You are on page 1of 5

This article has been accepted for publication in a future issue of this journal, but has not been

fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TCSII.2021.3058803, IEEE
Transactions on Circuits and Systems II: Express Briefs

Admissible Bipartite Consensus in Networks of


Singular Agents Over Signed Graphs
Tianya Liu, Andong Sheng, Guoqing Qi and Yinya Li

Abstract—This brief focuses on the network of singular agents with exogenous uncertainties in the system, how to deal with consensus of
unknown disturbances. An admissible bipartite consensus problem is MASs with unknown disturbances attracts increased attention. Many
considered under cooperative-competitive interactions modeled by signed
references are concerned with consensus problem of MASs subjected
graphs. A distributed control algorithm derived by the extended state
observer is designed, based on which singular multi-agent systems achieve to unknown disturbances over cooperative-competitive interactions
not only disturbance rejection but also admissible bipartite consensus. such as [16]–[20]. By using state continuous or discontinuous feed-
It is shown that, by using singular system theory and matrix theory, back methods, authors in [16] and [17] investigate bipartite consensus
the designed controller is solvable for admissible bipartite consensus control for linear MASs with unknown disturbances over signed undi-
under the assumption of connection and structural balance. Finally, the
simulation example illustrates the feasibility of the proposed method. rected graphs, respectively. In [19] and [20], distributed stabilization
of MASs subjected to external disturbances is investigated under the
Index Terms—Singular multi-agent systems (SMASs), Unknown dis-
turbances, Admissible bipartite consensus, Signed graphs.
strong-weak competition network.
From the aforementioned literatures, one can find that the dynamics
of all the systems are normal, which are established by differential
I. I NTRODUCTION equations. In fact, compared to the normal system, the singular system
The distributed coordination of multiple systems has been a hot is more general because it consists of both the differential equation
topic in recent years due to the applications of multi-agent systems and the algebraic equation. The singular system has been typically
(MASs) such as sensor networks [1], unmanned aerial vehicles [2] applied in circuit systems, three-link planar manipulator networks,
and social networks [3]. The consensus control for MASs is one of the multi-agent supporting systems and so on (see, [21], [22]). Many
fundamental issues in distributed coordination, which has obtained researchers have studied consensus control for SMASs ( [23]–[27]).
the attention from more and more researchers. One can refer to [4], In [23], the authors investigate the admissible consensus problem
[5] and references therein for more details. It is clear that all these of SMASs by designing an observer-type distributed control law. In
works on consensus problem adopt to the distributed cooperative [24] and [25], output consensus of heterogeneous SMASs is inves-
control. Different from cooperative interactions among agents, a lot tigated by considering static or switched communication topology,
of situations like engineering systems and social networks should be respectively. Furthermore, under cooperative-competitive interactions,
described by cooperative-competitive interactions. As shown in [6], authors in [26] investigate bipartite consensus of SMASs over a
a signed graph denotes the cooperative-competitive network, where signed graph. Although there exist some fruitful works on consensus
positive/negative edge denotes cooperative/competitive interaction analysis of SMASs, consensus control for SMASs subjected to
between agents, respectively. Under the condition that the signed unknown disturbances is still rare and worthwhile to research.
graph satisfies structurally balanced and is connected or strongly In this brief, we investigate the consensus problem of SMASs in the
connected, bipartite consensus for integrator MASs is achieved in presence of disturbances under cooperative-competitive interactions,
[6]. The so-called bipartite consensus is that two groups of agents where the unknown but bounded disturbances are considered, and
have common values in modulus and opposite in sign. More results satisfy constant steady-state values. We adopt the extended state
for MASs under cooperative-competitive interactions can refer to [7]– observer (ESO) method in designing consensus control law to reject
[11] and references therein. disturbances and reach bipartite consensus. Similar ESO method
In addition, it is clear that the influence of external disturbances emerges in [18], but for normal systems. With these observations,
or uncertainties is inevitable such that both the stability analysis of the contributions of this brief are summarized as follows: (i) Bi-
the system and the design of controller will become more intractable. partite consensus control for SMASs with unknown disturbances is
The appropriate control algorithm is necessary for addressing these investigated. It is noteworthy that, different from the normal system
problems such as neural network-based control [12], disturbance [9]–[11], [16]–[20], the algebraic equation is existing in singular
observer-based control [13], fuzzy adaptive control [14], [15], etc. system. This leads to that not only stability but also the impulse-free
The consensus problem of multiple agents, where each agent’s property needs to be considered. Moreover, the unknown disturbances
dynamics is affected by disturbances, turns out to be equally worth are introduced into this work, which also bring great challenges in
considering. There are two types of consensusability systems on the analyzing the closed-loop system. In view of this, we design appro-
basis of different external disturbances. On the one hand, external priate observer and control gains to reject disturbance, and expand
disturbances are generated by known exogenous system; on the other the convergent property of singular system by Lemma 3.1. (ii) The
hand, the disturbances are unknown but bounded, which can not cooperative-competitive interaction is considered in SMASs, where
be generated by existing exosystem. Due to the wide existence of cooperative-competitive interaction is modeled by a signed graph with
positive/negative edge between agents being cooperative/competitive
This work was supported in part by the National Natural Science Foundation
of China under Grant 61973183 and Grant 61871221; in part by the Natural interaction. Since this brief focuses on signed graphs, the Laplacian
Science Foundation of Shandong Province under Grant ZR2019MF041; in of topology graph does not satisfy zero row sum and the error
part by the Youth Creative Team Sci-Tech Program of Shandong Universities between agents should be constructed by the form of difference and
under Grant 2019KJI007; and in part by the National Defense Basic Research sum. By assuming structurally balanced signed graph and using the
Project of China under Grant JCKY2018209B010.
The authors are with the School of Automation, Nanjing University of Sci-
transformation matrix, the asymptotic stability of the error system can
ence and Technology, Nanjing 210094, China (e-mail: liutianya00@163.com; be achieved. (iii) Sufficient and necessary conditions on admissible
njustsheng@163.com; qiguoqing@njust.edu.cn; liyinya@njust.edu.cn). bipartite consensus of SMASs with unknown disturbances are estab-

