See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/273480475

Nonsingular fixed-time consensus tracking for second-order multi-agent

networks

Article in Automatica · April 2015

DOI: 10.1016/j.automatica.2015.01.021

CITATIONS READS

678 1,685

1 author:

Zongyu Zuo
Beihang University (BUAA)
170 PUBLICATIONS 7,701 CITATIONS

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Fixed-time stabilization control with applications to multi-agent systems and multi-UAV systems View project

All content following this page was uploaded by Zongyu Zuo on 08 April 2018.

The user has requested enhancement of the downloaded file.

 Automatica 54 (2015) 305–309

Contents lists available at ScienceDirect

Automatica
journal homepage: www.elsevier.com/locate/automatica

Technical communique

Nonsingular fixed-time consensus tracking for second-order

multi-agent networks✩
Zongyu Zuo 1
The Seventh Research Division, and Science and Technology on Aircraft Control Laboratory, Beihang University (BUAA), Beijing 100191, China

article info abstract

Article history: This paper investigates the fixed-time consensus tracking problem for second-order multi-agent systems
Received 7 May 2014 in networks with directed topology. Global well-defined nonlinear consensus protocols are constructed
Received in revised form with the aid of a newly-designed sliding surface for each double-integrator agent dynamics. In particular,
12 November 2014
the proposed framework eliminates the singularity and the settling time is assignable for any initial con-
Accepted 3 January 2015
ditions. This makes it possible for network consensus problems to design and estimate the convergence
time off-line. Finally, simulation is included to demonstrate the performance of the new protocols.
Keywords:
© 2015 Elsevier Ltd. All rights reserved.
Directed graph
Fixed-time stability
Multi-agent systems
Network consensus
Nonsingular
Sliding mode

1. Introduction Ren (2012) and Cortés (2006) proposed discontinuous finite-time

protocols for first-order agent networks based on different ways of
Network Consensus (Ren, 2008) means that a group of agents applying the signum function. However, it is nontrivial to extend
reaches an agreement upon a common value by local interaction. the finite-time consensus algorithm from the first-order case to
In some applications, groups of agents are required to track a the second-order case straightforwardly. Despite the difficulties,
dynamic leader (time-varying reference information), which is still some progress on this issue has been made recently. The
called consensus tracking. In the analysis of consensus problems, an works due to Wang and Hong (2008) and Zhao, Duan, Wen, and
important performance index for a proposed consensus protocol Zhang (2013) presented the finite-time consensus and tracking
is convergence rate (Wang & Xiao, 2010). Olfati-Saber and Murray control respectively for second-order multi-agent systems via
(2004) proposed a linear consensus protocol and demonstrated using homogeneity with dilation. Khoo, Xie, and Man (2009)
that the algebraic connectivity of a interaction graph qualified expanded the terminal sliding mode technique (Feng, Yu, & Man,
the convergence rate. On the other hand, finite-time consensus 2002) to the finite-time consensus problem of second-order multi-
problem has been promoted to achieve high-speed convergence. robot systems. The work due to Li, Du, and Lin (2011) proposed
The benchmark work due to Bhat and Bernstein (2000) related the continuous finite-time consensus algorithms for leaderless and
regularity properties of the Lyapunov functions to the settling time
leader–follower second-order multi-agent systems respectively by
function in finite-time stability analysis for autonomous systems.
adding a power integrator method.
Xiao and Wang (2007) and Xiao, Wang, and Chen (2011) expanded
However, the settling time functions derived in the finite-time
the finite-time control idea (Bhat & Bernstein, 2000) to multi-
consensus depend on initial states of the agents, which prohibits
agent systems with single-integrator kinematics. Chen, Cao, and
their practical applications if the knowledge of initial conditions is
unavailable in advance. Recently, Zuo and Tie (2014a,b) present a
novel class of nonlinear consensus protocols for single-integrator
✩ This work was supported by the National Natural Science Foundation of
multi-agent networks, called fixed-time consensus which assumes
China (61203022) and Aeronautic Science Foundation of China (2012CZ51029). uniform boundedness of a settling time regardless of the ini-
The material in this paper was not presented at any conference. This paper was
recommended for publication in revised form by Associate Editor Zhihua Qu under
tial conditions, i.e. fixed-time stability (Polyakov, 2012). Unfortu-
the direction of Editor André L. Tits. nately, the direct extension to second-order case has come across
E-mail address: [email protected]. a challenging issue: the singularity in the controller (Feng et al.,
1 Tel.: +86 10 82315584; fax: +86 10 82315584. 2002; Feng, Yu, & Man, 2013; Wu, Yu, & Man, 1998). Motivated
http://dx.doi.org/10.1016/j.automatica.2015.01.021
0005-1098/© 2015 Elsevier Ltd. All rights reserved.
306 Z. Zuo / Automatica 54 (2015) 305–309

