Distributed_Fault-Tolerant_Time-Varying_Formation_Control_for_Second-Order_Multi-Agent_Systems_With_Actuator_Failures_and_Directed_Topologies

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 5

774 IEEE TRANSACTIONS ON CIRCUITS AND SYSTEMS—II: EXPRESS BRIEFS, VOL. 65, NO.

6, JUNE 2018

Distributed Fault-Tolerant Time-Varying Formation


Control for Second-Order Multi-Agent Systems
With Actuator Failures and Directed Topologies
Yongzhao Hua, Xiwang Dong, Qingdong Li, and Zhang Ren

Abstract—This brief investigates the distributed fault-tolerant practical applications, the control input is force or moment,
time-varying formation control problems for second-order then the dynamics of each agent should be described by
multi-agent systems with directed interaction topologies in the second-order differential equations. Ren [15] extended a con-
presence of both bias and loss of effectiveness actuator failures. sensus algorithm to handle formation control problems for
Using the adaptive updating mechanism and the boundary layer second-order multi-agent systems and pointed out that the
theory, a continuous fault-tolerant formation control protocol
is constructed to compensate for the actuator failures and the
consensus-based approach is more general than many clas-
derivative of the time-varying formations. The proposed proto- sical formation strategies, such as behavior, virtual structure
col is totally distributed without requiring any global knowledge and leader-follower based ones. Distributed time-invariant for-
about the interaction topologies or the bounds of the actuator mation control problems for second-order multi-agent systems
failures. The formation feasible condition is provided, and it is were studied in [16] and [17]. Dong et al. [18], [19] inves-
proved that the formation errors are uniformly bounded and tigated the consensus-based time-varying formation control
can converge to an adjustable small neighborhood of zero by the problems for a cluster of second-order agents in the pres-
proposed continuous protocol under the influences of actuator ence of jointly connected topologies and switching directed
failures. topologies, respectively.
Index Terms—Time-varying formation, fault-tolerant control, It should be pointed out that the characteristics of actu-
adaptive control, multi-agent systems, actuator failures. ators were not considered in [11]–[19]. In some practical
circumstances, each agent may suffer from certain actuator
failures, which could make the actual outputs of actuators dif-
I. I NTRODUCTION
fer from the control inputs. In the framework of distributed
ORMATION control of multi-agent systems has attracted
F significant interest from a variety of scientific fields,
including formation control of multiple mobile robotic sys-
cooperative control of multi-agent systems, the actuator fail-
ure in a single agent can spread over neighboring agents
through the interaction topology and affect the performance
tems [1], unmanned aerial vehicles [2] and micro-satellites [3]. of the whole system. Thus, it is necessary to investigate the
If the states of a group of agents (e.g., positions and velocities) fault-tolerant control approaches for multi-agent systems with
can form and keep a prescribed shape, the desired formation is actuator failures. Fault-tolerant consensus control protocols for
said to be achieved by the multi-agent system. For a complex multi-agent systems with actuator bias faults were presented
system composed of a large quantity of autonomous agents, in [20]–[22], and loss of effectiveness faults were considered
centralized control approaches are no longer applicable and it in [23]–[25]. In [26]–[28], consensus control problems for a
is significant to develop distributed formation control strategies group of second-order agents subject to both bias and loss of
independent of the global information. effectiveness faults were studied. To the best of our knowl-
Inspired by the consensus control of multi-agent sys- edge, fault-tolerant time-varying formation control problems
tems (e.g., [4]–[10]), the consensus-based formation con- for multi-agent systems with directed topologies are still open.
trol approach has received great attention in formation Motivated by the facts stated above, this brief investigates
control field. Distributed formation control protocols for the fault-tolerant time-varying formation control problems for
multi-agent systems with first-order dynamics were presented second-order multi-agent systems with directed interaction
in [11]–[14] based on the consensus strategies. In many topologies in the presence of both bias and loss of effec-
tiveness actuator failures. Compared with the relevant results
Manuscript received June 13, 2017; revised August 4, 2017; accepted on formation control, the main contributions of this brief are
August 31, 2017. Date of publication September 5, 2017; date of current
version May 28, 2018. This work was supported in part by the National threefold. Firstly, the expected time-varying formations can
Natural Science Foundation of China under Grant 61503009, Grant 61333011, still be achieved by the multi-agent systems with actuator fail-
and Grant 61421063, in part by the Aeronautical Science Foundation of ures. In [11]–[19], the characteristics of actuators were not
China under Grant 2016ZA51005, in part by the Special Research Project addressed and the multi-agent systems could not keep the
of Chinese Civil Aircraft, in part by the Innovation Zone Project under
Grant 17-163-11-ZT-004-017-01, and in part by the Fundamental Research predefined formations under the influence of actuator failures.
Funds for the Central Universities under Grant YWF-17-BJ-Y-81. This brief The derivative of the time-varying formation was compen-
was recommended by Associate Editor M. Cao. (Corresponding author: sated by the control input directly in [18] and [19], but it
Xiwang Dong.) is no longer applicable in this brief since the efficiency fac-
The authors are with the School of Automation Science and Electrical
Engineering, Science and Technology on Aircraft Control Laboratory, Beihang
tors of actuators are assumed to be unknown and time-varying.
University, Beijing 100191, China (e-mail: xwdong@buaa.edu.cn). Secondly, the proposed time-varying formation control proto-
Digital Object Identifier 10.1109/TCSII.2017.2748967 col is totally distributed with no need for any global knowledge
1549-7747 c 2017 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: Harbin Engineering Univ Library. Downloaded on April 08,2024 at 09:41:52 UTC from IEEE Xplore. Restrictions apply.
HUA et al.: DISTRIBUTED FAULT-TOLERANT TIME-VARYING FORMATION CONTROL FOR SECOND-ORDER MULTI-AGENT SYSTEMS 775

