Available online at www.sciencedirect.

com

ScienceDirect
IFAC PapersOnLine 53-2 (2020) 4076–4081
Fault-Tolerant
Fault-Tolerant Fully
Fully Distributed
Distributed
Fault-Tolerant
Fault-Tolerant
Leader-Following Fully
Fully Distributed
Distributed
Consensus for Linear
Leader-Following
Fault-Tolerant
Leader-Following Consensus
Fully for
Distributed
Consensus for Linear
Linear
Leader-Following
Multi-agent
Multi-agent Systems
Systems Consensus
with
with for Linear
Non-cooperative
Non-cooperative
Leader-Following
Multi-agent Systems Consensus
with for Linear
Non-cooperative
Leader
Leader
Multi-agent Systems with Non-cooperative
Leader
Leader
Shahram Leader
Shahram Hajshirmohamadi
Hajshirmohamadi
∗∗∗
∗ ∗∗
∗ Farid Sheikholeslam ∗∗
∗ Farid Sheikholeslam
∗∗∗∗ ∗∗
Shahram
Shahram Nader
Nader Meskin
Hajshirmohamadi
Hajshirmohamadi
Meskin ∗∗∗ Jawhar
Jawhar ∗ Farid
∗
FaridGhommam
Ghommam Sheikholeslam
Sheikholeslam ∗∗∗∗ ∗∗
∗∗
Shahram Hajshirmohamadi ∗∗∗ Farid Sheikholeslam ∗∗∗∗
Shahram Nader
Nader Meskin
Meskin
Hajshirmohamadi ∗∗∗
∗∗∗ Jawhar
Jawhar ∗
FaridGhommam
Ghommam Sheikholeslam ∗∗∗∗
∗∗∗∗ ∗∗
∗ Nader Meskin Jawhar Ghommam
∗ Department Nader of Electrical
Meskin ∗∗∗and Computer
Jawhar Ghommam Engineering, ∗∗∗∗ Isfahan
∗ DepartmentUniversity
of Electrical and Computer Engineering, Isfahan
of
Department of Technology,
of Electrical
Electrical Isfahan,
and Computer
Computer Iran,
Iran, (e-mail:
Engineering, Isfahan
∗
∗ DepartmentUniversity
Department of and
Technology,
of Electrical Isfahan, Engineering,
and Computer (e-mail:Isfahan
Engineering, Isfahan
∗ University
University
Department of shahramhajshir@gmail.com)
of
of Technology,
Technology,
Electrical and Isfahan,
Isfahan,
Computer
shahramhajshir@gmail.com) Iran,
Iran, (e-mail:
(e-mail:
Engineering, Isfahan
∗∗ University of Technology, Isfahan, Iran, (e-mail:
∗∗ DepartmentUniversity of Electrical and Computer
shahramhajshir@gmail.com)
shahramhajshir@gmail.com)
of Technology, Isfahan, Engineering,
Iran, (e-mail: Isfahan
∗∗ Department of Electrical and Computer
shahramhajshir@gmail.com) Engineering, Isfahan
University
∗∗ Department
∗∗ Department
University of
of Technology,
of
of
Technology, Isfahan,
Electrical
Electrical and
and
Isfahan, Iran,
Computer
Computer
shahramhajshir@gmail.com)
Iran, (e-mail:
(e-mail: sheikh@cc.iut.ac.ir)
Engineering,
Engineering, Isfahan
Isfahan
sheikh@cc.iut.ac.ir)
∗∗∗ Department
∗∗Department
of
of Electrical
Electrical and Computer
Engineering, QatarEngineering,
University, Isfahan
Doha,
University
University
∗∗∗ Department
Department of
of Technology,
Technology,
of Isfahan,
Isfahan,
Electrical
of Electrical and Iran,
Iran,
Computer
Engineering, (e-mail:
(e-mail: sheikh@cc.iut.ac.ir)
sheikh@cc.iut.ac.ir)
Engineering,
Qatar sheikh@cc.iut.ac.ir)
University, Isfahan
Doha,
University
∗∗∗ of Technology,
Qatar, (e-mail:Isfahan, Iran, (e-mail:
nader.meskin@qu.edu.qa)
∗∗∗ Department
∗∗∗
Department
University of of
of
Technology,
Qatar, Electrical
Electrical Engineering,
Engineering,
Isfahan, Iran, Qatar
Qatar
(e-mail:
(e-mail: nader.meskin@qu.edu.qa) University,
University,
sheikh@cc.iut.ac.ir) Doha,
Doha,
∗∗∗∗Department
Department of Electrical
of Electrical Engineering,
and Computer Qatar University,
Engineering, Doha,
Sultan
∗∗∗
∗∗∗∗Department
Department Qatar,
Qatar, of of (e-mail:
(e-mail:
Electrical
Electrical nader.meskin@qu.edu.qa)
nader.meskin@qu.edu.qa)
Engineering,
and Computer Qatar University,
Engineering, Doha,
Sultan
∗∗∗∗
Qaboos Qatar,
University, (e-mail:
Muscat, nader.meskin@qu.edu.qa)
Oman, (e-mail:
(e-mail: jawher@squ.edu.om)
∗∗∗∗ Department
∗∗∗∗
Qaboos Department Qatar,of
University,
Department of
of
Electrical
Electrical
(e-mail:
Muscat,
Electrical
and
and Computer
Computer
nader.meskin@qu.edu.qa)
Oman,
and Computer
Engineering,
Engineering,
jawher@squ.edu.om)
Engineering,
Sultan
Sultan
Sultan
Qaboos
∗∗∗∗
Qaboos University,
University,
Department Muscat,
of Muscat, Oman,
ElectricalOman, (e-mail: jawher@squ.edu.om)
(e-mail:
and Computer jawher@squ.edu.om)
Engineering, Sultan
Qaboos University, Muscat, Oman, (e-mail: jawher@squ.edu.om)
Abstract: Qaboos University, Muscat, Oman, (e-mail: jawher@squ.edu.om)
Abstract: In In this
this paper,
paper, the the problem
problem of of distributed
distributed fault-tolerant
fault-tolerant control control for for multi-agent
multi-agent
systems
Abstract:
Abstract:
systems is studied.
In
In
is studied.this
this In the
paper,
paper,
In proposed
the
the
the proposed problem
problem method,
method,of
of the communication
distributed
distributed
the communication fault-tolerant
fault-tolerant graph of
control
control the
of the for
graphcontrol network
for is
is not
multi-agent
multi-agent
network not
Abstract:
needed to be In this
known paper,
by the the problem
agents. The of
leader distributed
agent is fault-tolerant
active with a bounded for
non-zeromulti-agent
control
systems
systems
Abstract:
needed is
tois studied.
studied.
be this In
Inknown In the
bythe
paper, proposed
theproposed
the problem
agents. Themethod,
method, the
the
of distributed
leader agent communication
communication
is active graph
withgraph
fault-tolerant of
of
control
a bounded the
the network
network
for
non-zero is
is
multi-agent not
not
control
systems
input. Theis studied.
control In
input the of proposed
the leader method,
and its the
bound communication
is not known graph
by any of the
follower. network
Moreover, is notit
needed
needed
systems
input. to
to
The be
is be known
known
studied.
control by
by
In
input the
the
the agents.
agents.
of proposed
the leaderThe
The leader
leader
method,
and agent
agent
the
its bound is
isis active
active
communication
not knownwith
with by a bounded
aa bounded
graph any of non-zero
thenon-zero
follower. network control
control
Moreover, is not it
needed
is assumedto be
thatknownthe by
leader the isagents. The
non-cooperative leader agent
and doesis active
not with
communicate bounded
with non-zero
any other control
agent.
input.
input.
needed
is The
The
assumedto be control
control
thatknown input
input
by the
theinput
leader of
of the
the leader
leader
agents.
isthe The and
and
non-cooperative its
its
leader bound
bound
agent
and is
is
does not
not
is not
active
not known
knownwithby
communicateby
by any
any
a boundedfollower.
follower.
with Moreover,
Moreover,
non-zero
any othercontrol
agent. it
it
input.
By The
considering controlall these of
limitations, leader the and its
control bound
input isof the known
followers any
are follower.
designed Moreover,
such that theit
is
is assumed
assumed
input.
By The
considering that
that
control the
the
all leader
leader
input
these of is
is non-cooperative
non-cooperative
the
limitations, leader and its
the controland and
and
bound
inputdoes
does
is not
not
not
of not communicate
communicate
known
the communicate by any
followers are designed with
with
follower.any
any other
other
Moreover,
such agent.
agent.
that theit
is assumed
followers canthattrackthe leader
the leader is non-cooperative
in the presence of does
bounded additive actuator with any
faults. Aother agent.
numerical
By
By
is considering
considering
assumed
followers canthat all
all
trackthethese
these
leader
the limitations,
limitations,
leader is the
the
non-cooperative
in the control
control
presence input
input
and
of does
bounded of
of the
the
not followers
followers
communicate
additive are
are
actuator designed
designed
with any
faults. such
such
Aother that
that the
the
agent.
numerical
By considering
simulation is all thesetolimitations, theeffectiveness
control inputofofthe theproposed
followers approach.
are designed such that the
followers
followers
By can
can
considering
simulation
followers canis provided
track
track the
the
all these
provided
track the to
illustrate
leader
leader in
in
limitations,
illustrate
leader
thethe
in the
presence
the
thethe controlof
presence of bounded
bounded
ofinput
effectiveness
presence ofofthe
bounded the additive
additive
followers
proposed
additive
actuator
actuator faults.