1549-7747 (c) 2021 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
Authorized licensed use limited to: ULAKBIM UASL - KOCAELI UNIVERSITESI. Downloaded on April 26,2021 at 08:48:26 UTC from IEEE Xplore. Restrictions apply.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TCSII.2021.3058803, IEEE
Transactions on Circuits and Systems II: Express Briefs

lished by combining singular system theory with matrix analysis and where xi ∈ Rn , ui , wi ∈ Rp and yi ∈ Rq are the state, control
graph theory. Moreover, parameters of the proposed control algorithm input, external disturbance and output of the ith agent, respectively.
can be designed by using linear quadratic regulator (LQR) based And E, A ∈ Rn×n , Bc ∈ Rn×p , C ∈ Rq×n .
optimal control and observer design technique. Problem 2.1: Our aim is to solve the following problem that
The remainder of this brief is organized as follows. Section II all the agents of (2) can reach admissible bipartite consensus by
provides preliminaries and the system model to be studied. Sec- designing a distributed controller ui (t). That is, the closed-loop
tion III shows admissible bipartite consensus obtained by observer- system is impulse-free and satisfies
based distributed controllers. Section IV verifies the theoretical result 
via simulation example and the concluding remark is presented in  lim kxi (t) − xj (t)k = 0, ∀i, j ∈ Vf1 or Vf2 ;
t→∞
Section V. (3)
 lim kxi (t) + xj (t)k = 0, ∀i ∈ Vf1 and j ∈ Vf2 .
Notations. R is real number set, C is complex number set. t→∞

