6th IFAC Workshop on Distributed Estimation and Control in

6th
6th IFAC
IFAC Workshop
Networked Systemson
Workshop on Distributed
Distributed Estimation
Estimation and
and Control
Control in
in
6th IFAC
NetworkedWorkshop
Systems on Distributed Estimation and Control in
Networked
September
Networked Systems
8-9, 2016.
Systems Tokyo, Japan Available online at www.sciencedirect.com
September
September 8-9, 2016. Tokyo, Japan
September 8-9,
8-9, 2016.
2016. Tokyo,
Tokyo, Japan
Japan
ScienceDirect
IFAC-PapersOnLine 49-22 (2016) 187191
Stability
Stability analysis
analysis of
of mobile
mobile robot
Stabilitybased
formations analysis
on of mobile robot
robot
synchronization of
formations based
formationscoupled on
based on synchronization
synchronization of
of
coupled oscillators
oscillators
coupled oscillators
Tadashi Nakamura Miyuki Tsukiji Naoyuki Hara
Tadashi
Tadashi Nakamura
Nakamura Miyuki Tsukiji
Miyuki Tsukiji
Naoyuki Hara
Naoyuki Hara
Tadashi Nakamura Keiji
Keiji
Konishi
Miyuki Tsukiji
Konishi

Naoyuki Hara
Keiji Konishi
Keiji Konishi

Department of Electrical and Information Systems,
Department of Electrical and Information Systems,
Department of
Department of Electrical
Osaka
OsakaElectrical and
Prefecture Information
Information Systems,
andUniversity,
Prefecture University, Systems,
Osaka
11 Gakuen-cho,
Osaka Prefecture
Naka-ku, University,
Sakai,
Prefecture Osaka 599-8531 Japan
University,
11
11 Gakuen-cho,
(e-mail:Naka-ku, Sakai, Osaka 599-8531 Japan
11 Gakuen-cho,
Gakuen-cho, Naka-ku,
(e-mail:Naka-ku,
Sakai,
Sakai, Osaka
Osaka 599-8531
599-8531 Japan
n-hara@eis.osakafu-u.ac.jp)
n-hara@eis.osakafu-u.ac.jp) Japan
(e-mail: n-hara@eis.osakafu-u.ac.jp)
(e-mail: n-hara@eis.osakafu-u.ac.jp)
Abstract: The present paper deals with mobile robot circular formations realized by a
Abstract:
Abstract: The
The present
present paper
paper deals
deals with by mobile robot circular formations realized by a
simple control
Abstract:
simple control
law.
Thelaw. This law
present
This paper
law is deals with
is established
established with by mobile
mobile robot
utilizing
robot
utilizing
circular
circular formations
synchronization
synchronization
of coupled
formations
of coupled
realized
realized by
by aa
limit-cycle
limit-cycle
simple
oscillators.
simple control
control law.
Thelaw. This
stability
This law law is established
of formations
is established by utilizing
is analyzed,
by utilizing synchronization
andsynchronization
the analytical of of coupled
results
coupled limit-cycle
givelimit-cycle
a simple
oscillators.
oscillators.
procedure for The
The stability
stability
designing aof
of formations
formations
control law. This is
is analyzed,
analyzed,
procedure and
and the
the
allows analytical
analytical
us to choose results
results give
give
a desired aacircular
simple
simple
oscillators.
procedure The
for stability
designing a of formations
control law. is
This analyzed,
procedure and the
allows analytical
us to chooseresults
a give
desired a simple
circular
procedure
formation. for
procedure for designing a
Furthermore,
designing athe
control law. This
analytical
control law. This
resultsprocedure
guarantee
procedure allows
thatus
allows uswetotocan choose a desired
specify
choose a desired
any radius circular
circular of
formation.
formation. Furthermore,
Furthermore,
circular formations. the
the
The results analytical
analytical
are veried results
results guarantee
guarantee
by some numericalthat
that we
we
examples.can
can specify
specify any
any radius
radius of
of
formation.
circular Furthermore,
formations. The the analytical
results are veried results
by guarantee
some numerical that we
examples.can specify any radius of
circular
circular formations.
formations. The
The results
results are
are of veried
veried by
by some
some numerical
numerical examples.
examples.