faults.
are designed
approach.
actuator faults.such
A
A numerical
A numerical
that the
numerical
simulation is provided to illustrate the
an open
This is the effectiveness
access of
article the
under proposed
the CC approach.
BY-NC-ND license
simulation©is2020
Copyright
simulation
followers canis track The Authors.
provided
provided the to illustrate
leader
to illustratein the the effectiveness
presence of of
bounded the proposed
additive
effectiveness of the proposed approach. approach.
actuator faults. A numerical
(http://creativecommons.org/
Keywords:
simulation Multi-agent
Keywords: is provided to
Multi-agent licenses/by-nc-nd/4.0)
systems,
illustrate
systems, Fully
thedistributed
Fully effectiveness
distributed control, Consensus,
of the Consensus,
control, proposed approach. Fault-tolerant,
Fault-tolerant,
Leader-following.
Keywords: Multi-agent
Keywords: Multi-agent
Multi-agent systems,
Leader-following. systems,
systems, Fully Fully distributed
Fully distributed control,
distributed control, Consensus,
control, Consensus, Fault-tolerant,
Consensus, Fault-tolerant,
Fault-tolerant,
Keywords:
Leader-following.
Leader-following.
Keywords: Multi-agent systems, Fully distributed control, Consensus, Fault-tolerant,
Leader-following.
1. INTRODUCTION
Leader-following. years(Cheng
1. INTRODUCTION years(Cheng and Ugrinovskii (2016);
and Ugrinovskii (2016); Xu Xu et et al.
al. (2015);
(2015);
1. INTRODUCTION
1. INTRODUCTION
INTRODUCTION Cheng
years(Cheng
years(Cheng and Li (2018);
and Wu
Ugrinovskii et al. (2017)).
(2016); Xu et al. (2015);
1. Cheng
years(Cheng and Liand (2018);
and Ugrinovskii
Wu et al.(2016);
Ugrinovskii (2017)).Xu
(2016); Xu et et al.
al. (2015);
(2015);
During
During the the past
past decade,
decade, the
the study
1. INTRODUCTION study of of cooperative
cooperative control control Cheng Cheng
years(Cheng
Since
and
and Li
Li
multi-agent
(2018);
(2018);
and Wu
Wu
Ugrinovskii
systems
et
et al.
al.
usually
(2017)).
(2017)).
(2016);performXu et al. tasks
(2015);
algorithms the for multi-agent systems has received control Since
Cheng and Li
multi-agent (2018); Wu
systems et al.
etusually perform their tasks in
(2017)). their in
During
During the
algorithms
During the
past
past
for
past
decade,
decade, the
multi-agent
decade, the
the
study
study of
systems
study of
of received aa consid-
hascooperative
cooperative
cooperative
control
consid-
control Cheng
unknown
Since
Since
unknown
andand Li (2018);
multi-agent
multi-agent
and sometimes Wu harsh
systems
systems
sometimes
al. (2017)).
usually
usually
harsh environments,
perform
perform
environments, their
their the
the agents
tasks
tasks
agentsin
in
erable
During
erable attention
algorithms
algorithms the for
for
attentionpast (Davoodi
multi-agent
multi-agent
decade,
(Davoodithe et al.,
systems
systems
et study 2016).
has Leader-following
received
hascooperative
of
al., 2016). received aa consid-
consid-
control unknown
Leader-following Since
are multi-agent
subject to the systems
occurrence usually
of perform
faults (Meskintheir tasks
and Kho-in
algorithms for multi-agent systems has received a consid- unknown
Since
are and
and
multi-agent
subject to sometimes
sometimes
the systems
occurrence harsh
harsh
usually
of environments,
environments,
perform
faults (Meskintheir the
the
and agents
agents
tasksKho-in
consensus
erable is
attention
erable attention
attention
algorithms
consensus a cooperative
(Davoodi
(Davoodi systems
isfora multi-agent
cooperative behaviour
et
et al.,
behaviour 2016).
al., 2016).
2016). in multi-agent
Leader-following
Leader-following
hasinreceived
multi-agent sys-
a consid- unknown
sys- rasani, 2009; and sometimes harsh environments, the agents
erable (Davoodi et al., Leader-following are
are subject
subject
unknown
rasani, 2009; andtoMeskin
to the
sometimes
Meskin and
the occurrence
occurrence
and Khorasani,
harsh of
Khorasani, faults2009;
of environments,
faults (Meskin
(Meskin
2009; Meskin
Meskin and
the et
andagents al.,
Kho-
Kho-
et al.,
tems. In
consensus
consensus
erable
tems. In this
is abehaviour,
cooperative
is aabehaviour,
attention
this cooperative
(Davoodithe the states
behaviour
behaviour
et of
al., 2016).
states of ina
in
in group of
multi-agent
multi-agent follower
Leader-following
a group of followersys-
sys- 2010;are subject
Meskin to the
and occurrence
Khorasani, of faults
2011; (Meskin
Chadli et and Kho-
consensus
agents track is
is the
cooperative
states behaviour multi-agent sys-
sys- 2010;
rasani,
rasani,
are subject2009;
2009;
Meskin toMeskin
Meskin and
and
the occurrence
and Khorasani, Khorasani,
and Khorasani, of faults
2011; 2009;
2009;
(Meskin
Chadli et al.,
Meskin
Meskin and2017).
al., et
et al.,
al.,
etKho-
2017).
tems.
tems.
agents
tems.
In
In
consensus
In
this
this
track
this states of
abehaviour,
behaviour,
thecooperative
behaviour,
a
ofthe
the
the
leader.
states In
states
a behaviour
leader.
states of the
of
In
of in
the
a
leader-following
aa group
group
group
of
of follower
multi-agent
of follower
leader-following
follower
rasani,
Due
2010;
2010;
rasani,
Due to
to
2009;
the
Meskin
Meskin
2009;
the
Meskin
communication
and
and
Meskin Khorasani,
Khorasani,
communication
Khorasani,
and
and Khorasani,
and2011;
2011;
2009;
cooperation
Chadli
Chadli
Meskin
et
et
2009; Meskin
cooperation among
al.,
al.,
among
al.,
the
2017).
2017).
et the
al.,
consensus
agents
agents
tems. track
track
In
consensus control,
this the
the the
states
states
behaviour,
control, objective
of
of a
a
the
the objective leader.
leader. is
states to
is toIn
In
of design
the
the
a a control
leader-following
leader-following
group of law
follower 2010; Meskin and Khorasani, 2011; Chadli et al., 2017).
agents
that track
guarantees the this
states of a leader.
behaviour among In design
the
the
a control law Due
leader-following
agents. Leader-
agents,
Due
2010;to
agents,
Due to
aathe
toMeskin fault
the
fault
the
in
and
in an
communication
communication
an agent
Khorasani,
communication agent not
notand
and
and
only
2011;
only affects
Chadli the
cooperation
cooperation
affects
cooperation
behaviour
among
et among
the al.,
behaviour
among
the
the
2017).
the
consensus
consensus control,
control, the
the objective
objective is to
is to
to design
design a control law
agents
that
consensustrack
guarantees the this
control, states
the of a leader.
behaviour
objective among
is In design
the aa control
control
the leader-following
agents. law of
Leader-
law that
agents,
agents,
Due to aaaagent,
fault
fault
the but
in
in an
an
communicationalso
agent
agentdegrades
not
notand only
onlythe performance
theaffects
affects
cooperation the
the of
of the
behaviour
behaviour
theamong the
following
that
that
consensus
that guarantees
guarantees
consensus
following control,
consensus
guarantees
this
this
is used
isbehaviour
thisthebehaviour
behaviour
for different
for among
objective
used among
among
the
the
is to design
different the
applications
agents.
agents.
agents. law of
such
Leader-
Leader-
a control
applications such
Leader-
that
agents,
whole
of
of that
that
agents,
agent,
team fault
agent,
agent,
a such
fault
but
but
but
in
analso
in that
an
agent
also
alsothe
agent
degrades
not only
consensus
degrades
degrades
not onlythe
the
performance
affects
objective
performance
performance
affects the
behaviour
can not
of
of
behaviour be
the
the
as
thatmultiple
following
following unmanned
consensus
consensus
guarantees this is underwater
used
isbehaviour
used for different
for among
differentvehicles
the (YanLeader-
applications
applications
agents. etsuch
al., of
such whole team
that agent, such that
but also the consensus
degrades objective
the performance can not
oflawsbe
the
as multiple
following unmanned
consensus is underwater
used for differentvehicles (Yan
applications et al.,
such achieved.
whole
whole
of thatteam
achieved. team Therefore,
agent, such
such that
butthat
Therefore, designing
alsothe
the fault-tolerant
consensus
consensus
degrades
designing objective
theobjective