Rm×n is m × n dimensional real matrices. Rm is m dimensional


real vectors. I denotes identity matrix with compatible dimensions.
1 = [1, 1, · · · , 1]T with compatible dimensions. σ(E, X) = {s ∈ III. A DMISSIBLE BIPARTITE CONSENSUS
C| det(sE − X) = 0}. C− denotes the open left half plane. A. Design of Distributed Controllers
A > 0(A ≥ 0) denotes positive definite (semi-definite) matrix. ⊗ First, an ESO is designed. To achieve this, one has the following
is the Kronecker product. sgn(·) is a sign function. extended form:
Ē v̇i (t) = Āvi (t) + B̄ui (t) + F̄ ϕi (t)
II. P RELIMINARIES AND P ROBLEM F ORMULATION
yi (t) = C̄vi (t), i = 1, · · · , N, (4)
A. Preliminaries    
xi (t) E 0
A signed graph is denoted by G = (Vf , E, A) with Vf = where vi (t) = , ϕi (t) = ẇi (t) and Ē = , Ā =
w
 i (t) 0 I
{1, · · · , N }, E ⊆ Vf × Vf and A = (aij ) ∈ RN ×N being node set,    
A Bc Bc 0  
edge set and weighted adjacency matrix, where aij 6= 0 ⇔ (j, i) ∈ E, , B̄ = , F̄ = , C̄ = C 0 . Then combining
0 0 0 I
otherwise, aij = 0. (j, i) ∈ E means that node i can receive with (4), an ESO is proposed as follows:
information from node j. If aij > 0, then i and j are cooperative; if
aij < 0, then they are competitive. Ni = {j ∈ V|(j, i) ∈ E} denotes Ē ξ˙i (t) = Āξi (t) + B̄ui (t) + F (yi (t) − C̄ξi (t)), (5)
the neighborPset of i. The Laplacian of G is L = H − A with  
ξxi (t)
H = diag( N
PN
j=1 |a1j |, · · · , j=1 |aN j |). G is connected if there where ξi (t) =
ξ
is the state of (5), and represents the
wi (t)
exists a path for any two nodes in G. 
xi (t)
A signed graph G = (Vf , E, A) is structurally balanced if it admits estimation of .
wi (t)
a partition Vf1 , Vf2 with Vf1 ∪ Vf2 = Vf , Vf1 ∩ Vf2 = ∅ such that Next, for the system in (2), the following distributed control
aij ≥ 0, ∀i, j ∈ Vfq (q ∈ {1, 2}), and aij ≤ 0, ∀i ∈ Vfq , j ∈ Vfm , q 6= protocol is designed:
m(q, m ∈ {1, 2}). It is said structurally unbalanced otherwise.
N
Lemma 2.1: [6] If G = (Vf , E, A) is connected and structurally
X 
ui (t) = µHc aij ξxj (t) − sgn(aij )ξxi (t) − ξwi (t), (6)
balanced, then L has a simple zero eigenvalue, and all the nonzero j=1
eigenvalues are positive.
where µ ∈ R is the scalar, Hc ∈ Rp×n is the state feedback matrix. A
For the system block diagram of control strategy for single agent is taken in Figure 1.
E ż(t) = Az(t) + Dβ(t) (1) For ensuring the solvability of the studied problem, the following
assumptions are listed.
with E, A ∈ Rn×n , 0 < rankE < n. We call that (E, A) is regular Assumption 1: The signed graph G = (Vf , E, A) is connected
if det(s0 E − A) 6= 0, ∃s0 ∈ C. (E, A) is impulse-free if (E, A) is and structurally balanced. 
regular and deg(det(sE − A)) = rankE, ∀s ∈ C. (E, A) is stable if E 0
Assumption 2: rank = n + rankE.
σ(E, A) ⊆ C− . (E, A) is admissible if (E, A) is impulse-free and A E
stable. Assumption 3: The unknown disturbances wi (t) satisfy: (1)
wi (t) and ẇi (t) are bounded; (2) lim ẇi (t) = 0.
Lemma 2.2: [28] Assume that (E, A) is regular. If there exist t→∞
W = W T ≥ 0 and V = V T > 0 satisfying E T W A + AT W E = Assumption 4: (E, A, Bc ) is R-controllable and (Ē, Ā, C̄) is
−E T V E, then (E, A) is admissible. R-observable.
Remark 3.1: It is clear that Assumption 1 guarantees the con-
Lemma 2.3: [28] Assume that (E, A) is regular and impulse- nectivity of agents. Assumption 2 is taken to guarantee the regularity
free, (E, A, D) is R-controllable. Then for any Ξ, R > 0, there exists and impulse-free property of the studied system and the existence and
Υ > 0 satisfying E T ΥA+AT ΥE +E T ΞE −E T ΥDR−1 DT ΥE = uniqueness of solutions. In Assumption 3, the external disturbances
0. are assumed as unknown but bounded, and have constant steady-
state values. Moreover, Assumption 4 is established for ensuring the
existence of distributed control laws ui (t) and the ESO.
B. Problem Formulation
An SMAS composed of N agents is considered as below, where
each agent’s dynamics is modeled by B. Main Results
By using the distributed controller in (6), this section provides
E ẋi (t) = Axi (t) + Bc [ui (t) + wi (t)] main results on ensuring admissible bipartite consensus for SMASs
yi (t) = Cxi (t), i = 1, · · · , N (2) subjected to unknown disturbances.

1549-7747 (c) 2021 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
Authorized licensed use limited to: ULAKBIM UASL - KOCAELI UNIVERSITESI. Downloaded on April 26,2021 at 08:48:26 UTC from IEEE Xplore. Restrictions apply.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TCSII.2021.3058803, IEEE
Transactions on Circuits and Systems II: Express Briefs

wi
Thus, combining (10) with (12), we have
ui yi
Hc Plant   
IN ⊗ Ē 0 δ̇(t)
0 IN ⊗ E ẋ(t)
  
IN ⊗ (Ā − F C̄) 0 δ(t)
=
ESO ∗ IN ⊗ A − µL ⊗ Bc Hc x(t)
 
