Professional Documents
Culture Documents
Research Article
Impulsive Controllability/Observability for Interconnected
Descriptor Systems with Two Subsystems
Copyright © 2015 Qingling Zhang et al. This is an open access article distributed under the Creative Commons Attribution License,
which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
The problem of decentralized impulse controllability/observability for large-scale interconnected descriptor systems with two
subsystems by derivative feedback is studied. Necessary conditions for the existence of a derivative feedback controller for the
first subsystem of the large-scale interconnected descriptor systems ensuring the second subsystem to be impulse controllable
and impulse observable are derived, respectively. Based on the results, a derivative feedback controller for the first subsystem of
the large-scale interconnected descriptor systems is constructed easily such that the second subsystem is impulse controllable or
impulse observable. Finally, examples are given to illustrate the effectiveness of the results obtained in this paper.
is necessary and sufficient for the existence of impulsive to be considered. That means that system equations of the
mode eliminating proportional state feedback controller. The vehicles should include the algebraic equations. In most cases,
concept of impulsive controllability is proposed in [14]. It is the parameters of hypersonic technology systems change
shown [15, 16] that the condition of impulse elimination via all the time. When derivative coefficient matrix is singular
proportional plus derivative (P-D) state feedback is that the matrix, we can describe the hypersonic technology systems
original system is impulsive controllable one. New criteria as descriptor systems. So, the impulse problem of vehicle
are proposed in [17] for impulsive mode controllability systems is the first problem to be solved. However, to the
of descriptor linear systems by adopting the null space best of our knowledge, the research on impulse control of
approach. In [18], the problem of impulse elimination via pro- hypersonic vehicle still receives little attention. Therefore,
portional plus derivative (P-D) output feedback has also been the research problem of hypersonic technology impulse
studied. In [19], a decentralized (P-D) feedback is explored for controllability is overwhelmingly crucial.
the impulsive elimination problem where the centralized (P- As is known to all, for the descriptor systems, propor-
D) feedback is regarded as a special case. In [20], a concept tional feedback can not alter the impulse controllability of
of the structured proportional plus derivative (P-D) feedback systems. Therefore, this paper focuses on designing a deriva-
is first introduced and an explicit necessary and sufficient tive feedback controller for the first subsystem of the large-
condition is constructively derived for the closed-loop system scale interconnected descriptor system such that the second
to be regular and impulse-free by the structured proportional subsystem is impulse controllable or impulse observable. The
plus derivative (P-D) feedback. A rank condition of impulse impulse controllability/observability problem of descriptor
observability is presented for regular descriptor systems [21]. systems involves the rank constraints of matrix pencil with
In [12], the presence of impulsive responses in descriptor parameter 𝐾. Firstly, we use matrix theory to transform the
systems and how they relate to impulse controllability and matrix pencil involving parameter 𝐾 into the form of 𝐴 +
impulse observability are considered. 𝐵𝐾𝐶. Further by comparing the maximum rank of the left
The large-scale interconnected descriptor systems are hand of the condition of impulse controllability/observability
a kind of dynamic systems which are more general and with the minimum rank of the right hand of the condition
have extensive applied background, such as power systems, of impulse controllability/observability, necessary conditions
economic systems, and network systems [22]. The class of for the existence of a derivative feedback controller for the
systems can be characterized by a large number of state first subsystem ensuring the second subsystem to be impulse
variables, parametric uncertainties, a complex structure, and controllable and impulse observable are derived, respectively.
a strong interaction between subsystems [23]. Decentral- If the maximum rank of the left hand of the condition of
ized/interconnected descriptor systems have attracted more impulse controllability is lower than the minimum rank of
and more attention because of their practical backgrounds; the right hand of that, the controller to be designed will
for example, [24, 25] investigated dynamic networks gov- not exist. If the maximum rank of the left hand of the
erned by decentralized/descriptor systems, respectively. The condition of impulse controllability/observability is higher
related research has been widely used in the research of than the minimum rank of the right hand of that, the
multiagent robot and four-rotor aircraft [26]. intersection of the set of the ranks of the left hand and the
Decentralized/interconnected descriptor systems also set of the ranks of right hand of the condition of impulse
contain impulse terms. However, to the best of our knowl- controllability/observability is not an empty set, and the
edge, the research on decentralized impulse controllabil- controller to be designed will exist and can be selected in
ity/observability of interconnected descriptor systems still the intersection. Finally, illustrative examples are provided to
receives little attention. So in this paper, we study the problem demonstrate the applicability of the proposed method.
of decentralized impulsive controllability/observability for This paper is structured as follows. In Section 2, defini-
interconnected descriptor systems. The derivative coefficient tions and lemmas are introduced for later use. Section 3 gives
matrix (𝐸 + 𝐵𝐾) is not required to be revertible, so our conditions of the impulsive controllability and impulsive
results include previous related ones as special cases. Decen- observability by derivative feedback. In Section 4, illustrative
tralized impulsive controllability/observability is investigated examples are proposed to testify the feasibility of the theo-
by using matrix rank properties, which is a novel way to rems. Finally, conclusive remarks are made in Section 5.