performance
fault-tolerant control
can
can
control not
not
oflawsbe
be
the
2019),
as areal
multiple
as multiple
multiple
following
2019), areal vehicles
unmanned
unmanned
consensus (Xuan-Mung
underwater
underwater
vehiclesis(Xuan-Mung
used for different and
vehicles
vehicles
and Hong,
Hong, 2019)
(Yan
(Yan
applications
2019) et and
al.,
etsuch whole
al., achieved.
and for the team such
multi-agent that the
system consensus
is of great objective
importance. can not
Different be
as unmanned underwater vehicles (Yan et al., achieved.
whole
for the team Therefore,
Therefore,
such
multi-agent that designing
designing
the
system fault-tolerant
fault-tolerant
consensus
is of great objective
importance. control
control
can notlaws
laws
Different be
power
2019),
2019),
as
power systems
areal
multiple
systems (Hu
vehicles
areal unmanned
vehicles et al., 2014).
(Xuan-Mung
(Xuan-Mung
underwater
(Hu et(Xuan-Mung
al., 2014). and and
and
vehiclesHong,
Hong,(Yan 2019)
2019) and
and
et and achieved.
al., fault-tolerant Therefore,
control designing
protocols fault-tolerant
have been control
proposed laws
for
2019), areal vehicles Hong, 2019) for
for the
the
achieved. multi-agent
multi-agent
Therefore, system
system is
is
designing of
of great
great importance.
importance.
fault-tolerant Different
Different
control laws
power
power systems
2019), systems (Hu
(Hu
areal vehicles et al., 2014).
et(Xuan-Mung
al.,algorithms
2014). and Hong, 2019) and fault-tolerant
for themulti-agent
linear multi-agent control protocols
system
systems. isIn haveimportance.
of Gallehdari
great been proposed
et Differentfor
Leader-following
power systems
Leader-following
power systems
(Hu
(Hu
control
et
control
et
al.,
al.,
2014).
algorithms
2014).
have
have been been proposed
proposed for
fault-tolerant
fault-tolerant
themulti-agent
linear multi-agent
fault-tolerant
control
control
control
protocols
protocols
system
systems. isIn
protocols
have
have
of Gallehdari
great
have
been
been et al.
importance.
been al. (2017b),
proposed
proposed
Different
(2017b),
proposed
a
for
for
fora
by considering
Leader-following
Leader-following different
control agent
controlagent models
algorithms
algorithms including
have
have been single
proposed
been single
proposed in- fault-tolerant
linear multi-agent
linear multi-agent
multi-agent leader-following
control systems.
systems. In consensus
Gallehdari
In consensus
protocols Gallehdari
have been control
et al.
et al.
al. algorithm
(2017b),
(2017b),
proposed fora
a
by considering different
Leader-following control modelshave
algorithms including
been proposed in- fault-tolerant
linear leader-following
systems. In Gallehdari control
et algorithm
(2017b), a
tegrator,
by double
considering
by considering
considering
Leader-following integrator,
different
different agent
agent
controlagent linear
models
models
algorithms and non-linear
including
including models.
single
single in-
in- is is proposed
fault-tolerant
fault-tolerant
linear multi-agentfor the case
leader-following
leader-following
systems. ofIn multiplicative
consensus
consensus
Gallehdari faults
control
control
et al. and
and in
algorithm
algorithm
(2017b), a
tegrator,
by doubledifferent
integrator, linear
modelsandhave been single
non-linear
including proposed
models. in- proposed for
fault-tolerant the case of multiplicative
leader-following consensus faults
control algorithm in
Linear
tegrator,
tegrator,
by multi-agent
double
double
considering systems
integrator,
integrator,
different agent have
linear
linear
modelsgained
and
and considerable
non-linear
non-linear
including models.
models.
single at-
in- Gallehdari
is
is proposed
proposed
fault-tolerant et al.
for
for (2017a),
the
the case
case
leader-following distributed
of
of control
multiplicative
multiplicative
consensus reconfigura-
faults
faults
control and
and
algorithm in
in
Linear multi-agent
tegrator, systems linear
double integrator, have gained considerable
and non-linear models. at- Gallehdari
is proposed etfor
al. the
(2017a),
case distributed
ofswitching controlfaults
multiplicative reconfigura-
and in
tention recently due to tion
at- Gallehdari strategies et for
al. directed
(2017a),
case distributed topology
control networked
reconfigura-
Linear
Linear multi-agent
tegrator,
tention
Linear multi-agent
double integrator,
recently
multi-agent due to their
systems
systems
systems their have
havegenerality.
linear
have
gained
gained
generality.
gained
In
In fact,
considerable
considerable
and non-linear single
fact,models.
considerable
at-
single
at- Gallehdari
is
tionproposed
strategies
Gallehdari et
et al.
for
al. (2017a),
forthe directed
(2017a), distributed
ofswitching
distributed control
multiplicative
topology
control reconfigura-
faults and in
networked
reconfigura-
integrator
tention
tention
Linear and
recently
recently
multi-agent double
due
due integrator
to
to
systems their
their have systems
generality.
generality.
gained canIn
In be modelled
fact,
fact,
considerable single
single
at- multi-agent
tion
tion strategies
strategies
Gallehdari
multi-agent etsystems
for
for
al.
systems are
directed
directed
(2017a),
are developed.
switching
switching
distributed
developed. In
In Hajshirmohamadi
topology
topology
control networked
networked
reconfigura-
Hajshirmohamadi
integrator
tention and
recently double
due integrator
to systems
their generality. can be
In fact, modelled
single et tion strategies for directed switching topology networked
as linear
integrator systems.
and
and double The leader-following
integrator systemsproblem
systems canIn be for linear
modelled al.
multi-agent
multi-agent (2019b,a); forDavoodi
systems
systems are et
are developed. al.
al. (2016),
developed. In the
the problem
Intopology
Hajshirmohamadi of
integrator
tention
as linearrecently
integrator systems.
and double
due
Theto
double integrator
their generality.
leader-following
integrator systems can
problem
can be
fact,
be for single tion
modelled
linear
modelled et al.strategies
multi-agent
simultaneous
(2019b,a); systems
fault
directed
Davoodi switching
et
are developed.
detection and In Hajshirmohamadi
(2016),
control
networked
problem
Hajshirmohamadi
for multi-agent
of
multi-agent
as
as linear
linear
integrator systems
systems.
systems.
and The
The
double has been extensively
leader-following
leader-following
integrator systems studied
problem
problem
can be in
for
for recent
linear
linear
modelled et
et al.
al.
multi-agent
simultaneous(2019b,a);
(2019b,a); systems
fault Davoodi
Davoodi
are
detection et
et al.
al.
developed.
and (2016),
(2016),
In
control the
the
for problem
problem
Hajshirmohamadi
multi-agent of
of
multi-agent
as systems
linear systems. Thehasleader-following
been extensivelyproblem studiedfor in recent
linear et al. (2019b,a); Davoodi et al. (2016), the problem of
multi-agent
multi-agent
as linear systems
systems
systems. Thehas
has been
been extensively
extensively
leader-following studied
studied
problem in
in
for recent
recent
linear systems
simultaneous
simultaneous
et al.
systems is considered.
(2019b,a);
is fault
fault
considered. DavoodiIn
In these
detection
detection
theseet papers,
and
and
al. an
control
control
(2016),
papers, an H
H for
for
∞
the performance
multi-agent
multi-agent
problem
performance of
multi-agent
 systems has been extensively studied in recent simultaneous
is considered fault
to detection
attenuate the and control
effect of for
∞ multi-agent
fault on the con-
 This paper was supported by Qatar National Research Fund
multi-agent
This paper systems
was has
supported beenby extensively
Qatar National studied in
Research recent
Fund
systems
systems
simultaneous
is considered is
is considered.
considered.
fault
to attenuate In
In these
these
detection the papers,
papers,
andeffect an
an
control
of H
H for
∞
fault performance
performance
multi-agent
on the con-
 This
(a member of wasQatar Foundation) underNational
NPRP Grants number systems
sensus is
error. considered.
In In these papers, an H ∞
∞ performance