by this, this paper extends the idea in Zuo and Tie (2014a,b) where x0 ∈ R, v0 ∈ R and u0 ∈ R denote the position, velocity
and constructs global well-defined fixed-time consensus protocols and input of the reference system. We introduce a virtual leader
for second-order multi-agent networks with directed information π0 with the states ξ0 = [x0 , v0 ]T for the multi-agent system in
flow. The key contributions include: (i) a guaranteed settling time (6). Note, however, that ξ0 is available not to all agents but to only
independent of initial conditions is obtained, i.e., fixed-time con- a portion of agents. Here, define a nonnegative diagonal matrix
sensus; (ii) a continuous sinusoid function is introduced into the B = diag{b1 , b2 , . . . , bN } to indicate the accessibility of ξ0 by the
protocols to eliminate the singularity. To the author’s knowledge, agents, where bi = 1 if ξ0 is accessible by the ith agent, and bi = 0
no results on fixed-time consensus tracking with assignable set- otherwise. The directed graph incorporating π0 into G is denoted
tling time for second-order multi-agent networks are available till by Ge .
now.
Assumption 3. Ge has a spanning tree with π0 being its root
2. Preliminaries vertex, i.e., B ̸= 0.

Assumption 4. The input u0 (t ) of the leader is unavailable to

2.1. Graph theory notions
the group members, but its upper-bound uM 0 is accessible by its
neighboring agents.
A weighted graph G(A) = {V , E , A} consists of a node set
V (G) = {π1 , π2 , . . . , πn }, an edge set E (G) ⊆ V × V and an With a given protocol ui , the closed-loop system in (6) is said
adjacent matrix A = [aij ] ∈ Rn×n . An edge (πi , πj ) on G(A) denotes to reach or achieve fixed-time consensus tracking, if, for ∀ξi (0) and
the state of node πi is available to node πj , but not necessarily vice ∀i, j ∈ IN there exists a constant Tmax > 0 such that the settling
versa. If (πi , πj ) ∈ E , then node πi is called a neighbor of node time T < Tmax and
πj . The index set of all neighbors of πj is denoted by Nj = {i :
lim |ξi (t ) − ξ0 (t )| → 0

(πi , πj ) ∈ E }. The weighted adjacent matrix A of a directed graph t →T (8)
is defined such that aij = 1 for (πj , πi ) ∈ E and aij = 0 otherwise. ξi (t ) = ξ0 (t ), ∀t ≥ T
Let L = [lij ] ∈ RN ×N be the graph Laplacian of G. A graph has a where ξi , [xi , vi ]T , | · | is defined componentwise.
directed spanning tree if a subset of the edges forms a spanning
tree (Ren & Beard, 2005). 3. Nonsingular fixed-time control

2.2. Some lemmas To clarify the core idea, we first consider the fixed-time control
of a single system defined by
Lemma 1 (Zuo & Tie, 2014a). Let ξ1 , ξ2 , . . . , ξN ≥ 0. Then ż1 (t ) = z2 (t ), ż2 (t ) = u(t ) (9)
where z = [z1 , z2 ]T ∈ R2 denotes the system state vector, u ∈ R
 p
N
 N