2016, IFAC (International Federation Automatic Control) Hosting by Elsevier Ltd. All rights reserved.
Keywords: Stability, Mobile robot, Formation control, Coupled oscillators, Synchronization
Keywords:
Keywords: Stability,
Stability, Mobile
Mobile robot,
robot, Formation
Formation control, control, Coupled
Coupled oscillators,
oscillators, Synchronization
Synchronization
Keywords: Stability, Mobile robot, Formation control, Coupled oscillators, Synchronization
1. INTRODUCTION does not allow the in-depth theoretical analysis of the
1.
1. INTRODUCTION
INTRODUCTION does
does not
formationnot allow
controlthe
allow the
law. in-depth
in-depth theoretical
theoretical analysisanalysis of of the
the
1. INTRODUCTION does
formationnot allow
control the
law. in-depth theoretical analysis of the
Unmanned vehicles have been used for a wide range of formation formation control
control law.
law.
Unmanned vehicles have Complex network science has created considerable interest
Unmanned vehicles have been used
used for
beenexploration, for a wide
wide rangerange of
a environmental of Complex
purposes such
Unmanned
purposes
as planetary
vehicles have been used for a wide range of in variousnetwork
Complex
Complex network science
science has
elds [Strogatz
network science has
has
created
created
(2003)].
created It isconsiderable
considerable
well accepted
considerable
interest
interest
that
interest
monitoringsuch
purposes
purposes andas
such
such as
planetary
planetary exploration,
planetary exploration, environmental
asinvestigation/restoration
exploration, environmental
work in dan- in
environmental in various
various
coupled elds
elds
oscillators[Strogatz
[Strogatz (2003)].
(2003)].
have played It
It is
is
aIt key well
well accepted
accepted
roleaccepted
in complex that
that
monitoring
monitoring and
and investigation/restoration
investigation/restoration work
workInin in dan-
inrecent in
dan- coupled various elds
oscillators[Strogatz
have (2003)].
played is
aa(2001)].
key well
role in that
complex
gerous places
monitoring and (e.g., damaged nuclear plants).
investigation/restoration work dan- coupled
network oscillators
science have
[Pikovsky played
et al. key roleA in complex
number of
gerous
gerous places
places (e.g., damaged
(e.g., damaged
damaged nuclear
nuclear plants).
plants). In
Inhasrecent
recent coupled oscillators have played a key role in complex
years, the
gerous placescooperative
(e.g., use of nuclear
multipleplants).vehiclesIn recentat- network
network
studies science
science
have been [Pikovsky
[Pikovsky
made on et
et
the al.
al. (2001)].
(2001)].
collective A
A number
number
phenomena of
of
in
years,
tractedthe
years,
years, the cooperative
interest
the in bothuse
cooperative
cooperative use of
of multiple
academia
use of multiple vehicles
vehicles [Kumar
and industry
multiple vehicles
has
has at-
has at- network
at- studies
studies
science
have
have been
been
[Pikovsky
made
made on
on
ettheal.collective
the
(2001)]. phenomena
collective
A number in
phenomena
of
in
tracted interest in both academia and industry [Kumar coupled
studies oscillators
have been in
made order
on to
the clarify
collective the fundamental
phenomena in
tracted
tracted interest
et al. (2005);
interest in
Mesbahi both
in bothand andacademia
Egerstedt
academia and industry
(2010)].
and(2010)].
industry [Kumar
[Kumar coupled
Formation coupled oscillators
oscillators in
in order
order to
to clarify
clarify the fundamental
et al.
al. (2005);
mechanism
coupled of various
oscillators in types
order of clarify the
tosynchronization the fundamental
appeared
fundamental
et
control
et al. is suchMesbahi
(2005);
(2005); Mesbahi
an example
Mesbahi and Egerstedt
and Egerstedt
where multiple
Egerstedt (2010)].