TABLE I
about the communication topology and the bounds of actua- D IFFERENT M ODES OF ACTUATOR FAILURES
tor failures. The formation protocol in [19] needs to know the
minimum nonzero eigenvalue of Laplacian matrix and is not
fully distributed essentially. In [23]–[25] and [27], the bounds
of actuator failures are considered to be known, but it is diffi-
cult to determine the exact values of these bounds in practice.
Thirdly, the interaction topology is directed and the proposed
control protocol is continuous. The topologies in [16]–[18]
and [22]–[26] are limited to be undirected. The sign function
denote the unknown efficiency factor and the unknown out-
was used to design the consensus control protocol in [27],
put bias of actuator channel j (j = 1, 2, . . . , n), respectively.
which would lead to large chattering in the control input.
Without loss of generality, let n = 1. All the results can be
Based on the boundary layer theory, the continuous control
extended to the higher dimensional case directly using the
protocol in this brief can avoid the large chattering efficiently.
Kronecker product.
Throughout this brief, 1N is used to denote a column vector
The different modes of actuator failures considered in
with size N composed of 1. Let · stand for the Euclidean
this brief are shown in Table I. Both ρi (t) and bi (t) (i =
norm of a vector and diag{·} represent a diagonal matrix. The
1, 2, . . . , N) are assumed to be unknown and time-varying, and
sign function is denoted by sgn(·). A vector is said to be
are required to satisfy the following bounded assumptions.
positive if all its elements are greater than zero. For a matrix
Assumption 1: The unknown efficiency factor ρi (t) is
Q ∈ Rn×m , σmax (Q) stands for its maximal singular value.
bounded, and there exists an unknown positive constant ρ i
such that 0 < ρ i ≤ ρi (t) ≤ 1, i = 1, 2, . . . , N.
II. P RELIMINARIES AND P ROBLEM D ESCRIPTION Assumption 2: The unknown output bias bi (t) is bounded,
A. Basic Graph Theory and there exists an unknown positive constant b̄i such that
Let G = {V, E, A} denote a directed graph with N bi (t) ≤ b̄i , i = 1, 2, . . . , N.
nodes, where V = {v1 , v2 , . . . , vN } denotes the set of nodes, The expected time-varying formation is specified by a
E ⊆ {eij = (vi , vj ) : vi , vj ∈ V; i = j} is the set of edges, and vector h(t) = [hT1 (t), hT2 (t), . . . , hTN (t)]T , where hi (t) =
A = [aij ] ∈ RN×N represents the adjacency matrix with non- [hxi (t), hvi (t)]T (i = 1, 2, . . . , N) is piecewise continuously dif-
negative weights aij . The weight aij > 0 if and only if eji ∈ E, ferentiable. hxi (t) and hvi (t) denote the position and velocity
and aij = 0 otherwise. Let Ni = {vj ∈ V : (vj , vi ) ∈ E} rep- components of hi (t), respectively.
resent the set of neighbors of node vi . The in-degree matrix Definition 1: For any given bounded initial states and
∀i, j ∈ {1, 2, . . . , N}, if
of G is denoted  by D = diag{degin (v1 ), . . . , degin (vN )}, where   
degin (vi ) = N j=1 aij . The Laplacian matrix L of G is defined limt→∞ (xi (t) − hxi (t)) − (xj (t) − hxj (t)) = 0,
as L = D − A. If there exists a directed path from every node   (2)
to every other node, the directed graph G is said to be strongly limt→∞ (vi (t) − hvi (t)) − (vj (t) − hvj (t)) = 0,
connected. then multi-agent system (1) is said to accomplish the time-
Lemma 1 [7]: If the directed graph G is strongly connected, varying formation specified by h(t).
then This brief mainly focuses on how to design a distributed
(i) 0 is a simple eigenvalue of the Laplacian matrix L and fault-tolerant control protocol such that an expected time-
all the other N − 1 eigenvalues have positive real parts. varying formation can be achieved by multi-agent system (1)
(ii)There exists a positive vector ϕ = [ϕ1 , ϕ2 , . . . , ϕN ]T with actuator faults under directed interaction topologies.
with N i=1 ϕi = 1 such that ϕ L = 0.
T