ξip ≥ ξi if 0 < p ≤ 1 (1) the control input.
i =1 i=1 Let mk , nk , pk , qk be positive odd integers satisfying mk > nk ,
 p p1 < q1 < 2p1 , p2 < q2 and m1 /n1 − p1 /q1 > 1, and αk , βk be
N N
 p
 positive constants, where k = 1, 2. Define a C 1 function µτ (·) :
ξ ≥N 1−p
ξi if 1 < p < ∞. (2)
i [0, +∞) → [0, 1] as
i =1 i =1
 π x
sin · if x ≤ τ
µτ ( x ) = 2 τ (10)
Lemma 2 (Zuo & Tie, 2014b). Consider a scalar system 1 otherwise
m
ẏ = −α y n − β y q ,
p
y(0) = y0 (3) where τ is a positive constant.
To circumvent the singularity problem (Feng et al., 2013), a new
where α > 0, β > 0, m, n, p, q are positive odd integers satisfying sliding surface is proposed:
m > n and p < q. The equilibrium of (3) is globally fixed-time stable q1
with settling time T bounded by s = z1 + [κ(z1 ) · z2 ] p1 (11)

1 n 1 q where κ(·) : R → R+ denotes a scalar positive function, given by

T < Tmax := + . (4)
α m−n β q−p 1
κ(x) = . (12)
If ε , [q(m − n)]/[n(q − p)] ≤ 1, a less conservative estimation of α1 xm1 /n1 −p1 /q1 + β1
the settling time can be obtained instead as In the sequel, the parameter in κ(·) will be omitted for simplicity
after its first definition. It can be straightforwardly verified that
α
  
q 1 1 s = 0 implies
T < Tmax := √ tan−1 + . (5)
q−p αβ β αε m1 p1

z2 = −α1 z1 1 − β1 z1 1 .
n q

A new fixed-time nonlinear control law is defined as

2.3. Problem formulation    m1 p1 q 
1 m1 p1 n1 − q1 −1 p1 1− pq1 2− p11
u = α1 − z1 (κ z2 ) − κ
2 1 z2
Consider a group of N continuous-time agents as κ n1 q1 q1
ẋi (t ) = vi (t ), v̇i (t ) = ui (t )
 q1  q
(6) p1 − pq1 1− p1
−1  m2 p2 
κ 1 · µτ α2 s n2 + β2 s q2
p
− z2 1 · z2 1
(13)
where xi ∈ R, vi ∈ R and ui ∈ R denote the position, velocity q1
and input (protocol) of the ith agent respectively, and i ∈ IN , q / p −1
where µτ (·) is defined in (10) with x = z2 1 1 . The fact sin ωx ≤
{1, 2, . . . , N }. x, ∀ω > 0 ensures the input (13) is always well-defined. The
Let the consensus reference state evolution be following theorem presents the novel fixed-time stability property
ẋ0 (t ) = v0 (t ), v̇0 (t ) = u0 (t ) (7) (Polyakov, 2012).
 Z. Zuo / Automatica 54 (2015) 305–309 307

gets very close to the sliding surface s = 0 where the fixed-

time convergence is guaranteed, and thus ϵ(τ ) → 0. It makes
sense to neglect ϵ(τ ) in practice for small τ due to the estimation
conservativeness of T2 .

Remark 7. It can be verified that the parameter constraints m1 /n1

− p1 /q1 > 1 and p1 /q1 > 1/2 lead to ε1 > 1. Thus, the estimation
(5) is only suitable for computing tighter T2 since the case ε2 ≤ 1
Fig. 1. The phase plot of the system. is possible.