Formation
(2010)].vehicles
Formation
Formation are mechanism
mechanism
in complex of
of various
various
networks types
types
[Boccaletti of
of synchronization
synchronization
et al. (2006); appeared
appeared
Arenas et al.
control
control is
is such
such an
an example
example where
where multiple
multiple vehicles
vehicles are
are mechanism
in complex of various
networks types
[Boccaletti of synchronization
et al. (2006); appeared
Arenas et al.
required
control isto form
such an a specic
example formation
where pattern.
multiple Motiva-
vehicles are in complex networks [Boccaletti et al. (2006); Arenas et
required to
to form
form aaaformations
specic
specic formation Motiva- in
(2008)].
complex In recent
networks years, the
[Boccaletti mechanism
et al. (2006); is widely
Arenas et al.
used in
al.
required
tions for achieving
required to form specic come frompattern.
formation
formation pattern.
pattern.
Motiva-
various reasons,
Motiva- (2008)].
(2008)].
engineering In
In recent
recent years,
years,
applications, the
the
such mechanism
mechanism
as locomotion is
is widely
widely used
used
control ina
of in
tions
tions for achieving
for achieving
achieving formations
formations come from
come from
from various
various reasons,
reasons, (2008)]. In recent years, the mechanism is widely used ina
depending
tions for on a specicformations application;
come e.g., to realize
various reasons, an engineering
engineering applications,
applications,
biomimetic applications,
underwater vehicle such
such as
as locomotion
locomotion
[Zhou and Low control
control of
of
(2012)], aa
depending
depending on a specic
on aa specic
specic application;
application; e.g.,
e.g., to realize
to realize
realize an
an engineering such as locomotion control of
optical interferometer
depending on array by makinge.g.,
application; satellites,
to each an of biomimetic
biomimetic
synchronization underwater
underwater
of vehicle
vehicle
pulse-coupled [Zhou
[Zhou and
and
oscillators Low
Lowfor (2012)],
(2012)],
wireless
optical
them is interferometer
optical
optical assumed to bearray
interferometer
interferometer array
equipped
array
by
by making
by with a satellites,
making
making telescope, each
satellites,
satellites, form of
each
each ofa biomimetic
of synchronization
synchronization
underwater
of
of
vehicle [Zhou
pulse-coupled
pulse-coupled
and Low
oscillators
oscillators for
for
(2012)],
wireless
wireless
them is assumed to be equipped with a telescope, form a sensor networks [Okuda
synchronization of et al. (2011)],
pulse-coupled peak power
oscillators for reduc-
wireless
them is assumed
desiredis spatial
them assumed to
pattern be equipped
[Mesbahi
to be [Mesbahi
equipped and with
and a telescope,
Egerstedt
withEgerstedt
a telescope, form
(2010)].
formIn a
In sensor
a sensor networks [Okuda et al.
al. (2011)],
(2011)], peak power reduc-
reduc-
desired spatial pattern (2010)]. tion in networks
sensor energy storage
networks [Okuda
[Okuda et
oscillators
et al. coupledpeak
(2011)], peak power
by delayed
power power
reduc-
desired
formation
desired spatial
spatial pattern
control,
pattern [Mesbahi interest
of particular
[Mesbahi and Egerstedt
and Egerstedt (2010)].
is the (2010)].
case when In tion
In tion in
in energy
energy storage
storage oscillators
oscillators coupled
coupled by
by delayed
delayed power
power
formation control, of particular interest is the case when pricein[Fukunaga
tion energy et al. oscillators
storage (2016)], and so on by
coupled [In delayed
et al. (2009)].
power
formation
each vehicle
formation control,
makes
control, of
ofa particular
decision
particular interest
by itself
interest is the
with
is the case
only
case when
local
when price
price [Fukunaga
[Fukunaga et
et al.
al. (2016)],
(2016)], and
and so
so on
on [In
[In et
et al.
al. (2009)].