deal with the impulse. The dynamic order of the closed-loop
descriptor systems can be assigned between the minimum
rank and maximum rank of (𝐸 + 𝐵𝐾). 2. Preliminaries
Hypersonic technology is the best stratagem in the realm
of aerospace technology in the 21st century. The hypersonic Consider a class of large-scale interconnected descriptor
vehicle is characterized by its high speed and high ability systems composed of two subsystems of the following form:
of penetration. So it is of great importance in both civilian
field and military field. Because of the advanced aerodynamic 𝐸1 𝑥1̇ = 𝐴 11 𝑥1 + 𝐴 12 𝑥2 + 𝐵1 𝑢1 ,
configuration, fuselage-engine integrated design, large flight 𝐸2 𝑥2̇ = 𝐴 21 𝑥1 + 𝐴 22 𝑥2 + 𝐵2 𝑢2 ,
envelope, and the tremendous changes of flight environment, (1)
the hypersonic aircraft has poor stability, serious couplings 𝑦1 = 𝐶1 𝑥1 ,
among the subsystems, and biggish model uncertainty. Mean-
while, various restrictions for the flight of the vehicles need 𝑦2 = 𝐶2 𝑥2 ,
Mathematical Problems in Engineering 3
where Σ11 ∈ 𝑅(𝑟𝑒1 𝑏1 −𝑟𝑏1 )×(𝑟𝑒1 𝑏1 −𝑟𝑏1) , Σ1𝐵 ∈ 𝑅𝑟𝑏1 ×𝑟𝑏1 , Σ21 ∈ Lemma 3 (see [2, 21]). For arbitrary matrices A ∈ 𝑅𝑛×𝑟 , B ∈
𝑅(𝑟𝑒2 𝑏2 −𝑟𝑏2 )×(𝑟𝑒2 𝑏2 −𝑟𝑏2 ) , and Σ2𝐵 ∈ 𝑅𝑟𝑏2 ×𝑟𝑏2 are diagonal pos- 𝑅𝑛×𝑚 , and C ∈ 𝑅𝑙×𝑟 the following relation holds:
1
itive definite matrices. 𝐸21 ∈ 𝑅𝑟𝑏1 ×(𝑟𝑒1 +𝑟𝑏1 −𝑟𝑒1 𝑏1 ) , 𝐸22
1
∈
𝑟𝑏1 ×(𝑟𝑒1 +𝑟𝑏1 −𝑟𝑒1 𝑏1 ) 2 𝑟𝑏2 ×(𝑟𝑒2 +𝑟𝑏2 −𝑟𝑒2 𝑏2 ) 2
(a) min {rank (A + BK)}
𝑅 , 𝐸21 ∈ 𝑅 , and 𝐸22 ∈ K
(22)
𝑅𝑟𝑏2 ×(𝑟𝑒2 +𝑟𝑏2 −𝑟𝑒2 𝑏2 ) are full column rank matrices: = rank ([A B]) − rank (B) ,
𝐴11 11 11
11 𝐴 12 𝐴 13
[ 11 11 11 ] where minK {∗} is the minimum rank of matrix pencil {∗} with
𝑈1 𝐴 11 𝑉1 = [ ]
[𝐴 21 𝐴 22 𝐴 23 ] , parameter matrix K ∈ 𝑅𝑚×𝑟 .
11 11 11 Consider the following:
[𝐴 31 𝐴 32 𝐴 33 ]
(b) min {rank (A + BKC)}
𝐴12 12 12
11 𝐴 12 𝐴 13
K
[ 12 12 12 ]
𝑈1 𝐴 12 𝑉2 = [ ]
[𝐴 21 𝐴 22 𝐴 23 ] , A
12 12 12
= rank ([A B]) + rank ([ ])
[𝐴 31 𝐴 32 𝐴 33 ] C (23)
𝐴21 21 21
11 𝐴 12 𝐴 13
A B
[ 21 21 21 ] (16) − rank ([ ]) ,
𝑈2 𝐴 21 𝑉1 = [ ] C 0
[𝐴 21 𝐴 22 𝐴 23 ] ,
21 21 21 where minK {∗} is the minimum rank of matrix pencil {∗} with
[𝐴 31 𝐴 32 𝐴 33 ]
parameter matrix K ∈ 𝑅𝑚×𝑙 .
𝐴22 22 22
11 𝐴 12 𝐴 13 Consider the following:
[ 22 22 22 ]
𝑈2 𝐴 22 𝑉2 = [ ]
[𝐴 21 𝐴 22 𝐴 23 ] , (c) max {rank (A + BKC)}
22 22 22 K
[𝐴 31 𝐴 32 𝐴 33 ] (24)
A
1
𝑈1 𝐶1 𝑉1 = [𝐶11 1
𝐶12 1
𝐶13 ], = min {rank ([A B]) , rank ([ ])} ,
C
2 2 2
𝑈2 𝐶2 𝑉2 = [𝐶11 𝐶11 𝐶11 ]. where maxK {∗} is the maximum rank of matrix pencil {∗} with
For linear descriptor systems of the form parameter matrix K ∈ 𝑅𝑚×𝑙 .