ξwi −IN ⊗ F̄
+ ϕ(t). (13)
ξxi 0

Figure 1. Block diagram of control strategy. In addition, Vf can be divided into two parts under structurally bal-
anced property. For convenience, we assume that Vf1 = {1, · · · , m0 }
and Vf2 = {m0 + 1, · · · , N }. Motivated by Problem 2.1, the error
Lemma 3.1: Assume that the system (1) is regular and impulse- between agents is defined as follows:
free. If β(t) is bounded and satisfies lim β(t) = 0 and (E, A) is
t→∞ ei (t) = xi (t) − x1 (t), i = 2, · · · , m0 ,
stable, then lim z(t) = 0.
t→∞ ei (t) = xi (t) + x1 (t), i = m0 + 1, · · · , N.
Proof: By means of the regularity of (E, A), there exist
nonsingular matrices Φ, Ψ satisfying Ê = ΦEΨ =   n1 , N ),
diag(I Obviously, equation (3) is solved if ei (t) → 0, i = 2, · · · , N as
D1 t → 0. To prove the error
 = ΦAΨ = diag(A1 , In2 ) and D̂ = ΦD = , where
  D2
z1 (t) lim ei (t) = 0, i = 2, · · · , N, (14)
n1 + n2 = n. Let z(t) = Ψ , then (1) is equivalent to slow t→∞
z2 (t)
subsystem and fast subsystem a nonsingular matrix Q as presented in [29] is considered as below.
ż1 (t) = A1 z1 (t) + D1 β(t), (7)  1
q1 q2

N
N ż2 (t) = z2 (t) + D2 β(t). (8) Q =  −1m0 −1 Im0 −1 0 
−1N −m0 0 −IN −m0
And (E, A) is stable ⇒ (Ê, Â) is stable ⇒ A1 is a stable matrix.
According to Lemma 2 of [13], one can obtain that the system with q1 = [ N1 , · · · , 1
N
] and q2 = [− N1 , · · · , − N1 ]. Denote
(7) is asymptotically stable. i.e. lim z1 (t) = 0. Moreover, there is
t→∞ ê(t) = (Q ⊗ In )x(t). (15)
N = 0 in (8) because the system (1) is impulse-free. It is clear that
lim z2 (t) = 0 as lim β(t) = 0. Therefore, one has lim z(t) = 0 Then,
t→∞ t→∞ t→∞
from the above analysis.
ê(t) = [ρT (t), eT2 (t), · · · , eTN (t)]T , [ρT (t), eT (t)]T . (16)
Theorem 3.1: Under Assumption 1-4, SMAS in (2) via the
controller in (6) can solve Problem 2.1 iff there exist matrices Hc and Further, combining (12) with (15), one can obtain that
F such that (E, A − µλi Bc Hc ) and (Ē, Ā − F C̄) are admissible,
where λi , i = 2, · · · , N are nonzero eigenvalues of L. ˙ = (IN ⊗ A − µQLQ−1 ⊗ Bc Hc )ê(t)
(IN ⊗ E)ê(t)
 
δxi (t) − (µQL ⊗ Bc Hc )δx (t) − (Q ⊗ Bc )δw (t). (17)
Proof: Choose δi (t) = ξi (t) − vi (t) = with
δwi (t)  
0 0
δxi (t) = ξxi (t) − xi (t), δwi (t) = ξwi (t) − wi (t) (9) Clearly, QLQ−1 = . According to Lemma 2.1, we can find
0 L∆
being the estimation error for the ith agent. Then, by (4) and (5), one that the eigenvalues of L satisfy 0 = λ1 < λ2 ≤ · · · ≤ λN under
has the following error system Assumption 1. This implies that λ2 , · · · , λN are also the eigenvalues
of matrix L∆ , and they are nonzero. By considering both (10) and
Ē δ̇i (t) = Āδi (t) − F C̄δi (t) − F̄ ϕi (t). (17), one can obtain
Furthermore, for all agents, a compact from can be written as follow: 
IN ⊗ Ē 0

δ̇(t)


0 IN ⊗ E ê(t) ˙
(IN ⊗ Ē)δ̇(t) = [IN ⊗ (Ā − F C̄)]δ(t) − (IN ⊗ F̄ )ϕ(t), (10)   
IN ⊗ (Ā − F C̄) 0 δ(t)
where δ(t) = [δ1T (t), · · · , δN
T
(t)]T , ϕ(t) = [ϕT1 (t), · · · , ϕTN (t)]T . = −1
∗ IN ⊗ A − µQLQ ⊗ Bc Hc ê(t)
Next, combining (6) with (2), we get  
−IN ⊗ F̄
N + ϕ(t). (18)
X  0
E ẋi (t) = Axi (t) + µBc Hc aij ξxj (t) − sgn(aij )ξxi (t)
j=1 Specially, from (16) and (18), we have
− Bc ξwi (t) + Bc wi (t). (11)   
IN ⊗ Ē 0 δ̇(t)
Define 0 IN −1 ⊗ E ė(t)
  
x(t) = [xT1 (t), · · · , xTN (t)]T , ξx (t) = [ξxT1 (t), · · · , ξxTN (t)]T , IN ⊗ (Ā − F C̄) 0 δ(t)
=
T T ∗ IN −1 ⊗ A − µL∆ ⊗ Bc Hc e(t)
δw (t) = [δw 1
(t), · · · , δw N
(t)]T , δx (t) = [δxT1 (t), · · · , δxTN (t)]T ,  
−IN ⊗ F̄
where δwi (t) and δxi (t) are presented in (9). From (11), one has + ϕ(t). (19)
0
(IN ⊗ E)ẋ(t) = (IN ⊗ A − µL ⊗ Bc Hc )x(t) − (µL ⊗ Bc Hc )
Obviously, there exists a nonsingular matrix P ∈ R(N −1)×(N −1)
δx (t) − (IN ⊗ Bc )δw (t). (12) satisfying P −1 L∆ P = diag(J1 , · · · , Jr ) , J∆ , where Jk , k =