(2009)].
each
each vehicle makes
vehicle makes a decision
makes aa decision
decision by by
by itself
itself with
with only only
only local price [Fukunaga et al. (2016)],
local Our previous study suggested that synchronization in and so on [In et al. (2009)].
measurements/communications,
each vehicle anditself
a global
with formationlocal
measurements/communications, Our previous study suggested
has beenthat synchronization in
pattern is achieved as a result ofand
measurements/communications,
measurements/communications, the a
and
and a
global
global formation
adecisions
global made by Our
formation
formation coupled
Our previous study
oscillators,
previous study whichsuggested
suggested that synchronization
intensively
that investigated
synchronization in
in
pattern
pattern is
is achieved
achieved as
as a
a result
result of
of the
the decisions
decisions made
made by
by coupled
coupled
in the oscillators,
oscillators,
eld of nonlinearwhich
which has
has
physics, been
beencan intensively
intensively
be used investigated
investigated
for control of
each vehicle,
pattern is i.e., there
achieved as a isresult
no centralized
of the controller
decisions made that
by coupled oscillators, which has been intensively investigated
each
each vehicle,
vehicle, i.e.,
i.e., there
there is
is no
no centralized
centralized controller
controller that
that in
in the
the
mobile eld
eld
robot of
of nonlinear
nonlinear
circular physics,
physics,
formations can
can
[Hara be
be used
used
et al. for
for control
control
(2013)]. Thisof
of
supervises
each vehicle,all the
i.e., vehicles. in the eld of nonlinear
there is no centralized controller that mobile robot circular formations [Hara et al. (2013)]. This physics, can be used for control of
supervises
supervises all all the
all the vehicles.
the vehicles.
vehicles. mobile
suggestion robot wascircular formations
experimentally [Hara
veried et al.
with (2013)].
two-wheeledThis
supervises mobile robot circular formations [Hara et al. (2013)]. This
In our previous study [Yamada et al. (2011)], we focused suggestion suggestion was
was experimentally
experimentally veried
veried with
with two-wheeled
In
In our
our previous
previous study
study [Yamada
[Yamada et
et al.
al. (2011)],
(2011)], we
we focused
focused
mobile robots
suggestion
mobile robotswas[Tsukiji
[Tsukiji
et al. (2014)].
experimentally
et al. veried
(2014)]. with two-wheeled
Unfortunately,
Unfortunately,
these
two-wheeled
these
on
In circular
our previousformations for
study [Yamada multiple mobile
et al.mobile
(2011)], robots based
we focused mobile
our robots
previous [Tsukiji
studies et
have al. (2014)].
major Unfortunately,
drawbacks: these
on circular
on circular
circular formations
formations for multiple
for multiple
multiple mobile robots
robots based
based mobile robots [Tsukiji et al. (2014)]. Unfortunately, these
on a controlformations
law we developedfor for amobile
single robots
robot based[Hara our our previous
previous studies
studies have
have major
major drawbacks:
drawbacks:
on
on a
a control
control law
law we
we developed
developed for
for a
a single
single robot
robot [Hara
[Hara our
(i) previous
It is studies
dicult to have
analyze major
the drawbacks:
stability of formations;
et al.
on (2010)].law
a control Circular formations
we developed ofa mobile
forof single robotrobots[Hara have
et al. (2010)]. Circular
Circular formations of mobile
mobile robots have (i)
(i) It
It is radius
dicult to analyze the stability of formations;
et
some
et al. possible
al. (2010)].applications
(2010)]. Circular formations
for enclosing/capturing
formations of mobile robots
robots have (ii)
a target
have (ii) It is
(i) The
The
dicult
dicultof
is radius of
to
to analyze
circular
analyze
circular
the
the stability
formations
formationsstability of
of formations;
cannot
cannot
be specied.