Consider the following:
𝐸𝑥̇ = 𝐴𝑥 + 𝐵𝑢,
(17)
𝑦 = 𝐶𝑥, 𝑀 = {K ∈ 𝑅𝑚×𝑙 | rank (A + BKC)
(25)
where 𝑥 ∈ 𝑅𝑛 is state vector, 𝑢 ∈ 𝑅𝑟 is control input, and 𝑦 ∈
𝑅𝑙 is controlled output, 𝐸, 𝐴 ∈ 𝑅𝑛×𝑛 , 𝐵 ∈ 𝑅𝑛×𝑟 , and 𝐶 ∈ 𝑅𝑙×𝑛 = max {rank (A + BKC)}} .
K
are constant matrices and rank(𝐸) = 𝑛0 ≤ 𝑛.
It is a Zariski open set, or equivalently almost every matrix
Definition 1 (see [6]). System (17) is said to be regular if there K ∈ 𝑅𝑚×𝑙 can make rank(A + BKC) to be maximal.
exists a constant scalar 𝑠 ∈ C such that
By Lemma 2, we have the following results.
det (𝑠𝐸 − 𝐴) ≠ 0. (18)
Lemma 2 (see [6]). (a) System (17) is impulse-free if and only Lemma 4. (a) System (6) is impulse controllable if and only if
if the following relation holds: the following relation holds:
𝐸 0 𝐸𝐾 0 0
rank ([ ]) = 𝑛 + rank (𝐸) . (19) rank ([ ]) = 𝑛 + rank (𝐸𝐾 ) . (26)
𝐴 𝐸 𝐴 𝐸𝐾 𝐵2
(b) System (17) is impulse controllable if and only if the (b) System (11) is impulse observable if and only if the
following relation holds: following relation holds:
𝐸 0 0
rank ([ ]) = 𝑛 + rank (𝐸) . (20) 𝐸𝐿 𝐴
𝐴 𝐸 𝐵 [ ]
rank ([ 0 𝐸𝐿 ]
[
]) = 𝑛 + rank (𝐸𝐿 ) . (27)
(c) System (17) is impulse observable if and only if the
following relation holds: [ 0 𝐶2 ]
𝐸 𝐴 Lemma 5 (see [29]). If 𝑟min = minK rank(A + BKC) and
[ 0 𝐸] 𝑟max = maxK rank(A + BKC), then there always exists K ∈
rank ([ ]) = 𝑛 + rank (𝐸) . (21)
𝑅𝑚×𝑙 such that 𝑟0 = rank(A + BKC) for any positive integer
[ 0 𝐶] 𝑟0 satisfying 𝑟min ≤ 𝑟0 ≤ 𝑟max .
Mathematical Problems in Engineering 5
Denote where
S1 = Diag {𝑈1 , 𝑈2 , 𝑈1 , 𝑈2 } , 𝐸𝐾 0 0
𝐾1𝑖 = {𝐾1 | rank ([ ]) = 𝑖} ,
(31) 𝐴 𝐸𝐾 𝐵 2
S2 = Diag {𝑉1 , 𝑉2 , 𝑉1 , 𝑉2 , 𝑄2 , 𝑄1 , 𝑄1 } . (35)
𝐾2𝑖 = {𝐾1 | 𝑛 + rank (𝐸𝐾 ) = 𝑖} ,
Diag{∗} denotes the block diagonal matrix. By Lemma 3, we
have then one can find 𝐾1 ∈ Ω𝐾 such that the closed-loop system is
impulse controllable.
𝐸𝐾 0 0
𝑅1 = max {rank ([ ])} Proof. The proof of Theorem 7 is similar to that of Theorem 6;
𝐾1 𝐴 𝐸𝐾 𝐵 2 the details are omitted.
𝐸 0 0 𝐵1 0 Remark 8. In the case of 𝑅1 < 𝑟1 , there does not exist
= max {rank ([ ])}
𝐾1 𝐴 𝐸 𝐵2 0 𝐵1 gain matrix 𝐾1 such that the closed-loop system is impulse
controllable. When 𝑅1 ≥ 𝑟1 , the existence of gain matrix 𝐾1
𝐸1 0 0 0 0 𝐵1 0 can be considered in terms of probability. Namely, assume
[ ] there exists a topological space, in which we can define a
[ 0 𝐸2 0 0 0 0 0]
= rank (S1 [ ] S2 ) ̃ that expresses the intersection of the dense set
nonnull set Ω
[𝐴 𝐴 𝐸 0 0 0 𝐵1 ]
[ 11 12 1 ] and any subset of that topological space. Then based on the
[𝐴 21 𝐴 22 0 𝐸2 𝐵2 0 0] ̃
perspective of probability, we can easily obtain this kind of Ω.