1549-7747 (c) 2021 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
Authorized licensed use limited to: ULAKBIM UASL - KOCAELI UNIVERSITESI. Downloaded on April 26,2021 at 08:48:26 UTC from IEEE Xplore. Restrictions apply.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TCSII.2021.3058803, IEEE
Transactions on Circuits and Systems II: Express Briefs

1, · · · , r are upper triangular matrices and their diagonal el- Theorem 3.2: Under Assumption 1-4, SMAS in (2) via (6) can
ements
 are λi , i =  2, · · · , N . By pre-multiplying matrix solve Problem 2.1 if choose
IN ⊗ In+p 0
in (19), one has (1) µ ≥ 2λ12 , Hc = R−1 BcT ΥE with Υ satisfying E T ΥA +
0 P −1 ⊗ In AT ΥE + E T ΞE − E T ΥBc R−1 BcT ΥE = 0;
Ẽc (t) = Ãc (t) + B̃c ϕ(t) (20) (2) F = Ē ῩC̄ T R−1 /2 with Ῡ satisfying Ē ῩĀT + ĀῩĒ T +
ĒΞĒ T − Ē ῩC̄ T R−1 C̄ ῩĒ T = 0, ∀Ξ, R > 0.
with Proof: From Theorem 3.1, it is clear that admissible bipartite
(t) = [δ T (t), ẽT (t)]T , ẽ(t) = (P −1 ⊗ In )e(t), consensus of SMAS in (2) via (6) is equivalent to the admissibility

IN ⊗ Ē 0
 
−IN ⊗ F̄
 of both (E, A − µλi Bc Hc ) and (Ē, Ā − F C̄). Choose µ ≥ 2λ12 and
Ẽc = , B̃c = , Hc = R−1 BcT ΥE. According to Lemma 2.3, the following equation
0 IN −1 ⊗ E 0
  holds:
IN ⊗ (Ā − F C̄) 0
Ãc = .
∗ IN −1 ⊗ A − µJ∆ ⊗ Bc Hc E T Υ(A − µλi Bc Hc ) + (A − µλi Bc Hc )T ΥE = −E T V E
From the above analysis, we divide Problem 2.1 into two steps. where V = Ξ + (2µλi − 1)ΥBc R−1 BcT Υ > 0. Then, from
Step 1. We prove that the system in (13) is impulse-free, i.e., Lemma 2.2, we can obtain that (E, A − µλi Bc Hc ) is admissible.
 I ⊗ Ē Moreover, choose F = Ē ῩC̄ T R−1 /2. Then

N 0
,
0 IN ⊗ E
  (Ā − F C̄)ῩĒ T + Ē Ῡ(Ā − F C̄)T = −ĒΞĒ T ,
IN ⊗ (Ā − F C̄) 0
, (Em , Am ) which guarantees the admissibility of the pair (Ē, Ā − F C̄). Thus,
∗ IN ⊗ A − µL ⊗ Bc Hc
−1
Problem 2.1 is solved if conditions (1) (2) hold.
is impulse-free. Since (Q ⊗ In )(IN ⊗ E)(Q ⊗ In ) = (IN ⊗ E),
0 0
 Remark 3.3: It is noteworthy that, different from conventional
−1
(Q⊗In )(IN ⊗A−µL⊗Bc Hc )(Q⊗In ) = IN ⊗A−µ ⊗ consensus under the cooperative network, the assumption of structural
0 L∆
Bc Hc , by calculation, balance is necessary for the bipartite consensus problem, one can
refer to [7] and [30] for more details. Furthermore, we can find that
det(sEm − Am ) = [det(sĒ − (Ā − F C̄))]N det(sE − A) the cooperative network can naturally be structurally balanced, all the
 
× det s(IN −1 ⊗ E) − (IN −1 ⊗ A − µL∆ ⊗ Bc Hc ) . agents under cooperative interactions achieve lim kxi (t)−xj (t)k =
t→∞
0. Therefore, the conventional consensus can be regarded as the
This, together with (P ⊗ In )−1 (IN −1 ⊗ E)(P ⊗ In ) = (IN −1 ⊗ E) special case of bipartite consensus, the result of this brief can equally
and (P ⊗ In )−1 (IN −1 ⊗ A − µL∆ ⊗ Bc Hc )(P ⊗ In ) = IN −1 ⊗ deal with the admissible consensus problem of SMASs.
A − µJ∆ ⊗ Bc Hc , yields that
det(sEm − Am ) = [det(sĒ − (Ā − F C̄))]N det(sE − A) IV. S IMULATION
× det(sE − (A − µλ2 Bc Hc )) · · · det(sE − (A − µλN Bc Hc )).
Therefore, one can obtain that the impulse-free property of (Em , Am ) 1
is equivalent to that the pairs (Ē, Ā − F C̄), (E, A) and (E, A − 2
-1