(iii) Let  = diag{ϕ1 , ϕ2 , . . . , ϕN }, then L̂ = L + LT  III. FAULT-T OLERANT T IME -VARYING F ORMATION
is symmetric and can be regarded as the Laplacian matrix C ONTROL P ROTOCOL D ESIGN AND A NALYSIS
associated with an undirected connected graph. Moreover, let
ς ∈ RN×1 stand for any positive column vector and λ2 (L̂) For agent i (i = 1, 2, . . . , N), consider the following fault-
represent the minimum nonzero eigenvalue of L̂. For φ(t) ∈ tolerant time-varying formation control protocol:
RN×1 , it holds that ui (t) = −α̂i (t)ξi (t) − f (ξi (t)), (3)
  λ (L̂) N
min φ T (t)L̂φ(t) >
2
φ T (t)φ(t). where ξi (t) = j=1 aij ((xi (t) − hxi (t)) − (xj (t) − hxj (t)))
φ T (t)ς =0 N +vi (t) − hvi (t). α̂i (t) is an adaptive parameter updated by

B. Problem Description α̂˙ i (t) = γi (−ηi α̂i (t) + ξi (t)2 + (1 + ḣvi (t))ξi (t)), (4)
Consider a second-order multi-agent system with N agents. where α̂i (t0 ) ≥ 0, γi is a positive constant, and ηi is a
The dynamics of agent i (i = 1, 2, . . . , N) with actuator small positive constant chosen by the designer. The nonlinear
failures is described by function f (ξi (t)) is defined as
  βi (t)ξi (t)
ẋi (t) = vi (t),
(1) ξi (t) , βi (t)ξi (t) > ki ,
v̇i (t) = ρi (t)ui (t) + bi (t), f (ξi (t)) = β 2 (t)ξ (t) (5)
i
ki
i
, βi (t)ξi (t) ≤ ki ,
where xi (t) ∈ Rn , vi (t) ∈ Rn and ui (t) ∈ Rn repre-
sent the position, velocity and control input vectors, respec- where βi (t) = α̂i (t)(1 + ḣvi (t)) and ki is a positive constant.
tively. ρi (t) = diag{ρi1 (t), ρi2 (t), . . . , ρin (t)} and bi (t) = According to the boundary layer theory in [29], the function
[bi1 (t), bi2 (t), . . . , bin (t)]T , where 0 < ρij (t) ≤ 1 and bij (t) f (ξi (t)) is the continuous approximation of the sign function

Authorized licensed use limited to: Harbin Engineering Univ Library. Downloaded on April 08,2024 at 09:41:52 UTC from IEEE Xplore. Restrictions apply.
776 IEEE TRANSACTIONS ON CIRCUITS AND SYSTEMS—II: EXPRESS BRIEFS, VOL. 65, NO. 6, JUNE 2018

and ki represents the size of the boundary layers. If ki → 0, z̃i (t) and ξi (t) (i = 1, 2, . . . , N), multi-agent system (6) can be
one can obtain that f (ξi (t)) → βi (t) sgn(ξi (t)). rewritten in the following form:
Remark 1: In [18] and [19], there are no actuator fail- ⎧˙  
⎨z̃(t) = IN − 1N ϕ ξ (t) − Lz̃(t),
T
ures and the derivative of the time-varying formations (i.e.,
ḣvi (t)) can be compensated by the control input directly. Due ξ̇ (t) = −Pα̂ξ (t) − Pf (ξ (t)) + b(t) (10)