 This paper
paper supported by Qatar Research Fund is
is considered
considered
systems toZhu
to et
etInal.
attenuate
attenuate
is considered. (2016),
thepapers,
the
al.these aa fault-tolerant
effect
effect of
of fault
anfault on tracking
on the
the
H∞ performance con-
con-
(a This paperof was
member
NPRP10-0105-170107.was supported
Qatar Foundation)
supported by
by Qatar
The statements
underNational
Qatar NPRP Grants
National
made
Research
Research
hereinGrants
are solely
Fund
number
Fund
the sensus
is error.
considered In Zhu
toisattenuate (2016),
the effectfault-tolerant
of fault tracking
on the con-
(a

(a member
member of
paperof
NPRP10-0105-170107.
(a This
member of
Qatar
Qatar
was
Foundation)
Foundation)
The statements
supported
Qatar Foundation)by Qatar
under
under
undermade NPRP
NPRP
National
NPRP hereinGrants number
number
are solely
Research
Grants the
Fund
number
control
sensus
sensus
is method
error.
error.
considered
control method In
In
to Zhu
Zhu proposed
et
et
attenuate
is proposed al.
al. based
(2016),
(2016),
the
based on
a
a
effect
on intermediate
fault-tolerant
fault-tolerant
of fault
intermediate on estima-
tracking
tracking
the con-
estima-
responsibility of the authors.
NPRP10-0105-170107. The statements made herein are solely the sensus error. In Zhu et al. (2016), a fault-tolerant tracking
NPRP10-0105-170107.
responsibility
(a member ofof Qatar
NPRP10-0105-170107. The
The statements
the authors.
Foundation)
statements undermade
made NPRP herein are
are solely
hereinGrants number
solely the
the control
control
sensus method
method
error. In is
is
Zhu proposed
proposed
et al. based
based
(2016), on
on
a intermediate
intermediate
fault-tolerant estima-
estima-
tracking
responsibility
responsibility of of the
the authors.
authors.
control method is proposed based on intermediate estima-
NPRP10-0105-170107.
responsibility of the The statements made herein are solely the
authors. control method is proposed based on intermediate estima-
2405-8963 Copyright
responsibility of the © 2020 The Authors. This is an open access article under the CC BY-NC-ND license.
authors.
Peer review under responsibility of International Federation of Automatic Control.
10.1016/j.ifacol.2020.12.2433
 Shahram Hajshirmohamadi et al. / IFAC PapersOnLine 53-2 (2020) 4076–4081 4077
tor. In Chen et al. (2015); Zou et al. (2015); Yadegar et al.
(2017); Yadegar and Meskin (2020), adaptive protocols are
used to eliminate the effect of fault on the tracking error.
In most of the fault-tolerant leader-following protocols
such as Gallehdari et al. (2017b); Zou et al. (2015);
Chen et al. (2015), the full state measurement is needed.
It should be noted that in many applications, the full
state can not be measured and only an output signal is
available. Moreover, in most of the proposed approaches
like Gallehdari et al. (2017b); Hajshirmohamadi et al.
(2019b,a); Davoodi et al. (2016), the control input of the
leader should be transmitted to a group of followers while
n many cases, this is not possible. An example of this
situation is the case that the leader is a non-cooperative
target. The proposed method in Zhu et al. (2016) is based
on output feedback and does not need communication with
the leader. However, the Laplacian matrix of the network is
needed to be known for controller design. Laplacian matrix
is based on global information and therefore this approach
is not fully distributed.
In this paper, a fully distributed approach is proposed to
achieve leader-following consensus among linear agents.
The proposed method can also eliminate the effect of
additive actuator faults and it is based on output feedback
and full state measurement is not needed. Moreover,
the proposed approach is fully distributed and a global
information of the network is not needed for designing the
distributed controllers. Based on the authors’ knowledge,
a leader-following method with all these features does not
exist in the literature.
The outline of this paper is organized as follows. Problem
formulation and preliminaries are given in Section 2. Sec-
tion 3 includes the main results of the paper. A simulation
example is presented in Section 4 and finally, Section 5
concludes the paper.
2. PROBLEM FORMULATION AND
PRELIMINARIES
Consider a team of N + 1 homogeneous agents including
N followers and one leader. Without loss of generality, it is
assumed that the leader is indexed by 0 and the followers
are indexed by 1, . . . , N . The dynamics of the i-th follower
is described as
ẋi (t) = Axi (t) + B ui (t) + fi (t) ,
zij (t) = C xi (t) − xj (t) , j ∈ Ni , (1)
for i = 1, . . . , N , where Ni is defined based on the
communication graph of the followers. The dynamics of
the leader is described by
ẋ0 (t) = Ax0 (t) + Bu0 (t), (2)
where xi (t) ∈ R , t ≥ 0, i = 0, . . . , N , is the state of
nx
the i-th agent and zij (t) ∈ Rnz , t ≥ 0, is the measured
relative output of agent i, i = 1, . . . , N, with respect to
agent j ∈ Ni ∪ 0 if the leader is a neighbour of agent
i and j ∈ Ni , otherwise. The inputs ui (t) ∈ Rnu and
fi (t) ∈ Rnu are the control and actuator fault inputs
of the agent i. The matrices A, B and C are known
constant matrices of appropriate dimensions. We assume
that only a non-empty subset of the followers can measure
their outputs relative to the leader. However, there is
no communication link between the followers and the
leader to exchange information. The objective is to design
distributed controllers for the followers such that the
followers can track the leader in the presence of bounded
actuator faults, i.e., xi (t) − x0 (t) → 0 for t → 0.
The following assumptions are used throughout this paper.
Assumption 1. It is assumed that the pair (A, B) is stabi-
lizable and the pair (A, C) is detectable.
Assumption 2. It is assumed that rank(CB) = rank(B) =
nu .
Assumption 3. For every complex number s with non-
negative real part, we have:
 