Theorem 5. Consider the second-order system in (9) with the control 4. Nonsingular fixed-time consensus tracking
feedback law designed as (13). Then the state z = [z1 , z2 ]T is globally
Let ep = [e1 , e2 , . . . , eN ]T and ev = [ev1 , ev2 , . . . , evN ]T be the po-
p p p
fixed-time stable and the settling time estimate is derived as
sition and velocity disagreement  vectors, respectively, with their
T < Tmax := T1 + T2 + ϵ(τ ) (14) p
elements defined by ei = j∈Ni aij (xi − xj ) + bi (xi − x0 ) and
where Tk , nk /[αk (mk − nk )] + qk /[βk (qk − pk )], (k = 1, 2), and evi = (v v ) (v − v0 ), taking time derivatives of

j∈Ni a ij i − j + b i i
ϵ(τ ) denotes a small time margin related to τ . which yields
Proof. Differentiating s in (11) against time yields
ėp (t ) = ev (t ), ėv (t ) = (L + B)u(t ) − B1u0 (t ) (17)
q1
−1
q1 (κ z2 )
 m1 p1
where u = [u1 , u2 , . . . , uN ] .
p1
  
m1 p1 n − q −1
T
ṡ = z2 − α1 − z1 1 1 (κ z2 )2 − κ u .
p1 n1 q1 To solve the fixed-time consensus tracking problem, a new
nonsingular distributed protocol may be derived as
Substituting (13) into the preceding equation obtains
  −1  
 q1   m p2 
−1 2
 
ṡ = −µτ
p1
z2 · α2 s n2 + β2 s q2 . (15) ui = aij + bi udi + aij uj − b i uM (si )
0 sgn (18)
j∈Ni j∈Ni

Observe that µτ (·) > 0 if z2 ̸= 0. For the convenience of the proof, where si denotes a sliding surface, defined by
the state space z ∈ R2 is divided into two different areas, as shown
q1
in Fig. 1,
si = ei + [κi (ei ) · evi ] p1
p p
(19)
 q1   q1 
p −1 p −1
S1 = (z1 , z2 )|z2 1 ≥τ , S2 = (z1 , z2 )|z2 1 <τ with the virtual control signal udi defined by
  
(i) When the system states (z1 , z2 ) belong to S1 , the function 1 m1 p1 m1 p
− 1 −1
udi = α1 − (epi ) n1 q1 (κi evi )2
µτ (·) takes value one. Applying Lemma 2, the states (z1 , z2 ) will κi n1 q1
reach the sliding surface s = 0 or enter S2 within fixed-time. q
p1 1− p11 v 2− pq1

p1 − p1
q  q1 
− κi 1 µτi (evi ) p1
−1
(ii) In S2 , 0 < µτ < 1 when z2 ̸= 0. Applying Lemma 2 for − κi (ei ) 1

(15), the sliding surface s = 0 is still an attractor. What remains q1 q1

m2 p2
is to prove that the z1 -axis in Fig. 1 is not attractive except for the q
 
1− p1
· (evi ) α2 si + β2 si
n2 q2
origin (z1 , z2 ) = (0, 0). It can be shown that in close proximity to 1 (20)
z1 -axis the control input (13) degenerates into
p1 − p1  m2 p2  where µτi (·) is defined in (10) with x = (evi )q1 /p1 −1 . It is noted that
u = − κ q1 α2 s n2 + β2 s q2 (16) the threshold parameter τ can be different for each µτi . Without
q1 loss of generality, we choose the same τ for simplicity in this paper.
q /p −1 1−p /q
where the fact µτ (z2 1 1 ) · z2 1 1 → 1 as z2 → 0 is used. In
view of κ −p1 /q1 > 0, we have ż2 < 0 for s > 0 and ż2 > 0 for Theorem 8. Consider a multi-agent system (6) and (7) in networks
s < 0. Thus, z (t ) will transgress S2 into S1 monotonically within with interaction topology satisfying Assumptions 3 and 4. The nonsin-
finite time ϵ(τ ) (refer to Fig. 1). gular distributed protocol proposed in (18) achieves fixed-time con-
It is thus concluded that the sliding surface s = 0 can be reached sensus tracking with the settling time T bounded by
from anywhere in the phase plane within the time t1 < T2 + ϵ(τ ).
T < Tmax := T1 + T2 + ϵ(τ ) (21)
Once the sliding surface s = 0 is reached, it follows from Lemma 2
that (z1 , z2 ) will reach the origin within the period t2 < T1 . Hence, where Tk = [N (mk −nk )/2nk
nk ]/[αk (mk − nk )]+ qk /[βk (qk − pk )], (k =
the uniform bound (14) follows. 1, 2), and ϵ(τ ) denotes a small time margin.