6 Mathematical Problems in Engineering
Based on this idea, we know that 𝐾1𝑅1 is a Zariski open Theorem 10. Consider system (1). If
set according to Lemma 3. We also know that almost every
𝐾1 can make rank([ 𝐸𝐴𝐾 𝐸0 𝐵0 ]) = 𝑅1 . Namely, 𝐾1𝑅1 is a dense 𝐸 𝐴 𝐵1 0
𝐾 2
[ ]
set. According to Lemma 5 we know that there always exists rank ([ ]
[ 0 𝐸 0 𝐵1 ])
𝐾1 such that 𝑛 + rank(𝐸𝐾 ) = 𝑅1 , and 𝐾2𝑅1 is a subset of
the topological space 𝑅𝑚×𝑛1 . Then based on the perspective [ 0 𝐶2 0 0 ]
̃ 𝐾 . If Ω
of probability, we can easily get this kind of Ω ̃ 𝐾 ≠ 0, we
𝐸 𝐴
̃
can find 𝐾1 ∈ Ω𝐾 such that the closed-loop system is impulse [ ]
[0 𝐸]
controllable, where [ ]
([ ])
̃ 𝐾 = 𝐾1𝑅 ∩ 𝐾2𝑅 . ≤ rank ( [
([ 0 𝐶2 ])
]) ,
Ω (36) [ ]
1 1
[𝐶 0]
[ 1 ] (42)
Remark 9. When Ω ̃ 𝐾 ≠ 0 and 𝑛 + rank(𝐸𝐾 ) = 𝑅1 , we can
obtain 𝐾1 as the following form: ([ 0 𝐶1 ])
−1
𝐴132 𝐴133 𝐴12
33
𝐾𝑑11 [− (Σ1𝐵 ) 1
𝐸22 0] + Δ [ 21 21 2 ]
𝐾1 = 𝑄[ ] 𝑉𝑇 , (37) [𝐴 𝐴 𝐴 ]
1
[ 22 23 23 ]
𝐾𝑑21 [𝐾𝑑22 𝐾𝑑23 ] rank ([ 21 21 2 ]) ≥ 𝑛 − 𝑟𝑒1 𝑏1 − 𝑟𝑏1 − 𝑟𝑒2 ,
[𝐴 𝐴 𝐴 ]
[ 32 33 33 ]
where 𝐾𝑑11 ∈ 𝑅𝑟𝑏 ×(𝑟𝑒1 𝑏1 −𝑟𝑏1) , 𝐾𝑑21 ∈ 𝑅(𝑙1 −𝑟𝑏1 )×(𝑟𝑒1 𝑏1 −𝑟𝑏1 ) , 𝐾𝑑22 ∈ [ 0 0 𝐶132
]
𝑅(𝑙1 −𝑟𝑏1 )×(𝑟𝑒1 +𝑟𝑏1 −𝑟𝑒1 𝑏1) , and 𝐾𝑑23 ∈ 𝑅(𝑙1 −𝑟𝑏1 )×(𝑛−𝑟𝑒1 ) are four
arbitrary parameters matrices, Δ ∈ 𝑅𝑟𝑏1 ×(𝑛−𝑟𝑒1 𝑏1 +𝑟𝑏1 ) is full Ω𝐿 = ⋃ {𝐿 1𝑖 ∩ 𝐿 2𝑖 } ≠ 𝜙,
𝑟2 ≤𝑖≤𝑅2
column rank parameter matrix, and rank(Δ) = 𝑅1 − 𝑛 − 𝑟𝑒1 𝑏1 +
𝑟𝑏1 −𝑟𝑒2 . At the same time, if 𝐾1 which is obtained by the above
process satisfies formula (27), then 𝐾1 is the gain matrix. Let where
Σ11 0 0 0 0 0 then one can find 𝐿 ∈ Ω𝐿 such that closed-loop system (11) is
[ 1 1 1 1 1 ]
[𝐸21 + Σ1 𝐾𝑑11 𝐸22 + Σ𝐵 𝐾𝑑12 Σ𝐵 𝐾𝑑13 0 0 0] impulse observable.
[ 1 1 ]
[ ]
[ 0 0 0 0 0 0] (39)
=[
[
],
]
Proof. Note that
[ 0 0 0 Σ21 0 0]
[ ]
[ 2 2 ]
[ 0 0 0 𝐸21 𝐸22 0] 𝐸𝐿 𝐴 𝐸 𝐴 𝐵1 0
[ ] [ ] [ ] 𝐶 0
[ 0 0 0 0 0 0] [ 0 𝐸 ] = [ 0 𝐸 ] + [ 0 𝐵 ] 𝐿1 [ 1 ], (44)
[ 𝐿] [ ] [ 1]
0 𝐶1
where 𝑈 = [ 𝑈1 𝑈2 ], 𝑉 = [ 𝑉1 𝑉2 ]. [ 0 𝐶2 ] [ 0 𝐶2 ] [ 0 0 ]
By rank(𝐸𝐾 ) = 𝑅1 − 𝑛 and (39) we can get the following
formula: where
1
rank ([𝐸22 + Σ1𝐵1 𝐾𝑑12 Σ1𝐵1 𝐾𝑑13 ]) 𝐿̃1 0
(40) 𝐿1 = [ ],
0 𝐿̃1
= 𝑅1 − 𝑛 − 𝑟𝑒1 𝑏1 + 𝑟𝑏1 − 𝑟𝑒2 .