sIn − A B
rank = n + nu . (3)
C 0
Assumption 4. The control input of the leader u0 (t) and
the fault inputs fi (t), satisfy u0  ≤ uM and fi  ≤ fM ,
where uM and fM are unknown positive constants.
Assumption 5. The communication and measurement
graphs of the follower agents are equal, un-directed and
connected. Moreover, at least one of the followers can
measure its output relative to the leader.
Remark 1. Assumptions 2 and 3 are common in studies
that consider sliding-mode observer design (Raoufi et al.,
2010; Lee et al., 2014; Menon and Edwards, 2014). As-
sumption 3 suggests that all invariant zeros of the triple
(A, B, C) are in the left half plane or equivalently the
triple (A, B, C) is minimum phase. Assumption 4 requires
that the control input of the leader is bounded that is a
general assumption in leader-following consensus problems
(Li et al., 2013; Zhu et al., 2016). Finally, Assumption
5 is common among multi-agent systems with undirected
communication graphs.
Lemma 1. ( Liu et al. (2007) ). Let L = [lij ]N ×N denote
a symmetric and irreducible matrix satisfying lij ≤ 0, i =
N
j, and j=1 lij = 0, for i = 1, . . . , N , and G =
diag(g1 , g2 , . . . , gN ) be a non-zero matrix with gi ≥ 0, i =
1, . . . , N . Then, all the eigenvalues of the matrix L + G are
positive.
3. MAIN RESULTS
3.1 Controller analysis
In this section, an observer-based distributed consensus
protocol is proposed. Before introducing the distributed
observer, we define the new states ζi (t) and ξi (t), i =
1, . . . , N , as
   
ζi (t) = xi (t) − xj (t) + qi xi (t) − x0 (t)
j∈Ni

ξi (t) = zij (t) + qi zi0 (t)
j∈Ni
where qi is defined as qi = 1 if the i-th follower has access
to the leader relative measurement and q i = 0, otherwise.
The proposed observer-based controller for the i-th agent
is now proposed as
4078 Shahram Hajshirmohamadi et al. / IFAC PapersOnLine 53-2 (2020) 4076–4081

   
˙ V ζ̂(t), e(t), η̃(t)
ζ̂i (t) = Aζ̂i (t)+qi Bui (t)+ B ui (t)−uj (t) +Hξi (t)
j∈Ni = ζ̂ T (t)(IN ⊗ P1−1 )ζ̂(t) + eT (t)(IN ⊗ P2 )e(t)
    