to the unknown and time-varying efficiency factors of actu- − ḣv (t) + Lξ (t) − L z̃(t),
2
ators, these control approaches are no longer applicable for
multi-agent systems with actuator failures. In this brief, an where P = diag{ρ1 (t), ρ2 (t), . . . , ρN (t)}, α̂ = diag{α̂1 (t),
adaptive updating term is proposed in (3) to deal with ḣvi (t) α̂2 (t), . . . , α̂N (t)} and f (ξ(t)) = [ f (ξ1 (t)), f (ξ2 (t)), . . . ,
in the presence of loss of effectiveness faults. f (ξN (t))]T . For simplicity, the symbol t of some variables is
Let zi (t) = xi (t) − hxi (t) and ri (t) = vi (t) − hvi (t), i = omitted in the following proof.
1, 2, . . . , N. Multi-agent system (1) can be transformed to The time derivative of V along the trajectory of (10) is
 obtained as
żi (t) = ri (t) + hvi (t) − ḣxi (t),  
(6) V̇ = ξ T Lξ − ξ T L2 z̃ + 2z̃T  − ϕϕ T ξ − z̃T L̂z̃
ṙi (t) = ρi (t)ui (t) + bi (t) − ḣvi (t).
 − ξ T Pα̂ξ − ξ T Pf (ξ ) + ξ T b − ξ T ḣv
Let z̃i (t) = zi (t) − N j=1 ϕj zj (t), where ϕj is defined in N
 
Lemma 1. If limt→∞ z̃i (t) = 0 and limt→∞ ξi (t) = 0 + ρ i α̃i −ηi (α̃i + ᾱ) + ξiT ξi + (1 + ḣvi )ξi  ,
(i = 1, 2, . . . , N), one can obtain that limt→∞ (zi (t) − zj (t)) = i=1
0 and limt→∞ (ri (t) − rj (t)) = 0 (i, j = 1, 2, . . . , N). (11)
Then, it follows from Definition 1 that the expected time-
varying formation is achieved by the multi-agent system (1). where L̂ = L + LT  is defined in Lemma 1.
Therefore, if z̃i (t) and ξi (t) are asymptotic stable, the In light of Assumption 1, one can obtain
time-varying formation problems are solved by the proto- N N
col (3). Let z̃(t) = [z̃T1 (t), z̃T2 (t), . . . , z̃TN (t)]T and ξ(t) = − ξ T Pα̂ξ = − ρi α̂i ξiT ξi ≤ − ρ i α̂i ξiT ξi . (12)
[ξ1T (t), ξ2T (t), . . . , ξNT (t)]T . Hereafter, z̃(t) and ξ(t) are regarded
i=1 i=1
as the time-varying formation errors.
In practical applications, the formation errors can hardly It follows from Assumption 2 that
converge to zero exactly under the influences of actuator fail- N N
ures. If the formation errors are uniformly ultimately bounded ξTb ≤ bi ξi  ≤ b̄i ξi . (13)
with sufficiently small bounds, then multi-agent system (1) is i=1 i=1
said to achieve the time-varying formation with certain small N
error, which is acceptable in most practical circumstances. Note that −ξ T ḣv ≤ i=1 ḣvi ξi  and −α̃i (α̃i + ᾱ) ≤
Theorem 1: Suppose that the directed graph G is strongly − 2 α̃i + 2 ᾱ . Then it holds from (11) that
1 2 1 2
connected and Assumptions 1 and 2 hold. If the desired time-  
varying formation specified by h(t) satisfies the feasibility V̇ ≤ ξ T Lξ − ξ T L2 z̃ + 2z̃T  − ϕϕ T ξ − z̃T L̂z̃
condition hvi (t) − ḣxi (t) = 0, then the time-varying forma- N N N
tion errors z̃(t) and ξ(t) and the adaptive control gains α̂i (t) − ξ Pf (ξ ) −
T
ρ i ᾱξiT ξi + b̄i ξi  + ḣvi ξi 
(i = 1, 2, . . . , N) are uniformly ultimately bounded under the i=1 i=1 i=1
fault-tolerant control protocol (3), and converge exponentially N N
to the bounded set 1
 + ρ i α̃i (1 + ḣvi )ξi  + ρ i ηi (−α̃i2 + ᾱ 2 ).
N 2
1 1 i=1 i=1
D = z̃(t), ξ(t), α̂i (t) : V(t) < ρ i ηi ᾱ 2 + ki , (14)
2δ 2
i=1
(7) Three cases are considered in the following proof.
(i) If α̂i (1 + ḣvi )ξi  > ki , i = 1, 2, . . . , N, then one can
where get from (5) that
1
N
ρ i α̃i2 (t) N
V(t) = z̃T (t)z̃(t) + ξ T (t)ξ (t) + , (8) − ξ T Pf (ξ ) = − ρi α̂i (1 + ḣvi )ξi 
2 2γi
i=1 i=1
 = diag{ϕ1 , ϕ2 , . . . , ϕN } and α̃i (t) = α̂i (t) − ᾱ. The positive N
constant δ satisfies δ < mini=1,...,N {λ2 (L̂)/(2N), γi ηi }, and the ≤− ρ i α̂i (1 + ḣvi )ξi . (15)
positive constant ᾱ is sufficiently large such that i=1