(45)
𝐿1 0
Obviously, the matrix [𝐾𝑑12 𝐾𝑑13 ] can be expressed as the ̃1 = [
𝐿 ].
following form: 0 𝐿1
−1
[𝐾𝑑12 𝐾𝑑13 ] = [− (Σ1𝐵 ) 𝐸22
1
0] + Δ, Denote
1
(41)
rank (Δ) = 𝑅1 − 𝑛 − 𝑟𝑒1 𝑏1 + 𝑟𝑏1 − 𝑟𝑒2 . S3 = Diag {𝑈1 , 𝑈2 , 𝑈1 , 𝑈2 , 𝑈1 , 𝑈2 } ,
(46)
Next, we discuss the problem of impulse observability. S4 = Diag {𝑉1 , 𝑉2 , 𝑉1 , 𝑉2 , 𝑄1 , 𝑄1 } ,
Mathematical Problems in Engineering 7
𝐸1 + 𝐵1 𝐾1 0 0 0 0
[ ]
𝐸𝐾 0 0 [ 0 𝐸2 0 0 0]
rank ([ ]) = rank ([
[
])
𝐴 𝐸𝐾 𝐵2 [ 𝐴 11 𝐴 12 𝐸1 + 𝐵1 𝐾1 0 0]
]
[ 𝐴 21 𝐴 22 0 𝐸2 𝐵2 ]
𝑎 𝑏
+1 +3 0 0 0 0 0 0 0
[2 ]
[ 𝑎 + 2 2𝑏 + 6 0 0 0 0 0 0 0]
[ ]
([ ])
([ 0 0 −2.3 4.5 0 0 0 0 0] )
([ ])
([ ])
([ 0 0 1.15 −2.25 0 0 0 0 0] )
([ ])
= rank ([ ]) ≤ 5, (54)
([ 𝑎 𝑏 ])
([ −1 0.5 4 0.7 +1 +3 0 0 0] )
([ 2 2 ])
([ ])
([ −2 1 8 1.4 𝑎 + 2 𝑏 + 6 0 0 0 ])
[ ]
[ ]
[ −3 1.5 1 6 0 0 −2.3 4.5 2]
([ −1.5 0.75 −0.5 −3 0 0 1.15 −2.25 −1])
𝑎 𝑏
+1 +3 0 0
[2 2 ]
𝐸1 + 𝐵1 𝐾1 0 [ ]
[𝑎 + 2 𝑏 + 6 0 0 ]
𝑛 + rank (𝐸𝐾 ) = 𝑛 + rank ([ ]) = 4 + rank ([ ]) ≥ 6.
0 𝐸2 [ 0 0 −2.3 4.5 ]
[ ]
[ 0 0 1.15 −2.25]
Mathematical Problems in Engineering 9
For any derivative feedback gain 𝐾1 , equality (26) in Lemma 4 𝑅1 > 𝑟1 satisfies the necessary condition of impulse
can not hold. That means there does not exist 𝐾1 such that controllability. By Theorem 7, we can design proportional
closed-loop system (6) is impulse controllable. feedback controller for the first subsystem as follows:
𝐸𝐾 0 0 = 5,
𝑅1 = max {rank ([ ])}
𝐾1 𝐴 𝐸𝐾 𝐵2 1 0 0 0
[ ]
𝐸1 0 0 0 0 𝐵1 0 [0 0 0 0]
𝑛 + rank (𝐸) = 4 + rank ([ ]
[0 0 2 0]) = 6,
[ ] [ ]
[ 0 𝐸2 0 0 0 0 0]
= rank ([
[𝐴 𝐴 𝐸 0
])
[ 11 12 1 0 0 𝐵1 ]
] [0 0 0 0]
[𝐴 21 𝐴 22 0 𝐸2 𝐵2 0 0] 𝐸 0 0
rank ([ ]) ≠ 𝑛 + rank (𝐸) ,
= 6, 𝐴𝐾 𝐸 𝐵2
𝑛 + rank (𝐸𝐾 ) = 5 + rank ([1 + 𝑎 𝑏]) . Various restrictions for the flight of the hypersonic tech-
nology need to be considered. In most cases, the parameters
of hypersonic technology systems change all the time, with
In fact, by the above derivation, we can give arbitrary derivative coefficient matrix into singular matrix; by [29]
derivative feedback controller such that closed-loop system we describe the hypersonic technology systems as descriptor
(6) is impulse controllable. When 𝐾1 = [1 −1], we have systems. By Assumption 2.1 of [30] if the equilibrium of
nonlinear singular system 𝐸𝑥̇ = 𝑓(𝑥) is 𝑥𝑒 = 0 then
the locally nonimpulsiveness of this system is equivalent