N
+B ηi (t)sgn φi (t) −ηj (t)sgn φj (t)  1 2
+ η̃ (t), (9)
j∈Ni
  τ i
i=1 i
+ qi ηi (t)Bsgn φi (t) ,
The time derivative of (9) along the trajectory of (5) is
η̇i (t) = τi ψiT (t)Γψi (t) obtained as
   
   V̇ ζ̂(t), e(t), η̃(t)
+ τi  q F
 i i ξ (t) + F ξ i (t) − ξ j (t)  ,

j∈Ni 1 = 2ζ̂ T (t)(IN ⊗ P1−1 A − L̄G(t)L̄ ⊗ P1−1 BB T P1−1 )ζ̂(t)
 
ui (t) = −ηi (t)Kψi (t)−ηi (t)sgn φi (t) , i = 1, . . . , N, (4) + 2ζ̂ T (t)(IN ⊗ P1−1 HC)e(t)
 
where  + 2eT (t) IN ⊗ P2 (A − HC) e(t)
  
φi (t) = qi ηi (t)F ξi (t) + ηi (t)F ξi (t) − ξj (t) − 2eT (t)(L̄G(t) ⊗ P2 B)sgn (G(t)L̄ ⊗ F C)e(t)
j∈Ni
− 2eT (t)(Q ⊗ P2 B)u0 (t) + (L̄ ⊗ P2 B)f (t)
and τi is a positive scalar. The matrices F ∈ Rnu ×nz ,
N

H ∈ Rnx ×nz , Γ ∈ Rnx ×nx , and K ∈ Rnu ×nx are the
parameters of the controller that will be designed later. +2 ηi (t)ψiT (t)Γψi (t)
 T i=1

Let ζ̂(t) = ζ̂1T (t), . . . , ζ̂NT

(t) , ei (t) = ζi (t) − ζ̂i (t), N   
 
 T  T T +2 ηi (t)qi F Cei (t) + F C ei (t) − ej (t) 
e(t) = eT T
1 (t), . . . , eN (t) , ζ(t) =
T
ζ1 (t), . . . , ζN (t) , i=1 j∈Ni
1

Q = [q1 , . . . , qN ]T , and G(t) = diag(η1 (t), . . . , ηN (t)). N


Using (1), (2), and (4), the dynamics of the aggregated − 2α ψiT (t)Γψi (t)
state vectors ζ̂(t) and e(t) can be written as i=1
  N 
ζ̂(t) = IN ⊗ A − L̄G(t) ⊗ BK ζ̂(t) + (IN ⊗ HC)e(t),    

  − 2α qi F Cei (t)+F C ei (t)−ej (t)  . (10)
ė(t) = IN ⊗ (A − HC) e(t) i=1 j∈Ni
1
    
− L̄G(t) ⊗ B sgn G(t)L̄ ⊗ F C e(t) Using the fact that xT sgn(x) = x1 , and B T P2 = F C, it
+ (L̄ ⊗ B)f (t) − (Q ⊗ B)u0 (t), (5) is easy to show that
 
where L̄ is the modified Laplacian matrix defined as L̄ = eT (t)(L̄G(t) ⊗ P2 B)sgn (G(t)L̄ ⊗ F C)e(t)
L + diag(q1 , . . . , qN ). N   
 
Lemma 2. (Davoodi et al. (2016)). The modified Lapla- = ηi (t)qi F Cei (t) + F C ei (t) − ej (t)  , (11)
1
cian matrix L̄ is positive definite. i=1 j∈Ni

and
Let xei (t) = xi (t) − x0 (t) and xe (t) = [xT T T
e1 (t), . . . , xeN ] . It
2ζ̂ T (t)(L̄G(t)L̄ ⊗ P1−1 BB T P1−1 )ζ̂(t)
can be easily seen that ζ(t) = (L̄ ⊗ Inx )xe (t). If ζ̂(t) → 0
N

and e(t) → 0, then ζ(t) → 0. Since L̄ is non-singular, this
leads to xe (t) → 0 and the consensus will be achieved. =2 ηi ψiT (t)Γψi (t). (12)
In the following theorem, a sufficient condition is given i=1
to design the parameters of the leader-following consensus Moreover, using the Hölder’s inequality we have
protocol (4). − eT (t)(Q ⊗ P2 B)u0 (t) + (L̄ ⊗ P2 B)f (t)
Theorem 1. Suppose that Assumptions 1-4 hold. If there
exist positive definite matrices P1 ∈ Rnx ×nx and P2 ∈ = −eT (t)(L̄1N ⊗ P2 B)u0 (t) + (L̄ ⊗ P2 B)f (t)
 
Rnx ×nx such that T
= −e (t)(L̄ ⊗ P2 B) (1N ⊗ Inu )u0 (t) − f (t)
Ξ < 0, (6)
B T P2 = F C, (7) ≤ L̄ ⊗ B T P2 e(t)1 (1N ⊗ Inu )u0 (t) − f (t)∞ . (13)
where, By using (11), (12) and (13), the inequality (10) leads to
   
AP1 +P1 AT −2BB T HC V̇ ζ̂(t), e(t), η̃(t)
Ξ= ,
C TH T P2 (A−HC)+(A−HC)T P2
(8) ≤ 2ζ̂ T (t)(IN ⊗ P1−1 A − αL̄2 ⊗ P1−1 BB T P1−1 )ζ̂(t)
then, the leader following consensus objective is achieved + 2ζ̂ T (t)(IN ⊗ P1−1 HC)e(t)
 
in presence of actuator fault by considering K = B T P1−1 + 2eT (t) IN ⊗ P2 (A − HC) e(t)
and Γ = P1−1 BB T P1−1 . Moreover the gains ηi converge to
+ L̄ ⊗ B T P2 e(t)1 ((1N ⊗ Inu )u0 (t) − f (t)∞ − α).
a constant value.
(14)
Proof. Define η̃i (t) = ηi (t)−α, η̃(t) = [η̃1 (t), . . . , η̃N (t)]T , Since u0 (t) and fi (t) are bounded according to Assumption
where α is a positive constant. Consider the positive 4 and α can be any positive scalar, for a large enough α
definite Lyapunov candidate function we have
 Shahram Hajshirmohamadi et al. / IFAC PapersOnLine 53-2 (2020) 4076–4081 4079