µλi Bc Hc ), i = 2, · · · , N are impulse-free. Under Assumption 2, the


pair (E, A) is impulse-free, therefore the system in (13) is impulse- 2 5
-2

free iff (Ē, Ā − F C̄) and (E, A − µλi Bc Hc ), i = 2, · · · , N are 1


-1
4

impulse-free. -3
3 4
Step 2. We prove that (14) holds. One can see that, from (16),
lim e(t) = 0. Obviously, it is turned into the stability of the system
t→∞ Figure 2. Communication topology G = {Vf , E, A}.
in (20). Under Assumption 3, it is clear that lim ϕ(t) = 0. From
t→∞
Step 1, one knows that (Ẽc , Ãc ) is impulse-free when (Ē, Ā − F C̄) Example 4.1: Consider an SMAS of five agents, where each agent
and (E, A − µλi Bc Hc ) are impulse-free. Therefore, by Lemma 3.1, is represented by the system (2), and the matrices satisfy
the stability of the system (20) is satisfied iff (Ẽc , Ãc ) is stable. And    
hence, (14) holds iff (Ē, Ā − F C̄) and (E, A − µλi Bc Hc ) are stable 1 0 0 2 1 0
with λi , i = 2, · · · , N being nonzero eigenvalues of L. Combining E = 0 1 0 , A = −5 −2 0
with Step 1, admissible bipartite consensus of SMASs with unknown 0 0 0 2 0 −1
T
disturbances is achieved iff (Ē, Ā−F C̄) and (E, A−µλi Bc Hc ), i =
  
Bc = 2 0 1 , C = 1 0 0
2, · · · , N are admissible.
and unknown disturbances are as follows: w1 (t) = −3 + exp(−2t),
w2 (t) = 2 + 2 exp(−t), w3 (t) = 1 − 2 exp(−t), w4 (t) =
Remark 3.2: Bipartite consensus control for SMASs with un-
−1 + exp(−3t), w5 (t) = −2 + 3 exp(−2t). It is clear that
known disturbances is investigated in this work due to the inevitability
Assumption 2-4 hold. Next, the communication topology is modeled
of external disturbances. Compared to [26], the influence of distur-
by Figure 2 with blue solid lines being cooperative interaction
bances is considered. To reject disturbances, a distributed controller
and red dotted lines being competitive interaction of two nodes,
via the ESO is proposed. Moreover, in absence of disturbances, the
which satisfies Assumption 1. By calculation, the eigenvalues of
proposed method also work. It is different from the state feedback
L in Figure 2 are λ1 = 0, λ2 = 2.7595, λ3 = 5.5828,
emerged in [26].
λ4 = 7.0000, λ5 = 12.6577, respectively. Thus, the  gains
Furthermore, from Theorem 3.1, the controller and observer design are chosen as µ = 12 , Hc = 2.4307 −0.0094 0 , F =
procedures are shown by LQR technique.  T
1.6666 −1.8056 0 0.5000 such that the pairs (E, A −

1549-7747 (c) 2021 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
Authorized licensed use limited to: ULAKBIM UASL - KOCAELI UNIVERSITESI. Downloaded on April 26,2021 at 08:48:26 UTC from IEEE Xplore. Restrictions apply.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TCSII.2021.3058803, IEEE
Transactions on Circuits and Systems II: Express Briefs