Proof. (i) Substituting (18) into (17) obtains

Remark 6. Note that the traveling time ϵ(τ ) across S2 cannot be
estimated precisely. However, for sufficiently small τ and z1 ̸= 0, ėvi = udi − bi uM
0 sgn(si ) − bi u0 . (22)
integrating both sides of ż2 = u subject to (16) with s → z1 obtains
Differentiating (19) against time and using (22), we have
q1 p1 1 p1
ϵ(τ ) = [κ(|z1 |)] q1 · 2τ q1 −p1   m2 p2 
α2 |z1 |m2 /n2 + β2 |z1 |p2 /q2
 q1
p1 τ
v p1 −1
ṡi = −µi (ei ) · α2 si + β2 si 2
n2 q

where p1 /(q1 − p1 ) > 1. This implies that ϵ(τ ) can be made very
small by choosing sufficiently small τ . For very small z1 including
q1
q1 q1
κi 1 (evi ) p1
−1
0 sgn(si ) + bi u0 .
p
b i uM
 
the case z1 = 0, z2 → 0 in S2 implies the system trajectory − (23)
p1
308 Z. Zuo / Automatica 54 (2015) 305–309

Consider Lyapunov function V2 = (1/2)

N 2
i =1 s i and its time
derivative is
N q1   m2 +1 p2 
q +1

µτi (evi ) p1
−1

α2 si 2 + β2 si 2
n
V̇2 = −
i=1
N q1 q1
κi 1 (evi ) p1
−1

( b i uM
0 |si | + bi u0 si )
p
−
i=1 Fig. 2. The extended information flow Ge .
N  q1  m2 +n2 p2 +q2

µτi (evi ) p1
−1

≤− α2 (s2i ) 2n2
+ β2 (s2i ) 2q2

i=1
where the virtual signal ūdi is given by
N q1
  
q1 1 m1 p1 m1 p
− q 1 −1
v −1 ūdi α1 p
(ēi ) n1 (κi evi )2

κi (ei ) (
p1
− p1
b i uM − |u0 |)|si | = − 1
0 κi n1 q1
i=1
q  q
p1 1− p1 1 q p1 − p11 q1
 
v 2− p1
µτi (evi ) p1
−1
N 
κi 1
(ei ) κi

τ
 m2 +n2 p2 +q2 − −
≤ −µm α ( )2
2 si
2n2
+β ( ) 2
2 si
2q2
q1 q1
i =1 q
 m2 p2 
1− p1
· (evi ) α2 s̄i + β2 s̄i
  n2 q2
n2 −m2 m2 +n2 p2 +q2 1 (27)
≤ −µτm α2 N 2n2
(2V2 ) 2n2
+ β2 (2V2 ) 2q2

q /p
where κi 1 1 · (evi )q1 /p1 −1 ≥ 0, µτm , min{µτ1 , . . . , µτN }, Lemma 1 Theorem 9. Consider a multi-agent system (6) and (7) in networks
√
is used in the last inequality. If V2 ̸= 0, then let y2 = 2V2 be the with interaction topology satisfying Assumptions 3 and 4. The
solution to the differential equation nonsingular distributed protocol proposed in (26) achieves fixed-time
 n2 −m2 m2 p2  consensus tracking with the settling time bounded by
ẏ2 (t ) = −µτm α2 N 2n2 y2 2 (t ) + β2 y2 2 (t )
n q

T < Tmax := T1 + T2 + ϵ(τ ) (28)