1 δ 4 (L) + 4)
N(σmax b̄i 1 (ii) If α̂i (1 + ḣvi )ξi  ≤ ki , i = 1, 2, . . . , N, then it
ᾱ > max + σmax (L) + , , , follows from (5) that
i=1,...,N ρ 2 λ2 (L̂) ρi ρ
N 2
(9) ρi α̂i2 (1 + ḣvi ) ξi 2
− ξ Pf (ξ ) = −
T
where ρ = min{ρ 1 , ρ 2 , . . . , ρ N }. i=1
ki
Proof: Let b(t) = [bT1 (t), bT2 (t), . . . , bTN (t)]T and ḣv (t) = N 2
ρ i α̂i2 (1 + ḣvi ) ξi 2
[ḣv1 (t), ḣTv2 (t), . . . , ḣTvN (t)]T . Since the time-varying formation
T
≤− . (16)
feasibility condition is satisfied, based on the definitions of ki
i=1

Authorized licensed use limited to: Harbin Engineering Univ Library. Downloaded on April 08,2024 at 09:41:52 UTC from IEEE Xplore. Restrictions apply.
HUA et al.: DISTRIBUTED FAULT-TOLERANT TIME-VARYING FORMATION CONTROL FOR SECOND-ORDER MULTI-AGENT SYSTEMS 777

Furthermore, from (16), one has 4N λ2 (L̂)


2ξ z̃ ≤ ξ 2 + z̃2 . (23)
N λ2 (L̂) 4N
−ξ T Pf (ξ ) + ρ i α̂i (1 + ḣvi )ξi  Then it follows from (20)-(23) that
i=1    4  
N
 2
 δ N σmax (L) + 4
α̂i2 (1 + ḣvi ) ξi 2  ≤ − ρ ᾱ − + σmax (L) + ξ 2
≤ ρi − + α̂i (1 + ḣvi )ξi  2 λ2 (L̂)
ki  
i=1
N λ2 (L̂)
1 − − δ z̃2 . (24)
≤ ρ i ki , (17) 2N
4
i=1
Substituting (24) into (19) yields
  