𝐸𝐾 0 0 to (deg(det(𝑠𝐸 − 𝐴)) = rank(𝐸) ∀𝑠 ∈ C, where 𝐴 =
rank ([ ])
𝐴 𝐸𝐾 𝐵2 (𝜕𝑓/𝜕𝑥)(0)). Therefore nonlinear large-scale interconnected
descriptor systems (62) can be linearized at equilibrium point
2 −1 0 0 0 0 0 0 (𝛾, 𝛼, 𝛽, 𝜔𝑥 , 𝜔𝑦 , 𝜔𝑧 ) = (0, 0, 0, 0, 0, 0); then we obtain the
[ ] following system:
[0 0 2 0 0 0 0 0]
[ ]
([ ]) 𝐸1 𝑥1̇ = 𝐴 11 𝑥1 + 𝐴 12 𝑥2 + 𝐵1 𝑢1 ,
([ 1 2 7 3 2 −1 0 0])
= rank ( [
([ 5
]) = 6, (61) (63)
([ 6 4 2 0 0 0 0]])
) 𝐸2 𝑥2̇ = 𝐴 21 𝑥1 + 𝐴 22 𝑥2 + 𝐵2 𝑢2 ,
[ ]
[2 1 1 3 0 0 2 3] where
[ ]
1 0 0
( [ 4 6 2 4 0 0 0 0] )
[0 1 − 𝑏 0 ]
𝐸1 = [ 4 ],
2 −1 0
𝑛 + rank (𝐸𝐾 ) = 4 + rank ([ ]) = 6, [0 −𝑎4 −1]
0 0 2
1 0 0
[0 1 − 𝑎 1]
and we can conclude that closed-loop system (6) is impulse 𝐸2 = [ 5 ],
controllable under the above derivative feedback controller.
[0 −𝑏5 1]
Illustrative Examples 2 and 3 are provided for demon-
strating the different effects of proportional feedback and 1 0 0
derivative feedback. We can find that our method of deriva- [0 𝑏 𝑎 ]
tive feedback can change impulsive controllability when the 𝐴 11 =[ 4 3] ,
method of proportional feedback is unable to do it. [0 𝑎4 𝑎4 ]
Mathematical Problems in Engineering 11
1 0 0 𝑎3 = 2,
[ ]
𝐴 12 =[
[
0 0 0] ,
] 𝑎4 = 0.5,
[ 0 0 0]
𝑎5 = 1,
𝑐1 𝑐5 𝑐2
[ ] 𝑎6 = 0,
𝐴 21 = [ ]
[𝑎1 𝑎5 𝑎2 ] ,
𝑎7 = 0.6,
[ 𝑏1 𝑏5 𝑏2 ]
𝑎8 = 0.5,
0 0 0
[ ] 𝑏1 = 0.3,
𝐴 22 =[ ]
[0 0 0] ,
[0 𝑐4 0] 𝑏2 = 0.4,
0 𝑏4 = 1,
[ ]
𝐵1 = [ ]
[ 𝑏8 ] , 𝑏5 = 1,
[𝑎8 ] 𝑏6 = 0,
𝑏6 𝑏7 𝑏7 = 1,
[ ]
𝐵2 = [ ]
[𝑎6 𝑎7 ] , 𝑏8 = 0.5,
[ 𝑏6 𝑏6 ]
𝑐1 = 0.1,
𝛾
[ ] 𝑐2 = 0.4,
𝑥1 = [ ]
[𝛼] ,
𝑐4 = 0.3,
[𝛽]
𝑐5 = 0.4.
𝜔𝑥 (65)
[ ]
𝑥2 = [ ]
[𝜔𝑦 ] ,
[𝜔𝑧 ]
Note that
𝑢1 = 𝛿𝑧 ,
𝛿𝑥
𝑢2 = [ ] . { 𝐸𝐾 0 0 }
𝛿𝑦 𝑅1 = max {rank ([ ])
𝐾1 }
(64) { [ 𝐴 𝐸𝐾 𝐵2 ] }
System (63) is the large-scale interconnected descriptor 𝐸1 0 0 0 0 𝐵1 0
systems for the conditions that 1 − 𝑏4 = 0, 1 − 𝑎5 = 0, [ ]
and 𝐸1 and 𝐸2 are singular matrices. Furthermore, we find [ 0 𝐸 0 0 0 0 0]
[ 2 ]
rank([ 𝐴𝐸 𝐸0 ]) ≠ 𝑛 + rank(𝐸), so impulse exists in the system. = rank ([
[
])
]
[ 𝐴 1 𝐴 12 𝐸1 0 0 0 𝐵1 ]
For hypersonic vehicle, it may generate impulse response to [ ]
the vehicle with the vehicle flight with acceleration, decelera- ([𝐴 21 𝐴 22 0 𝐸2 𝐵2 0 0 ])
tion, and sharp turns, which may lead to the instability of the
system or may even destroy the vehicle, and hence it is not = 11,
expected to exist.