√
where ẏ2 = V̇2 / 2V2 is used. Since µτm = 1 if min{|ev1 |, |ev2 |, . . . , where
|evN |} > τ p1 /(q1 −p1 ) , by Theorem 5 and Comparison Principle of N (m1 −n1 )/2n1 n1 1 q1
differential equations Khalil (2005), we obtain limt →t2 V2 (si ) = 0, T1 = (m1 +n1 )/2n1
+ (p1 +q1 )/2q1
λ α1 m1 − n1 λ β1 q1 − p1
implying limt →t2 si = 0 for all i ∈ IN , where the settling time t2 is
bounded by t2 < T2 + ϵ(τ ). N (m2 −n2 )/2n2 n2 1 q2
T2 = + ,
(ii) Note that on the sliding surface si = 0, (19) becomes α2 m2 − n2 β2 q2 − p2
m1 p1
p p
ėi = −α1 (ei ) n1 − β1 (ei ) q1
p
(24) λ denotes the smallest eigenvalue of matrix (L̂ + B), ϵ(τ ) denotes a
small time margin.
for all i ∈ IN . Consider Lyapunov function V1 = (1/2) (epi )2
N
i=1
and its time derivative is The proof follows the same lines of Theorem 8.
N m1 +n1 N p1 +q1
  Remark 10. It is worthwhile noticing that the settling time
V̇1 = −α1 (epi ) n1 − β1 (epi ) q1
estimation (28) also relies on the interaction topology property
i =1 i=1
n1 −m1 m1 +n1 p1 +q1
(i.e. λ). In addition, the introduction of the interaction topology
≤ −α1 N 2n1
(2V1 ) 2n1
− β1 (2V1 ) 2q1 into the definition of the sliding variables guarantees the
√ consensus during the transition. However, the protocol in (18) only
where Lemma 1 is inserted. If V1 ̸= 0, then let y1 = 2V1 be the achieves the reference goal seeking of the multiple agents within
solution to the differential equation fixed-time.
n1 −m1 m1 p1

ẏ1 (t ) = −α1 N y1 1 (t ) − β1 y1 1 (t ).
2n1 n q
5. Simulation example
p
Similarly as (i), we have = 0 with settling time t1
limt →t1 ei
Consider a group of six agents in a network with the interaction
bounded by t1 < T1 for all i ∈ IN .
Thus, ep converges to zero within the period T < T1 + T2 +ϵ(τ ). graph shown in Fig. 2. The time-varying control input of the
Applying Theorem 4 in Khoo et al. (2009), (8) follows, i.e., the fixed- virtual leader (7) is designed as u0 = − sin(x0 )/(1 + exp(−t ))
time consensus tracking is achieved. verifying uM0 = 1. Consider the initial scenario that x(0) = [−5,
−3, 7, 9, 4, 5]T and v(0) = 0 are set for the group of agents (6) and
To incorporate the interaction topology into the sliding surface, x0 (0) = π /2 and v0 (0) = 0 for the virtual leader (7). The design
we may modify the sliding variables as parameters α1 = β1 = α2 = β2 = 2, m1 = 9, n1 = 5, p1 =
q1
7, q1 = 9, m2 = 11, n2 = 9, p2 = 5, q2 = 7 and τ = 0.1 are
s̄i = ēi + [κi (ēi ) · evi ] p1
p p
(25) set for protocols (18) and (26). It can be verified that the parameter
p
( p p
)+ p
with âij = (aij + aji )/2 for constraints m1 /n1 − p1 /q1 > 1 and p1 /q1 > 1/2 are satisfied. The

where ēi = j∈Ni âij ei − ej bi e i
bounds for settling time T in (21) and (28) are 8.025 s and 24.985 s,
all i, j ∈ IN . Let L̂ = (L + L )/2 be the Laplacian matrix of G(Â)
T
respectively. Since ε2 = 7/9 < 1, a less conservative bounds can
with  , [âij ]n×n . Thus, the corresponding protocol for fixed-time be calculated as 7.889 and 24.425 s.
consensus tracking is then obtained: The simulation results in Fig. 3 show that the consensus tracking


 −1 

 objective is achieved within fixed time. Note that the settling times
ui = aij + bi ūdi + aij uj − bi umax
0 sgn(s̄i ) (26) under the two protocols are about respectively 2 s and 20 s, which
j∈Ni j∈Ni demonstrate the performance claimed in Theorems 8 and 9. It can
 Z. Zuo / Automatica 54 (2015) 305–309 309

group seeks consensus among neighbors during transition and

reaches the goal state together.

