t − e15/2−5t + e−5 )
25 4 2 4
1 1 31−10t 2 27/2−5t
d = ( e − e + e−4 )
100 9 3
1 1 11−5t 2 13−5t 1 1
e = ( e + e − e57/2−10t − e61/2−10t − e−9/2 )
100 6 3 18 9
1 1 26−10t 1 28−10t 1 30−10t 1 21/2−5t 2 25/2−5t
f = ( e + e + e − e − e + e−5 ).
100 36 9 9 3 3
e2 − 41 e−2 − 43 7 −1/2
− 21 e−3/2 − e3/2 − 14 e−5/2
4
e
G2 = 7 −1/2 ,
4
e − 21 e−3/2 − e3/2 − 41 e−5/2 e − 11
4
e−1 + e−2 + 74 e−3
and
a b
G3 =
b c
where
1
a = − (−54e−4 + 23e−9 + e−14 − 60e−13/2 )
9000
1
b = − (120e−7 + 30e−9 + 108e−9/2 − 46e−19/2 − 29e−23/2 −
18000
−2e−29/2 − e−33/2 )
1
c = (216e−5 + 36e−9 − 92e−10 − 116e−12 − 35e−14 − 4e−15 − 4e−17
36000
−e−19 + 240e−15/2 + 120e−19/2 ).
Then
4 Observability
Consider the following impulsive dynamical system
∆
x = Ak (t)x + Bk (t)u, t ∈ [tk−1 , tk ),
x(t+
k ) = (1 + ck )x(tk ), k = 1, 2, ...,
(30)
y(t) = Ck (t)x + Dk (t)u,
x(t0 ) = x0 ,
is invertible, where
R ti
M(t0 , ti−1 , ti ) = ti−1 ΩTi (τ, t0 )CiT (τ )Ci (τ )Ωi (τ, t0 )∆τ , i = 1, 2, ..., k − 1,
R tf
M(t0 , tk−1, tf ) = tk−1
ΩTk (τ, t0 )CkT (τ )Ck (τ )Ωk (τ, t0 )∆τ,
and
Ωi (τ, t0 ) = ΦAi (τ, ti−1 )ΦAi−1 (ti−1 , ti−2 )...ΦA1 (t1 , t0 )
for τ ∈ (ti−1 , ti ] and i = 1, 2, ..., k.
Proof . Suppose that M(t0 , tf ) is invertible. From (5) and (30) we obtain
Z t
y(t) = C1 (t)ΦA1 (t, t0 )x0 + C1 (t) ΦA1 (t, σ(τ ))B1 (τ )u(τ )∆τ + D1 (t)u(t)
t0
(31)
l−1
Q 1
Q
= Cl (t)ΦAl (t, tl−1 ) (1 + ci ) ΦAi (ti , ti−1 )x0 +
i=1 i=l−1
" #
l−1 l−1
Z ti i+1
P Q Q
Cl (t) (1 + cj ) ΦAl (t, σ(τ )) ΦAr (tr , tr−1 )ΦAi (ti , σ(τ ))Bi (τ )u(τ )∆τ
i=1 j=i ti−1 r=l−1
Z t
+Cl (t) ΦAl (t, σ(τ ))Bl (τ )u(τ )∆τ + Dl (t)u(t),
tl−1
(32)
for t ∈ (tl−1 , tl ], l = 2, 3, ...k. It is easy to see from the Definition 2 that the
observability of system (30) is equivalent to the observability of y(t) given by
C1 (t)ΦA1 (t, t0 )x0 , t ∈ [t0 , t1 ]
y(t) = l−1
Y (33)
(1 + ci )Cl (t)Ωl (t, t0 )x0 , t ∈ (tl−1 , tl ], l = 1, 2, ..., k,
i=1
as u(t) = 0. Now, multiplying ΩTl (t, t0 )ClT (t) to both sides of (33) and
integrating with respect to t from t0 to tf , we have
Z tf Z t1
ΩTl (τ, t0 )ClT (τ )y(τ )∆τ = ΦTA1 (τ, t0 )C1T (τ )C1 (τ )ΦA1 (τ, t0 )∆τ
t0 t0
k−1 ti
P i−1
Z
ΩTi (τ, t0 )CiT (τ )Cl (τ ) Ωi (τ, t0 )∆τ
Q
+ (1 + cj )
i=2 j=1 ti−1
#
k−1
Z tf
ΩTk (τ, t0 )CkT (τ )Ck (τ )Ωk (τ, t0 )∆τ x0
Q
+ (1 + cj )
j=1 tk−1
k−1
P i−1
Q
= [M(t0 , t0 , t1 ) + (1 + cj )M(t0 , ti−1 , ti ) (34)
i=2 j=1
k−1
Q
+ (1 + cj )M(t0 , tk−1, tf )]x0 .
j=1
Obviously, the left-hand side of (34) depend on y(t), t ∈ [t0 , tf ]. Since the
matrix M(t0 , tf ) is invertible, so from linear algebraic equation (34) we de-
duce that x(t0 ) = x0 is uniquely determined by the corresponding system
output y(t) for t ∈ [t0 , tf ].
Conversely, if we suppose that the matrix M(t0 , tf ) is not invertible, then
there exist nonzero xα ∈ Rn such that xTα M(t0 , tf )xα = 0. Since 1 + ci ≥ 0,
i = 1, 2....., k, M(t0 , ti−1 , ti ), i = 1, 2, ....k − 1, and M(t0 , tk−1, tf ) are positive
semidefinite matrices, we have
Z t1
= xTα ΦTA1 (τ, t0 )C1T (τ )C1 (τ )ΦA1 (τ, t0 )xα ∆τ
t0
" #2 Z
k−1 i−1 ti
xTα ΩTi (τ, t0 )CiT (τ )Ci (τ )Ωi (τ, t0 )xα ∆τ
P Q
+ (1 + cj )
i=2 j=1 ti−1
" #2 Z
k−1 tf
xTα ΩTk (τ, t0 )CkT (τ )Ck (τ )Ωk (τ, t0 )xα ∆τ.