Consider system (63) with the following parameters [22, 𝑟1 = min {𝑛 + rank (𝐸𝐾 )}
23]: 𝐾1
𝑎1 = 0.5, 𝐸1 + 𝐵1 𝐾1 0
= min {𝑛 + rank ([ ])}
𝑎2 = 0.8, 𝐾1 0 𝐸2
12 Mathematical Problems in Engineering
𝐸1 0 𝐵1 0 𝑛 + rank (𝐸)
= 𝑛 + rank ([ ])
0 𝐸2 0 0
1 0 0 0 0 0
𝐵1 0 [ ]
[0 0 0 0 0 0]
− rank ([ ]) = 10. [ ]
0 0 ([ ])
([0 −0.5 −1 0 0 0])
= 6 + rank ( [
([0 0
]) = 11.
(66)
([ 0 1 0 0]
])
)
[ ]
[0 0 0 0 0 1]
𝑅1 > 𝑟1 satisfies the necessary condition of impulse [ ]
controllability. By Theorem 7, we can design a proportional
([0 0 0 0 −1 1])
feedback controller for the first subsystem as follows:
(68)
𝑢1 = 𝐾𝑑 𝑥, 𝐾𝑑 = [𝑎 𝑏 𝑐] . (67)
Obviously,
Then we have
𝐸 0 0
𝐸 0 0 rank ([ ]) ≠ 𝑛 + rank (𝐸) . (69)
rank ([ ]) 𝐴𝐾 𝐸 𝐵2
𝐴𝐾 𝐸 𝐵2
By the above derivation, for any proportional feedback gain
𝐸1 0 0 0 0 𝐾𝑑 , equality (26) in Lemma 4 can not hold. That means
[ ] there does not exist 𝐾𝑑 such that closed-loop system (6) is
[ 0 𝐸2 0 0 0]
= rank ([
[𝐴 + 𝐵 𝐾 𝐴 𝐸 0
]) impulse controllable. Further, by Theorem 7 we can design
[ 11 1 𝑑 12 1 0]
] a derivative feedback controller for the first subsystem as
𝐴 21 𝐴 22 0 𝐸2 𝐵2 ] follows:
[
−0.5 −1 𝑢1 = −𝐾1 𝑥1̇ , 𝐾1 = [𝑎 𝑏 𝑐] . (70)
= 9 + rank ([ ]) = 10,
1 2 Then we have
𝐸1 + 𝐵1 𝐾1 0 0 0 0
[ ]
𝐸𝐾 0 0 [ 0 𝐸2 0 0 0]
[ ]
rank ([ ]) = rank ([
[
])
]
[ 𝐴 11 𝐴 12 𝐸1 + 𝐵1 𝐾1 0 0 ]
[ 𝐴 𝐸𝐾 𝐵2 ] [ ]
([ 𝐴 21 𝐴 22 0 𝐸2 𝐵2 ])
𝐸𝐾 0 0
rank ([ ]) = 𝑛 + rank (𝐸𝐾 ) = 11. (73)
1 + 0.1𝑎 0.1𝑏 0.1𝑐 𝐴 𝐸𝐾 𝐵2
rank ([ ]) = 2, (72)
[ 0.2𝑎 −0.5 + 0.2𝑏 1 + 0.2𝑐] Then closed-loop system (6) is impulse controllable.
Mathematical Problems in Engineering 13
In the practical example, we can find that our method [7] M. A. Shayman, “Homogeneous indices, feedback invariants
of derivative feedback can change impulse controllability and control structure theorem for generalized linear systems,”
when the method of proportional feedback is unable to do SIAM Journal on Control and Optimization, vol. 26, no. 2, pp.
it. So, derivative feedback method is more effective than 387–400, 1988.
proportional feedback in the problem of impulse elimination. [8] J. J. Loiseau, “Pole placement and connected problems,” Kyber-
netika, vol. 28, no. 2, pp. 90–99, 1992.
[9] P. Zagalak, V. Kučera, and J. J. Loiseau, “Dynamics assignment
5. Conclusions by PD state feedback in linear reachable systems,” Kybernetika,
vol. 30, no. 2, pp. 153–158, 1994.
In this paper, the problem of impulse controllability and [10] J. J. Loiseau and P. Zagalak, “On pole structure assignment in
impulse observability of large-scale interconnected descrip- linear systems,” International Journal of Control, vol. 82, no. 7,
tor system by derivative feedback has been studied. Necessary pp. 1179–1192, 2009.
conditions for the existence of a derivative feedback con- [11] J. J. Loiseaua, K. Özçaldiran, M. Malabre, and N. Karcanias,
troller for the first subsystem of the large-scale interconnected “Feedback can onical forms of singular systems,” Kybernetika,
descriptor systems ensuring the second subsystem to be vol. 27, no. 4, pp. 289–305, 1991.
impulse controllable and impulse observable are derived, [12] J. Y. Ishihara and M. H. Terra, “Impulse controllability and
respectively. The examples have been presented to demon- observability of rectangular descriptor systems,” IEEE Transac-
strate the applicability of the proposed approach. Recently, tions on Automatic Control, vol. 46, no. 6, pp. 991–994, 2001.