 
V̇ ζ̂(t), e(t), η̃(t)
Algorithm 1. Obtaining the parameters of the controller (4)
≤ 2ζ̂ T (t)(IN ⊗ P1−1 A − αL̄2 ⊗ P1−1 BB T P1−1 )ζ̂(t) ——————————————————————-
T
+ 2ζ̂ (t)(IN ⊗ P1−1 HC)e(t) 1: Solve LMI AP1 + P1 AT T
 − 2BB < 0 forT P1 .
  2: Solve Her P̄2 (A−HC) < −βInx and B P̄2 = F̄ C for
+ 2eT (t) IN ⊗ P2 (A − HC) e(t). (15)
H and F̄ by using Theorem 3 in Hui and Zak (2005).
Based on Lemma 2, the modified Laplacian L̄ is positive 3: Find γ such that −(HC)T (AP1 + P1 AT −
definite. Hence, there exists a unitary matrix U such that 2BB T )−1 HC < γβInx .
−1
U T L̄U = diag(λ1 , . . . , λN ), where λN > . . . > λ1 > 0. 4: Set F = γ F̄ and K = B T P1 .
Define the transformation ζ̃ = (U T ⊗ P1−1 )ζ̂. Then, we ——————————————————————-
have
  Her(AP1 + P AT − 2BB T ) < 0, (17)
V̇ ζ̃(t), ẽ(t), η̃(t)  
Her P2 (A − HC)
≤ 2ζ̃ T (t)(IN ⊗ AP1 − αΛ2 ⊗ BB T )ζ̃(t)
− (HC)T (AP1 + P AT − 2BB T )−1 HC < 0, (18)
+ 2ζ̃ T (t)(IN ⊗ HC)ẽ(t)
  hold. Since (A, B) is stabilizable according to Assumption
+ 2ẽT (t) IN ⊗ P2 (A − HC) ẽ(t) 1, the linear matrix inequality (17) has a positive definite
N
 solution for P1 (Li et al., 2013). Moreover, since based on

=2 ζ̃iT (t)(AP1 − αλ2i BB T )ζ̃i (t) Assumption 1, (A, C) is detectable and Assumptions 2 and
i=1 3 hold, based on Theorem 3 in Hui and Żak (2005), there

+ ζ̃iT (t)HC ẽi (t) + ẽT
i (t)P2 (A − HC)ei (t) exist H, F , and P2 > 0 such that the Lyapunov inequality
N
 Her P̄2 (A − HC) < −βInx and the equality B T P̄2 = F̄ C

≤2 ζ̃iT (t)(AP1 − BB T )ζ̃i (t) hold. On the other hand, the scalar γ can be found such
i=1 that

+ ζ̃iT (t)HC ẽi (t) + ẽT
i (t)P2 (A − HC)ẽi (t)
−(HC)T (AP1 + P1 AT − 2BB T )−1 HC < γβInx
 T   and consequently we have
 
= ζ̃(t) Ξ ζ̃(t)  −W ζ̃(t), ẽ(t) , (16)  
ẽ(t) ẽ(t) γHer P̄2 (A − HC)
where the last inequality is obtained if α is large enough − (HC)T (AP1 + P1 AT − 2BB T )−1 HC < 0

such that αλi ≥ 1. Since Ξ < 0, we have V ζ̃(t), e(t) ≤ 0 By defining P2 = γ P̄2 , the inequality (18) is satisfied
and the system (5) is stable.
 This means
 thatζ̃(t), e(t) and
 and hence, the inequality (6) holds. Moreover, by setting
∞
ζ(t) are bounded. t=0 W ζ̃(t), e(t) dt = V ζ̃(0), e(0) − F = γ F̄ , the equality constraint (7) holds. Algorithm 1
˙ can be used to find the parameters.
V∞ exists and is bounded. Since ζ̃(t) and ė(t) are bounded,
then Ẇ (ζ̃(t), e(t)) is bounded and W (ζ̃, e) is uniformly
continuous. By using the Barbalat Lemma, W (ζ̃(t), e(t)) 4. SIMULATION
converges to zero which means that ζ̃(t), ζ(t) converge to
zero and as a result xe (t) converges to zero. On the other In this section, a simulation example is provided to demon-
hand, η(t) is nondecreasing and bounded which means it strate the effectiveness of the proposed method. Consider a
converges to a constant. network of unmanned underwater vehicles (UUVs) includ-
ing four followers and a leader. The linearized dynamics of
Remark 2. The controller (4) and the conditions of The- the diving system is described by (1) with xi = [ωi θi zi ]T
orem 1 illustrate the advantage of the proposed method where ωi is the pitch rate, θi is the pitch angle, and zi is
compared with the existing approaches. It can be seen that the depth of the i-th UUV (Healey and Lienard (1993)).
different from Gallehdari et al. (2017b), relative output The parameters of the model are given as
measurements are used instead of relative states. Com-      
−0.7 −0.3 0 0.35
pared to Gallehdari et al. (2017b); Khalili et al. (2020), in 1 0 0
A= 1 0 0 ,B = 0 ,C = . (19)
the proposed method none of the followers need communi- 0 0 1
0 −0.3 0 0
cation of the control input or any variable with the leader.
Moreover, different from Zhu et al. (2016), in the proposed The control input ui (t) is the deflection of the control
method, no global information such as the eigenvalues of surface from the stern plane. Each UUV can measure
the Laplacian matrix is needed. its depth and pitch rate relative to its neighbours. The
communication graph of the network is given in Fig. 1.
The control input of the leader is selected as u0 (t) =
3.2 Controller synthesis Kr x0 (t) + sin(0.5t), t ≥ 0, where Kr = [12.85 24.02 −
44.68]. The parameters of the controller is designed by
In this subsection, we propose a method to design the using Algorithm 1 and setting τ = 100. It is assumed that
parameters of the controller (4). Algorithm 1 is proposed at t = 50s, a bias fault f2 (t) = 1 occurs in the second
for this purpose. In the following lines we explain how agent. The trajectories of the states of the agents and the
this algorithm works. According to Schur complement adaptive gains ηi are shown in Fig. 4 and 5, respectively.
lemma, the inequality (6) holds if and only if the following It can be seen that the consensus is achieved in presence
inequalities of actuator fault in agent 2.
4080 Shahram Hajshirmohamadi et al. / IFAC PapersOnLine 53-2 (2020) 4076–4081

-8
4 1 0

Depth (m)
-10
3
-12

Fig. 1. The communication/measurement graph of the -14

agents where the dashed arrows denote the communi- 0 50 100 150 200
cation among the agents and the solid arrows denote Time (s)
the relative measurements available for each agent.

Fig. 4. The depth trajectories of agents.

1
Pitch rate (rad/s)

3
0

i
2
-1
0 50 100 150 200
Time (s)
1
0 50 100 150 200
Time (s)
Fig. 2. The pitch rate trajectories of agents.