References

Bhat, S. P., & Bernstein, D. S. (2000). Finite-time stability of continuous autonomous

systems. SIAM Journal on Control and Optimization, 38(3), 751–766.
Chen, F., Cao, Y., & Ren, W. (2012). Distributed average tracking of multiple
time-varying reference signals with bounded derivatives. IEEE Transactions on
Automatic Control, 57(12), 3169–3174.
Cortés, J. (2006). Finite-time convergent gradient flows with applications to
network consensus. Automatica, 42(11), 1993–2000.
Feng, Y., Yu, X., & Man, Z. (2002). Non-singular terminal sliding mode control of
rigid manipulators. Automatica, 28(11), 2159–2167.
Feng, Y., Yu, X., & Man, Z. (2013). On nonsingular terminal sliding-mode control of
nonlinear systems. Automatica, 49(6), 1715–1722.
Khalil, H. K. (2005). Nonlinear systems (3rd ed.). Upper Saddle River, NJ: Prentice
Hall.
Khoo, S., Xie, L., & Man, Z. (2009). Robust finite-time consensus tracking
algorithm for multirobot systems. IEEE/ASME Transactions on Mechatronics,
(a) Protocol (18). 14(2), 219–228.
Li, S., Du, H., & Lin, X. (2011). Finite-time consensus algorithm for multi-agent
systems with double-integrator dynamics. Automatica, 47(8), 1706–1712.
Olfati-Saber, R., & Murray, R. M. (2004). Consensus problems in networks of
agents with switching topology and time-delays. IEEE Transactions on Automatic
Control, 49(9), 1520–1533.
Polyakov, A. (2012). Nonlinear feedback design for fixed-time stabilization of linear
control systems. IEEE Transactions on Automatic Control, 57(8), 2106–2110.
Ren, W. (2008). On consensus algorithms for double-integrator dynamics. IEEE
Transactions on Automatic Control, 53(6), 1503–1509.
Ren, W., & Beard, R. W. (2005). Consensus seeking in multiagent systems under
dynamically changing interaction topologies. IEEE Transactions on Automatic
Control, 50(5), 655–661.
Wang, X., & Hong, Y. (2008). Finite-time consensus for multi-agent networks with
second-order agent. In Proc. 17th IFAC world congress. Seoul, Korea, 6-11 July 2008
(pp. 15185–15190).
Wang, L., & Xiao, F. (2010). Finite-time consensus problem for networks of dynamic
agents. IEEE Transactions on Automatic Control, 55(4), 950–955.
Wu, Y., Yu, X., & Man, Z. (1998). Terminal sliding mode control design for uncertain
dynamic systems. Systems & Control Letters, 34(5), 281–288.
Xiao, F., & Wang, L. (2007). Reaching agreement in finite time via continuous local
state feedback. In Proc. 26th Chinese control conference. Hunan, China, July 2007
(pp. 711–715).
Xiao, F., Wang, L., & Chen, T. (2011). Finite-time consensus of multi-agent systems
with directed and intermittent links. In Proc. 30th Chinese control conference.
Yantai, China, July 2011 (pp. 6539–6543).
Zhao, Y., Duan, Z., Wen, G., & Zhang, Y. (2013). Distributed finite-time tracking
(b) Protocol (26). control for multi-agent systems: an observer-based approach. Systems &
Control Letters, 62(1), 22–28.
Fig. 3. Consensus tracking results. Zuo, Z., & Tie, L. (2014a). A new class of finite-time nonlinear consensus protocols
for multi-agent systems. International Journal of Control, 87(2), 363–370.
Zuo, Z., & Tie, L. (2014b). Distributed robust finite-time nonlinear consensus
be further observed from Fig. 3(a) that each agent in the group protocols for multi-agent systems. International Journal of Systems Science,
achieves the goal state seeking directly, while from Fig. 3(b) the Published online.

View publication stats