Q
+ (1 + cj )
j=1 tk−1
" #2
k−1 i
= xTα M(t0 , t0 , t1 )xα + xTα M(t0 , ti−1 , ti )xα
P Q
(1 + cj )
i=1 j=1
" #2
k−1
xTα M(t0 , tk−1 , tf )xα = 0
Q
+ (1 + cj )
j=1
and so, Z tf
ky(τ )k2 ∆τ = 0.
t0
It follows that
C1 (t)ΦA1 (t, t0 )x0 , t ∈ [t0 , t1 ],
l−1
Q (1 + c )C (t)Ω (t, t )x , t ∈ (t , t ], l = 1, 2, ...k − 1,
j l l 0 α l−1 l
0 = y(t) = j=1
k−1
Q
(1 + cj )Ck (t)Ωk (t, t0 )xα , t ∈ (tk−1 , tf ].
j=1
The last equality implies, by Definition 2, that the impulsive system is not
observable on [t0 , tf ] (tf ∈ (tk−1 , tk ]).
Z ti
= [Ci Ωi (τ, t0 )xα ]T [Ci Ωi (τ, t0 )xα ]∆τ = 0, i = 1, 2, ..., k − 1
ti−1
and
Z tf
xTα M(t0 , tk−1, tf )xα = [Ci Ωk (τ, t0 )xα ]T [Ci Ωk (τ, t0 )xα ]∆τ = 0.
tk−1
Since Ωi (τ, t0 ) = eAi (τ, ti−1 )...eA2 (t2 , t1 )eA1 (t1 , t0 ) for i = 1, 2, ..., k, it follows
that
Ci eAi (τ, ti−1 )...eA2 (t2 , t1 )eA1 (t1 , t0 )xα = 0 (37)
for τ ∈ (ti−1 , ti ], i = 1, 2, ..., k − 1, and
Ck eAk (τ, tk−1 )...eA2 (t2 , t1 )eA1 (t1 , t0 )xα = 0 (38)
for τ ∈ (tk−1 , tf ]. Obviously, at τ = ti−1 , we have Ci Υi−1 xα = 0, i = 1, 2, ..., k,
and differentiating in (37) and (38) j times and evaluating the result at
τ = ti−1 , i = 1, 2, ..., k, we obtain
Ci Aji Υi−1 xα = 0, i = 1, 2, ..., k, j = 0, 1, 2, ...n − 1. (39)
where
2 0
A1 = , C1 = 0 e−3 (0, t)
1 3
1 2
0 e3 ( 21 , t)
A2 = , C2 = (41)
0 3
−3 −2
0 e3 ( 25 , t)
A3 = , C3 = .
3 4
5/2
Z
M(0, 12 , 25 ) = ΩT2 (τ, 0)C2T (τ )C2 (τ ) Ω2 (τ, 0)∆τ (42)
1/2
9/2
Z
5 9
M(0, , )
2 2
= ΩT3 (τ, 0)C3T (τ )C3 (τ ) Ω3 (τ, 0)∆τ,
5/2
and
Ωi (s, 0) = ΦAi (s, ti−1 )ΦAi−1 (ti−1 , ti−2 )...ΦA1 (t1 , 0), s ∈ (ti−1 , ti ], i = 1, 2, 3.
If T = R then
1 1
(−e5 + 1) − 11 −e11/2 + 1
1 − 10
M(0, 0, ) = 1 11/2 1 ,
2 − 11 −e +1 − 12 (−e6 + 1)
2e2 2e5/2
1 5
M(0, , ) =
2 2 2e5/2 2e3
1 1
(−e5 + 1) − 11 −e11/2 + 1
1 − 10
M(0, 0, ) = 1 1 ,
2 − 11 −e11/2 + 1 − 12 (−e6 + 1)
e2 e5/2
1 5
M(0, , ) =
2 2 e5/2 e3
and
5 9 a1 a2
M(0, , ) =
2 2 a2 a3
5 Applications
5.1. Consider the following application to population growth model with
impulse
△
N (t) = rk N(t) + ck U(t), t 6= tk ,
N(t+k ) = (rk+1 − rk )N(tk ), t = tk ,
N(0) = N0 ,
where N(t) is the number of population at the time t, rk is the rate of popu-
lation growth between two consecutive impulsive points and U(t) is a control
input. Such model can be describe the evaluation of cicada magicicada sep-
tendecim. In this case is need to consider the time scale T = P1,1 (see [13,
Example 1.39] ) Using the Theorem 2 it is easy to see that the system is
controllable.
5.2. Next application is a impulsive model in Nonelectronic [44, Example
6 Conclusion
In this paper, the issue on the controllability and observability criteria for
linear impulsive time-varying systems on time scales has been addressed. Sev-
eral sufficient and necessary criteria for state controllability and observability
of such systems have been established, respectively, by the variation of pa-
rameters for time-varying impulsive systems on time scales. In addition, two
examples and two applications have been presented to show the effectiveness
of proposed results. As it has been shown that a larger class of systems are
considered, the results generalize some known results in [8, 21, 23, 25, 36, 47].
References
[1] R. P. Agarwal, M. Bohner, Basic calculus on time scales and some of its
applications, Results Math., 35 (1999) 3–22.
[33] H. Liu, X. Xiang, A class of the first order impulsive dynamic equations
on time scales, Nonlinear Analysis, 69 (2008) 2803–2811.