Fig. 5. Adaptive gains ηi , i = 1, . . . , 4.

and switching network topology will be considered in our
2 future work.
Pitch angle (rad)

0 Chadli, M., Davoodi, M., and Meskin, N. (2017). Dis-
-2
0 50 100
Time (s)
Fig. 3. The pitch angle trajectories of agents.
5. CONCLUSION
In this paper, a fault-tolerant control protocol has been
proposed for linear multi-agent systems. The proposed
method does not need full-state measurement of the
agents. The leader is considered to be non-cooperative and
its control input is assumed to be unknown to all followers.
Moreover, the global information of the network is not used
in controller design. By considering all these limitations,
an algorithm is proposed to design the parameters of the
controller by using linear matrix inequalities. A numerical
example has been presented to illustrate the effectiveness
of the proposed method. The case of multiplicative faults
REFERENCES
tributed state estimation, fault detection and isola-
tion filter design for heterogeneous multi-agent linear
parameter-varying systems. IET Control Theory & Ap-
plications, 11, 254–262.
150 200
Chen, S., Ho, D.W.C., Li, L., and Liu, M. (2015). Fault-
tolerant consensus of multi-agent systems with dis-
tributed adaptive protocol. IEEE Trans. Cybern.,
45(10), 2142–2155.
Cheng, B. and Li, Z. (2018). Fully distributed event-
triggered protocols for linear multiagent networks. IEEE
Transactions on Automatic Control, 64(4), 1655–1662.
Cheng, Y. and Ugrinovskii, V. (2016). Event-triggered
leader-following tracking control for multivariable multi-
agent systems. Automatica, 70, 204–210.
Davoodi, M., Meskin, N., and Khorasani, K. (2016). Si-
multaneous fault detection and consensus control design
for a network of multi-agent systems. Automatica, 66,
185–194.
Davoodi, R., Gallehdari, Z., Saboori, I., Rezaee, H.,
Semsar-kazerooni, E., Meskin, N., Abdollahi, F., and
Khorasani, K. (2016). An overview of cooperative and
consensus control of multiagent systems. Wiley Ency-
clopedia of Electrical and Electronics Engineering, 1–35.
 Shahram Hajshirmohamadi et al. / IFAC PapersOnLine 53-2 (2020) 4076–4081 4081
Gallehdari, Z., Meskin, N., and Khorasani, K. (2017a).
Distributed reconfigurable control strategies for switch-
ing topology networked multi-agent systems. ISA Trans-
actions, 71, 51 – 67. Special issue on Distributed Coordi-
nation Control for Multi-Agent Systems in Engineering
Applications.
Gallehdari, Z., Meskin, N., and Khorasani, K. (2017b).
An H∞ cooperative fault recovery control of multi-agent
systems. Automatica, 84, 101–108.
Hajshirmohamadi, S., Sheikholeslam, F., Davoodi, M., and
Meskin, N. (2019a). Event-triggered simultaneous fault
detection and tracking control for multi-agent systems.
International Journal of Control, 92(8), 1928–1944.
Hajshirmohamadi, S., Sheikholeslam, F., and Meskin, N.
(2019b). Distributed simultaneous fault detection and
leader-following consensus control for multi-agent sys-
tems. ISA Tran., 87, 129–142.
Healey, A.J. and Lienard, D. (1993). Multivariable sliding
mode control for autonomous diving and steering of un-
manned underwater vehicles. IEEE journal of Oceanic
Engineering, 18(3), 327–339.
Hu, J., Li, Y., Yong, T., Cao, J., Yu, J., and Mao, W.
(2014). Distributed cooperative regulation for multia-
gent systems and its applications to power systems: A
survey. The Scientific World Journal.
Hui, S. and Zak, S. (2005). Observer design for systems
with unknown inputs. International Journal of Applied
Mathematics and Computer Science, 15, 431–446.
Hui, S. and Żak, S. (2005). Observer design for systems
with unknown inputs. International Journal of Applied
Mathematics and Computer Science, 15(4), 431–446.
Khalili, M., Zhang, X., Cao, Y., Polycarpou, M.M., and
Parisini, T. (2020). Distributed fault-tolerant control
of multiagent systems: An adaptive learning approach.
IEEE Transactions on Neural Networks and Learning
Systems, 31(2), 420–432.
Lee, D.J., Park, Y., and Park, Y.S. (2014). Robust H∞
sliding mode descriptor observer for fault and output
disturbance estimation of uncertain systems. IEEE
Transactions on Automatic Control, 57(11), 2928–2934.
Li, Z., Liu, X., Ren, W., and L.Xie (2013). Distributed
tracking control for linear multiagent systems with a
leader of bounded unknown input. IEEE Transactions
on Automatic Control, 58(2), 518–523.
Liu, Z.X., Chen, Z.Q., and Yuan, Z. (2007). Pinning
control of weighted general complex dynamical networks
with time delay. Physica A, 375(1), 345–354.
Menon, P.P. and Edwards, C. (2014). Robust fault esti-
mation using relative information in linear multi-agent
networks. IEEE Transactions on Automatic Control,
59(2), 477–482.
Meskin, N. and Khorasani, K. (2009). Actuator fault
detection and isolation for a network of unmanned
vehicles. IEEE Transactions on Automatic Control,
54(4), 835–840.
Meskin, N. and Khorasani, K. (2009). Fault detection and
isolation of discrete-time markovian jump linear systems
with application to a network of multi-agent systems
having imperfect communication channels. Automatica,
45(9), 2032 – 2040.
Meskin, N. and Khorasani, K. (2011). Fault Detection and
Isolation: Multi-Vehicle Unmanned Systems. Springer-
Verlag New York.
Meskin, N., Khorasani, K., and Rabbath, C.A. (2010).
A hybrid fault detection and isolation strategy for
a network of unmanned vehicles in presence of large
environmental disturbances. IEEE Transactions on
Control Systems Technology, 18(6), 1422–1429.
Raoufi, R., Marquez, H.J., and Zinober, A.S.I. (2010). H∞
sliding mode observers for uncertain nonlinear Lipschitz
systems with fault estimation synthesis. International
Journal of Robust and Nonlinear Control, 20(16), 1785–
1801.
Wu, Z.G., Xu, Y., Lu, R., Wu, Y., and Huang, T. (2017).
Event-triggered control for consensus of multiagent sys-
tems with fixed/switching topologies. IEEE Trans-
actions on Systems, Man, and Cybernetics: Systems,
48(10), 1736–1746.
Xu, W., Ho, D.W., Li, L., and Cao, J. (2015). Event-
triggered schemes on leader-following consensus of gen-
eral linear multiagent systems under different topolo-
gies. IEEE transactions on cybernetics, 47(1), 212–223.
Xuan-Mung, N. and Hong, S.K. (2019). Robust adap-
tive formation control of quadcopters based on a
leader–follower approach. International Journal of Ad-
vanced Robotic Systems, 16(4).
Yadegar, M., Meskin, N., and Afshar, A. (2017). Fault-
tolerant control of multi-agent systems based on adap-
tive fault hiding framework. In 2017 American Control
Conference (ACC), 4111–4116.
Yadegar, M. and Meskin, N. (2020). Output feedback
fault-tolerant control of heterogeneous multi-agent sys-
tems. Asian Journal of Control.
Yan, Z., Pan, X., Yang, Z., and Yue, L. (2019). Forma-
tion control of leader-following multi-uuvs with uncer-
tain factors and time-varying delays. IEEE Access, 7,
118792–118805.
Zhu, J.W., Yang, G.H., Zhang, W.A., and Yu, L. (2016).
Cooperative fault tolerant tracking control for multi-
agent systems: an intermediate estimator-based ap-
proach. IEEE Transactions on Cybernetics., 58(6), 318–
327.
Zou, Z., Zhang, J., and Wang, Y. (2015). Adaptive fault
tolerant tracking control for linear and Lipschitz non-
linear multi-agent systems. IEEE Trans. Ind. Electron.,
62(6), 3923–3931.