µλi Bc Hc ) and (Ē, Ā − F C̄) are admissible. Take the initial con- [7] C.-Q. Ma and L. Xie, “Necessary and sufficient conditions for leader-
ditions as: Ex1 (0) = [−0.4, 0, 0]T , Ex2 (0) = [−0.1, −0.2, 0]T , following bipartite consensus with measurement noise,” IEEE Trans.
Syst., Man, Cybern., Syst., vol. 50, no. 5, pp. 1976–1981, 2020.
Ex3 (0) = [0.05, −0.12, 0]T , Ex4 (0) = [0.3, −0.3, 0]T , Ex5 (0) =
[8] J. Li, X. Chen, F. Hao, and J. Xie, “Event-triggered bipartite consensus
[−0.1, −0.25, 0]T , Ēξ1 (0) = [−0.2, 0.1, 0, 0.15]T , Ēξ2 (0) = for multi-agent systems with antagonistic interactions,” Int. J. Control
[−0.3, 0, 0, −0.15]T , Ēξ3 (0) = [0.3, −0.05, 0, 2]T , Ēξ4 (0) = Autom. Syst., vol. 17, no. 8, pp. 2046–2058, 2019.
[−0.15, 0.1, 0, −0.5]T , Ēξ5 (0) = [−0.25, −0.4, 0, 0]T . Figure 3 (a)- [9] G. Wen, H. Wang, X. Yu, and W. Yu, “Bipartite tracking consensus of
(c) show state trajectories of the five agents subjected to unknown linear multi-agent systems with a dynamic leader,” IEEE Trans. Circuits
Syst. II, Exp. Briefs, vol. 65, no. 9, pp. 1204–1208, 2018.
disturbances. One can find that the trajectories of all the agents [10] H. Zhang and J. Chen, “Bipartite consensus of multi-agent systems over
are parted into two groups as time t increases, i.e. agent 1,2,3 signed graphs: State feedback and output feedback control approaches,”
put together a group, agent 4,5 make the other group, and they Int. J. Robust Nonlin. Control, vol. 27, no. 1, pp. 3–14, 2017.
have common values in modulus and opposite in sign. Obviously, [11] J. Hu, Y. Wu, T. Li, and B. K. Ghosh, “Consensus control of general
linear multiagent systems with antagonistic interactions and communica-
admissible bipartite consensus is reached under the given controller. tion noises,” IEEE Trans. Autom. Control, vol. 64, no. 5, pp. 2122–2127,
Moreover, Figure 3 (d) shows estimation values of the disturbances, 2019.
they converge to -3,2,1,-1,-2 as time t increases, respectively, which [12] K. Sun, J. Qiu, H. R. Karimi, and H. Gao, “A novel finite-time control
are consistent with wi (t), i = 1, · · · , 5. for nonstrict feedback saturated nonlinear systems with tracking error
constraint,” IEEE Trans. Syst., Man, Cybern., Syst., to be published,
doi: 10.1109/TSMC.2019.2958072.
8
x
11 20 x
12
[13] S. Li, J. Yang, W.-H. Chen, and X. Chen, “Generalized extended state
6

4
x21
x
31
15
x22
x
32
observer based control for systems with mismatched uncertainties,” IEEE
x x
x51
41
10 42
x52 Trans. Ind. Electron., vol. 59, no. 12, pp. 4792–4802, 2012.
xi1 (t)

xi2 (t)

2
5
0
0
[14] K. Sun, J. Qiu, H. R. Karimi, and Y. Fu, “Event-triggered robust
-2

-4
-5 fuzzy adaptive finite-time control of nonlinear systems with pre-
0 5 10
(a)
15 20
-10
0 5 10
(b)
15 20 scribed performance,” IEEE Trans. Fuzzy Syst., to be published, doi:
10.1109/TFUZZ.2020.2979129.
8
15 x
x23
13

6
Agent 1
Agent 2
[15] K. Sun, L. Liu, J. Qiu, and G. Feng, “Fuzzy adaptive finite-time fault-
10 x
x43
33
4
Agent 3
Agent 4
Agent 5
tolerant control for strict-feedback nonlinear systems,” IEEE Trans.
Estimate

5 x53
xi3 (t)