2
where the Young’s inequality in the following form: − ac +a ≤  4
δ N σmax (L) + 4
4 c with a ≥ 0 and c > 0 has been used.
1
V̇ ≤ −δV − ρ ᾱ − + σmax (L) + ξ 2
(iii) In the case where some agents satisfy (i) but oth- 2 λ2 (L̂)
ers correspond to (ii), without loss of generality, assume   N
that α̂i (1 + ḣvi )ξi  ≤ ki (i = 1, 2, . . . , s) and λ2 (L̂)
− − δ z̃2 − (ρ i ᾱ − b̄i )ξi 
α̂i (1 + ḣvi )ξi  > ki (i = s + 1, s + 2, . . . , N), where s 2N
i=1
is a positive integer and 1 ≤ s ≤ N − 1. Similarly, based on
N N
the results in the two cases, it holds that 1 δ
− (ρ i ᾱ − 1)ḣvi ξi  − ρ i ηi − α̃ 2
N
1
s 2 γi i
− ξ T Pf (ξ ) + ρ i α̂i (1 + ḣvi )ξi  ≤ ρ i ki . (18) i=1 i=1
4 N
i=1 i=1 1 1
+ ρ i ηi ᾱ 2 + ki . (25)
Therefore, according to the above three cases (15), (17) 2 2
i=1
and (18), it follows from (8) and (14) that
δ Let the positive constant δ be sufficiently small such that
V̇ ≤ −δV + ξ T IN + L ξ − ξ T L2 z̃ + 2z̃T ( − ϕϕ T )ξ δ < mini=1,...,N {λ2 (L̂)/(2N), γi ηi } and design ᾱ large enough
2 to satisfy (9). Then it follows from (25) that
N
N
+ z̃T (δ − L̂)z̃ − ρ ᾱξ T ξ − (ρ i ᾱ − b̄i )ξi  1 1
i=1
V̇ ≤ −δV + ρ i ηi ᾱ 2 + ki . (26)
2 2
N N i=1
1 δ
− (ρ i ᾱ − 1)ḣvi ξi  − ρ i ηi − α̃ 2 Using the Comparison lemma in [30], one can obtain
2 γi i  
i=1 i=1 N
1 1
ρ i ηi ᾱ + ki e−δt
N
1 1 V(t) ≤ V(0) − 2
+ ρ i ηi ᾱ 2 + ki , (19) 2δ 2
2 2 i=1
i=1
N
where δ is a positive constant. Let 1 1
+ ρ i ηi ᾱ 2 + ki . (27)
δ 2δ 2
i=1
 = ξ T ( IN + L)ξ − ξ T L2 z̃ + 2z̃T ( − ϕϕ T )ξ
2 It follows from (27) that V(t) converges exponentially to the
+ z̃T (δ − L̂)z̃ − ρ ᾱξ T ξ. bounded set defined in (7) with a convergence rate faster than
Note that e−δt , where the positive constant δ gives a lower bound of
the convergence rate of V(t). Therefore, the time-varying for-
ξ T Lξ ≤ σmax (L)ξ 2 , −ξ T L2 z̃ ≤ σmax
2
(L)ξ z̃, mation errors z̃(t) and ξ(t) and the adaptive control gains
2z̃ ( − ϕϕ )ξ ≤ 2σmax ( − ϕϕ )ξ z̃ ≤ 2ξ z̃,
T T T α̂i (t) (i = 1, 2, . . . , N) are uniformly ultimately bounded. This
completes the proof of Theorem 1.
δz̃T z̃ ≤ δ maxi=1,...,N {ϕi }z̃2 ≤ δz̃2 . Remark 2: Due to the influences of actuator failures,
Then one has the formation errors are proved to be uniformly ultimately
δ bounded in Theorem 1. From (7) and (9), one sees that the
≤ + σmax (L) − ρ ᾱ ξ 2 + (σmax
2
(L) + 2)ξ z̃ bounded set D in (7) depends on the interaction topology
2 of the multi-agent system, the bounds of actuator failures,
+ δz̃2 − z̃T L̂z̃. (20) and the design parameters ki and ηi (i = 1, 2, . . . , N). In
view of the actual performance limitations of actuators, one
Due to the fact that z̃T ϕ = 0 and ϕ > 0, one can obtain from
can design ki and ηi relatively small to make the forma-
Lemma 1 that
tion errors converge to a small neighborhood of zero, which
λ2 (L̂) can meet the requirements of the most practical applications.
z̃2 .
− z̃T L̂z̃ ≤ − (21)
N The nonlinear function f (ξi (t)) is a continuous approxima-
Using the Young’s inequality, one gets tion of the discontinuous function βi (t) sgn(ξi (t)) using the
boundary layer theory. Although the formation errors can
4 (L)
Nσmax λ2 (L̂)
σmax
2
(L)ξ z̃ ≤ ξ 2 + z̃2 , (22) converge to zero using the sign function, it will lead to
λ2 (L̂) 4N large chattering in the control input, which is not allowed in

Authorized licensed use limited to: Harbin Engineering Univ Library. Downloaded on April 08,2024 at 09:41:52 UTC from IEEE Xplore. Restrictions apply.
778 IEEE TRANSACTIONS ON CIRCUITS AND SYSTEMS—II: EXPRESS BRIEFS, VOL. 65, NO. 6, JUNE 2018