to ensure the reliability and safety of modern large-scale [13] M. Hou, “Controllability and elimination of impulsive modes
industrial processes, data-driven methods have been receiv- in descriptor systems,” IEEE Transactions on Automatic Control,
ing considerably increasing attention. The research on the vol. 49, no. 10, pp. 1723–1727, 2004.
modeling and the impulsive controllability/observability of [14] G. R. Duan and A. G. Wu, “I-controllablizability in descriptor
interconnected descriptor systems based on large data and linear systems,” Dynamics of Continuous, Discrete and Impulsive
fault diagnosis of complex system will be the next research Systems, vol. 13, no. 1, pp. 1197–1204, 2006.
focus [31–33]. [15] G.-R. Duan, A.-G. Wu, Y. Zhao, and H.-H. Yu, “A revisit to I-
controllablisability for descriptor linear systems,” IET Control
Theory & Applications, vol. 1, no. 4, pp. 975–978, 2007.
Conflict of Interests [16] A.-G. Wu, G.-R. Duan, and S.-M. Zhao, “Impulsive-mode
controllablisability in descriptor linear systems,” IET Control
The authors declare that there is no conflict of interests Theory & Applications, vol. 1, no. 3, pp. 558–563, 2007.
regarding the publication of this paper. [17] A.-G. Wu, G.-R. Duan, and H.-H. Yu, “Impulsive-mode con-
trollablizability revisited for descriptor linear systems,” Asian
Acknowledgments Journal of Control, vol. 11, no. 3, pp. 358–365, 2009.
[18] D. L. Chu and D. W. C. Ho, “Necessary and sufficient conditions
This work is supported by National Natural Science Foun- for the output feedback regularization of descriptor systems,”
dation of China under Grant 61203001 and Fundamen- IEEE Transactions on Automatic Control, vol. 44, no. 2, pp. 405–
tal Research Funds for Central Universities under Grant 412, 1999.
N130405013. [19] D. H. Wang and C. B. Soh, “On regularizing singular systems
by decentralized output feedback,” IEEE Transactions on Auto-
matic Control, vol. 44, no. 1, pp. 148–152, 1999.
References [20] G. S. Zhang and W. Q. Liu, “Impulsive mode elimination
for descriptor systems by a structured P-D feedback,” IEEE
[1] Q. L. Zhang, C. Liu, and X. Zhang, Complexity, Analysis and Transactions on Automatic Control, vol. 56, no. 12, pp. 2968–
Control of Singular Biological Systems, vol. 421 of Lecture Notes 2973, 2011.
in Control and Information Sciences, Springer, Berlin, Germany, [21] M. Hou and P. C. Muller, “Causal observability of descriptor
2012. systems,” IEEE Transactions on Automatic Control, vol. 44, no.
[2] D. M. Yang, Q. L. Zhang, and B. Yao, Descriptor Systems, Science 1, pp. 158–163, 1999.
Press, Beijing, China, 2003. [22] Q. L. Zhang, Singular Large-Scale Systems Decentralized Control
[3] Z. G. Wu, H. Y. Su, and J. Chu, “Delay-dependent robust expo- and Robust Control, Northwest Industry University Press, Xian,
nential stability of uncertain singular systems with time delays,” China, 1997.
International Journal of Innovative Computing, Information and [23] D. D. Siljak, Large-Scale Dynamic Systems: Stability and Struc-
Control, vol. 6, no. 5, pp. 2275–2283, 2010. ture, North-Holland, Amsterdam, The Netherlands, 1978.
[4] Y.-F. Feng, X.-L. Zhu, and Q.-L. Zhang, “An improved 𝐻∞ [24] P. Lu, Y. Yang, Z. Li, and L. Huang, “Decentralized dynamic
stabilization condition for singular time-delay systems,” Inter- output feedback for globally asymptotic stabilization of a class
national Journal of Innovative Computing, Information and of dynamic networks,” International Journal of Control, vol. 81,
Control, vol. 6, no. 5, pp. 2025–2034, 2010. no. 7, pp. 1054–1061, 2008.
[5] Y. F. Teng, Q. L. Zhang, and G. S. Zhang, “Study of controlla- [25] W. J. Xiong, D. W. C. Ho, and J. D. Cao, “Synchronization
bility of descriptor systems,” Journal of Northeastern University, analysis of singular hybrid coupled networks,” Physics Letters A,
supplement, pp. 1–6, 2003. vol. 372, no. 44, pp. 6633–6637, 2008.
[6] D. Cobb, “Feedback and pole placement in descriptor variable [26] Y. J. Zhou, X. H. Yu, C. Y. Sun, and W. W. Yu, “Robust synchroni-
systems,” International Journal of Control, vol. 33, no. 6, pp. 1135– sation of second-order multi-agent system via pinning control,”
1146, 1981. IET Control Theory & Applications, vol. 9, no. 5, article 775, 2015.
14 Mathematical Problems in Engineering
International
Journal of Journal of
Mathematics and
Mathematical
Discrete Mathematics
Sciences