0
2
Fuzzy Syst., to be published, doi: 10.1109/TFUZZ.2020.2965890.
-5
0

-2
[16] Y. Wu, Y. Zhao, and J. Hu, “Bipartite consensus control of high-order
-10 -4
multiagent systems with unknown disturbances,” IEEE Trans. Syst., Man,
0 5 10 15 20 0 5 10 15
(c) (d) Cybern., Syst., vol. 49, no. 10, pp. 2189–2199, 2019.
[17] M. Liu, X. Wang, and Z. Li, “Robust bipartite consensus and tracking
Figure 3. (a)-(c) State trajectories xi , i = 1 · · · 5. (d) Estimation values control of high-order multiagent systems with matching uncertainties
ξwi (t), i = 1 · · · 5. and antagonistic interactions,” IEEE Trans. Syst., Man, Cybern., Syst.,
vol. 50, no. 7, pp. 2541–2550, 2020.
[18] S. Bhowmick and S. Panja, “Leader-follower bipartite consensus of
linear multiagent systems over a signed directed graph,” IEEE Trans.
Circuits Syst. II, Exp. Briefs, vol. 66, no. 8, pp. 1436–1440, 2019.
V. C ONCLUSION [19] H.-X. Hu, G. Wen, W. Yu, T. Huang, and J. Cao, “Distributed stabiliza-
The admissible bipartite consensus problem of SMASs with ex- tion of multiple heterogeneous agents in the strong-weak competition
ternal disturbances under cooperative-competitive interactions has network: A switched system approach,” IEEE Trans. Cybern., to be
published, doi: 10.1109/TCYB.2020.2995154.
been investigated in this brief. The ESO is introduced into the [20] H.-X. Hu, G. Wen, X. Yu, Z.-G. Wu, and T. Huang, “Distributed stabi-
studied system and a distributed control protocol based on ESO is lization of heterogeneous MASs in uncertain strong-weak competition
designed to reject disturbance. Furthermore, singular system theory, networks,” IEEE Trans. Syst., Man, Cybern., Syst., to be published, doi:
matrix analysis and graph theory are adopt for achieving admissible 10.1109/TSMC.2020.3034765.
[21] X.-R. Yang and G.-P. Liu, “Necessary and sufficient consensus condi-
bipartite consensus of SMASs. Sufficient and necessary conditions tions of descriptor multi-agent systems,” IEEE Trans. Circuits Syst. I,
are established by the admissibility of low dimension matrices, and Reg. Papers, vol. 59, no. 11, pp. 2669–2677, 2012.
the designed parameters are obtained by optimal LQR technique. [22] J. Xi, Z. Shi, and Y. Zhong, “Admissible consensus and consensualiza-
This brief focuses on homogeneous agents, heterogeneous SMASs tion of high-order linear time-invariant singular swarm systems,” Physica
A, vol. 391, no. 23, pp. 5839–5849, 2012.
subjected to unknown disturbances is considerable in further research. [23] L. Gao, Y. Cui, and W. Chen, “Admissible consensus for descrip-
tor multi-agent systems via distributed observer-based protocols,” J.
R EFERENCES Franklin Inst., vol. 354, no. 1, pp. 257–276, 2017.
[24] Q. Ma, S. Xu, F. L. Lewis, B. Zhang, and Y. Zou, “Cooperative output
[1] M. Zheng, C.-L. Liu, and F. Liu, “Average-consensus tracking of sensor regulation of singular heterogeneous multiagent systems,” IEEE Trans.
network via distributed coordination control of heterogeneous multi- Cybern., vol. 46, no. 6, pp. 1471–1475, 2016.
agent systems,” IEEE Control Syst. Lett., vol. 3, no. 1, pp. 132–137, [25] S. Wang and J. Huang, “Cooperative output regulation of singular multi-
2019. agent systems under switching network by standard reduction,” IEEE
[2] X. Dong, Y. Zhou, Z. Ren, and Y. Zhong, “Time-varying formation Trans. Circuits Syst. I, Reg. Papers, vol. 65, no. 4, pp. 1377–1385, 2018.
tracking for second-order multi-agent systems subjected to switching [26] L. Zhang and G. Zhang, “Bipartite consensus for descriptor multiagent
topologies with application to quadrotor formation flying,” IEEE Trans. systems with antagonistic interactions,” IEEE Trans. Circuits Syst. II,
Ind. Electron., vol. 64, no. 6, pp. 5014–5024, 2017. Exp. Briefs, vol. 67, no. 11, pp. 2602–2606, 2020.
[3] A. V. Proskurnikov, A. S. Matveev, and M. Cao, “Opinion dynamics in [27] X. Zhang, X. Liu, and Z. Feng, “Distributed containment control of
social networks with hostile camps: Consensus vs. polarization,” IEEE singular heterogeneous multi-agent systems,” J. Franklin Inst., vol. 357,
Trans. Autom. Control, vol. 61, no. 6, pp. 1524–1536, 2016. pp. 1378–1399, 2020.
[4] Z. Feng, G. Hu, and G. Wen, “Distributed consensus tracking for multi- [28] G. R. Duan, Analysis and design of descriptor linear systems. New
agent systems under two types of attacks,” Int. J. Robust Nonlin. Control, York: Springer, 2010.
vol. 26, pp. 896–918, 2016. [29] M. E. Valcher and P. Misra, “On the consensus and bipartite consensus
[5] Y. Lv, J. Fu, G. Wen, T. Huang, and X. Yu, “Distributed adaptive in high-order multi-agent dynamical systems with antagonistic interac-
observer-based control for output consensus of heterogeneous MASs tions,” Syst. Control Lett., vol. 66, no. 4, pp. 94–103, 2014.
with input saturation constraint,” IEEE Trans. Circuits Syst. I, Reg. [30] H.-X. Hu, G. Wen, G. Chen, X. Yu, and T. Huang, “Structural balance
Papers, vol. 67, no. 3, pp. 995–1007, 2020. preserving and bipartite static consensus of heterogeneous agents in
[6] C. Altafini, “Consensus problems on networks with antagonistic interac- cooperation-competition networks,” IEEE Trans. Netw. Sci. Eng., vol. 7,
tions,” IEEE Trans. Autom. Control, vol. 58, no. 4, pp. 935–946, 2013. no. 4, pp. 3223–3234, 2020.

1549-7747 (c) 2021 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
Authorized licensed use limited to: ULAKBIM UASL - KOCAELI UNIVERSITESI. Downloaded on April 26,2021 at 08:48:26 UTC from IEEE Xplore. Restrictions apply.

You might also like