practical situations. The continuous control protocol (3) can [9] X. L. Wang, X. F. Wang, H. S. Su, and G. R. Chen, “Fully distributed
avoid the large chattering efficiently. event-triggered semiglobal consensus of multi-agent systems with input
saturation,” IEEE Trans. Ind. Electron., vol. 64, no. 6, pp. 5055–5064,
Remark 3: Inspired by the so-called σ -modification tech- Jun. 2017.
nique in [31], the robust adaptive updating law (4) is proposed [10] H. F. Hong, W. W. Yu, X. H. Yu, G. H. Wen, and A. Alsaedi, “Fixed-
in this brief, and it can be verified that the adaptive gain time connectivity-preserving distributed average tracking for multi-agent
α̂i (t) updated by (4) satisfies α̂i (t) ≥ 0 for all t ≥ t0 in systems,” IEEE Trans. Circuits Syst. II, Exp. Briefs, to be published,
the case where the initial value α̂i (t0 ) is nonnegative. Using doi: 10.1109/TCSII.2017.2661380.
[11] W. Ren and N. Sorensen, “Distributed coordination architecture for
the adaptive updating gains, the proposed protocol (3) can be multi-robot formation control,” Robot. Auton. Syst., vol. 56, no. 4,
determined in a totally distributed form with no need for any pp. 324–333, Apr. 2008.
global knowledge about the communication topology or the [12] F. Xiao, L. Wang, J. Chen, and Y. P. Gao, “Finite-time formation control
bounds of actuator failures. In [19], the minimum nonzero for multi-agent systems,” Automatica, vol. 45, no. 11, pp. 2605–2611,
Nov. 2009.
eigenvalue of L is used to design the formation control pro- [13] K.-K. Oh and H.-S. Ahn, “Formation control and network localization
tocol, which means that the approach in [19] is not fully via orientation alignment,” IEEE Trans. Autom. Control, vol. 59, no. 2,
distributed essentially. In [23]–[25] and [27], the fault-tolerant pp. 540–545, Feb. 2014.
consensus control protocols all need the bounds of actuator [14] Z. Y. Lin, L. L. Wang, Z. M. Han, and M. Y. Fu, “A graph Laplacian
approach to coordinate-free formation stabilization for directed net-
failures, but it is difficult to determine the exact values of these works,” IEEE Trans. Autom. Control, vol. 61, no. 5, pp. 1269–1280,
bounds in practice. Moreover, Theorem 1 can be used to solve May 2016.
the fault-tolerant formation control problems with directed [15] W. Ren, “Consensus strategies for cooperative control of vehicle for-
topologies, while the topologies in [16]–[18] and [22]–[26] mations,” IET Control Theory Appl., vol. 1, no. 2, pp. 505–512,
are limited to be undirected. Mar. 2007.
[16] A. Abdessameud and A. Tayebi, “Formation control of VTOL unmanned
aerial vehicles with communication delays,” Automatica, vol. 47, no. 11,
IV. C ONCLUSION pp. 2383–2394, Nov. 2011.
[17] H. B. Du, S. H. Li, and X. Z. Lin, “Finite-time formation control of
Fault-tolerant time-varying formation control problems for multiagent systems via dynamic output feedback,” J. Robust Nonlin.
second-order multi-agent systems with actuator failures and Control, vol. 23, no. 14, pp. 1609–1628, Sep. 2013.
directed interaction topologies were investigated. Both bias [18] X. W. Dong, L. Han, Q. D. Li, and Z. Ren, “Time-varying formation
and loss of effectiveness faults were considered, and the control for double-integrator multi-agent systems with jointly connected
bounds of the faults can be unknown. With the adaptive updat- topologies,” Int. J. Syst. Sci., vol. 47, no. 16, pp. 3829–3838, 2016.
[19] X. W. Dong, Q. D. Li, R. Wang, and Z. Ren, “Time-varying forma-
ing mechanism, a fully distributed fault-tolerant formation tion control for second-order swarm systems with switching directed
control protocol was presented to compensate for the actuator topologies,” Inf. Sci., vol. 369, pp. 1–13, Nov. 2016.
failures and the derivative of the time-varying formations. An [20] J. Q. Li and K. D. Kumar, “Decentralized fault-tolerant control for satel-
interesting topic for future work is to deal with the case where lite attitude synchronization,” IEEE Trans. Fuzzy Syst., vol. 20, no. 3,
pp. 572–586, Jun. 2012.
only the relative position and velocity information is available. [21] Q. K. Shen, B. Jiang, P. Shi, and J. Zhao, “Cooperative adaptive fuzzy
Another future research direction is to consider the case where tracking control for networked unknown nonlinear multiagent systems
the directed topology contains a spanning tree. with time-varying actuator faults,” IEEE Trans. Fuzzy Syst., vol. 22,
no. 3, pp. 494–504, Jun. 2013.
[22] G. Chen and Y. D. Song, “Fault-tolerant output synchronisation con-
R EFERENCES trol of multi-vehicle systems,” IET Control Theory Appl., vol. 8, no. 8,
[1] W. Wang, J. S. Huang, C. Y. Wen, and H. J. Fan, “Distributed pp. 574–584, May 2014.
adaptive control for consensus tracking with application to formation [23] Z. Q. Zuo, J. Zhang, and Y. J. Wang, “Adaptive fault-tolerant
control of nonholonomic mobile robots,” Automatica, vol. 50, no. 4, tracking control for linear and Lipschitz nonlinear multi-agent sys-
pp. 1254–1263, Apr. 2014. tems,” IEEE Trans. Ind. Electron., vol. 62, no. 6, pp. 3923–3931,
[2] X. W. Dong, B. C. Yu, Z. Y. Shi, and Y. S. Zhong, “Time-varying Jun. 2015.
formation control for unmanned aerial vehicles: Theories and applica- [24] S. Chen, D. W. C. Ho, L. L. Li, and M. Liu, “Fault-tolerant consensus
tions,” IEEE Trans. Control Syst. Technol., vol. 23, no. 1, pp. 340–348, of multi-agent system with distributed adaptive protocol,” IEEE Trans.
Jan. 2015. Cybern., vol. 45, no. 10, pp. 2142–2155, Oct. 2015.
[3] A.-M. Zou, A. H. J. de Ruiter, and K. D. Kumar, “Distributed finite-time [25] L. Zhao and Y. M. Jia, “Neural network-based adaptive consensus track-
velocity-free attitude coordination control for spacecraft formations,” ing control for multi-agent systems under actuator faults,” Int. J. Syst.
Automatica, vol. 67, pp. 46–53, May 2016. Sci., vol. 47, no. 8, pp. 1931–1942, 2016.
[4] G. H. Wen, Z. S. Duan, G. R. Chen, and W. W. Yu, “Consensus tracking [26] Y. J. Wang, Y. D. Song, and F. L. Lewis, “Robust adaptive fault-tolerant
of multi-agent systems with Lipschitz-type node dynamics and switching control of multiagent systems with uncertain nonidentical dynamics and
topologies,” IEEE Trans. Circuits Syst. I, Reg. Papers, vol. 61, no. 2, undetectable actuation failures,” IEEE Trans. Ind. Electron., vol. 62,
pp. 499–511, Feb. 2014. no. 6, pp. 3978–3988, Jun. 2015.
[5] Z. K. Li, Z. S. Duan, and F. L. Lewis, “Distributed robust consensus con- [27] G. Chen and Y.-D. Song, “Robust fault-tolerant cooperative control of
trol of multi-agent systems with heterogeneous matching uncertainties,” multi-agent systems: A constructive design method,” J. Frankl. Inst.,
Automatica, vol. 50, no. 3, pp. 883–889, Mar. 2014. vol. 352, no. 10, pp. 4045–4066, Oct. 2015.
[6] Y. Zhao, G. H. Wen, Z. S. Duan, and G. R. Chen, “Adaptive consensus [28] Y. J. Wang, Y. D. Song, M. Krstic, and C. Y. Wen, “Fault-tolerant finite
for multiple nonidentical matching nonlinear systems: An edge-based time consensus for multiple uncertain nonlinear mechanical systems
framework,” IEEE Trans. Circuits Syst. II, Exp. Briefs, vol. 62, no. 1, under single-way directed communication interactions and actuation
pp. 85–89, Jan. 2015. failures,” Automatica, vol. 63, pp. 374–383, Jan. 2016.
[7] J. Mei, W. Ren, and J. Chen, “Distributed consensus of second-order [29] C. Edwards and S. Spurgeon, Sliding Mode Control: Theory and
multi-agent systems with heterogeneous unknown inertias and control Applications. London, U.K.: Taylor & Francis, 1998.
gains under a directed graph,” IEEE Trans. Autom. Control, vol. 61, [30] H. K. Khalil, Nonlinear Systems. Englewood Cliffs, NJ, USA:
no. 8, pp. 2019–2034, Aug. 2016. Prentice-Hall, 2002.
[8] X. L. Wang, H. S. Su, X. F. Wang, and G. R. Chen, “Nonnegative edge [31] G. Wheeler, C.-Y. Su, and Y. Stepanenko, “A sliding mode controller
quasi-consensus of networked dynamical systems,” IEEE Trans. Circuits with improved adaptation laws for the upper bounds on the norm of
Syst. II, Exp. Briefs, vol. 64, no. 3, pp. 304–308, Mar. 2017. uncertainties,” Automatica, vol. 34, no. 12, pp. 1657–1661, Dec. 1998.

Authorized licensed use limited to: Harbin Engineering Univ Library. Downloaded on April 08,2024 at 09:41:52 UTC from IEEE Xplore. Restrictions apply.

You might also like