formations;
be
some
some
object,
some
possible
possible
and
possible
applications
applications
collecting
applications
for
for
environmentalenclosing/capturing
enclosing/capturing
data
for enclosing/capturing [Dunbabin a
a target
target
a targetand (ii)
(ii) The
The radius
radius of
of circular
circular formations
formations cannot
cannot be specied.
be specied.
specied.
object,
object, and
and collecting
collecting environmental
environmental data
data [Dunbabin
[Dunbabin and
and The drawback (i) represents that we cannot design param-
Marquesand
object, (2012)]. The control
collecting environmental law [Hara dataet [Dunbabin
al. (2010)] and was The The
eters drawback
drawback
for desired (i)
(i) represents
represents
formations. that
that
The we
we cannot
cannot
drawback design
design
(ii) impliesparam-
param-
that
Marques
designed (2012)].
Marques (2012)]. The
The control
using a nonlinear control law
law [Hara
system [Hara et
et al.
with limit (2010)]
al. cycle,
(2010)] and was
wasa The eters drawback
for desired (i)formations.
represents The thatdrawback
we cannot(ii) design
impliesparam-
that
Marques
designed (2012)].
using a The control
nonlinear law
system [Hara
with et
limit al. (2010)]
cycle, and wasa eters
all
eters for desired
robots
for desired
run onformations.
a common
formations. The
The drawback
circle path;
drawback (ii)other
in
(ii) implies
implies that
words,
that
designed
formationusing
designed using
controla nonlinear
a nonlinear
law was obtainedsystem with
system with
basedlimit
limit cycle,
oncycle, and aa all
this result.
and all robots
robots run
run on
on a
a common
common circle
circle path;
path; in
in other
other words,
words,
formation
formation control
control law
law waswas
was obtained
obtained based
based on on
on this
this result.
result. we we
all cannot
robots avoid
run on collision
a common of robots.
circle path; in other words,
Although this
formation controlmethod
law wasobtained
demonstratedbased by thisnumerical
result. we cannot
cannot avoid
avoid collision
collision of
of robots.
robots.
Although
simulation this
Although and method
this method
experiment, was
was the demonstrated
demonstrated
approach adopted by
by numerical there we
numerical Thecannot
main avoidpurpose collision
of theofpresent
robots. paper is to overcome
Although
simulation this
and method
experiment, was demonstrated
the approach by
adopted numerical
there The main purpose of the present paper is to overcome
simulation and experiment, the approach
simulation and experiment, the approach adopted there The adopted there The main
the drawbacks purpose
main purpose of
(i) (ii). the
of the present
Wepresent
show that paper
paperthe is to
to overcome
is key idea for
overcome
the
the drawbacks
drawbacks (i)
(i) (ii).
(ii). We
We show
show that
that the
the key
key idea
idea for
This research was partially supported by JSPS KAKENHI

overcoming
the drawbacks
overcoming
them
them (i) is
is
to
(ii).
to
use
We
use
only
show
only
the
the
polar
that
polar thecoordinates
key
coordinatesidea for
for
for
This
This research
research
(26289131).
was
was partially
partially supported
supported by
by JSPS
JSPS KAKENHI
KAKENHI overcoming
stability
overcoming them
analysis,
them is
is to
while
to use
useouronly
only the
previous
the polar
polar coordinates
studies used both
coordinates for
for
This research was partially supported by JSPS KAKENHI stability analysis, while our previous studies used both
(26289131).
(26289131).
(26289131).
stability
stability analysis,
analysis, while
while our
our previous
previous studies
studies used
used both
both
Copyright
2405-8963 2016,
2016 IFAC 187Hosting by Elsevier Ltd. All rights reserved.
IFAC (International Federation of Automatic Control)
Copyright
Peer review
Copyright 2016
2016 IFAC
IFAC 187
187
Copyright under
2016 responsibility
IFAC of International Federation of Automatic
187Control.
10.1016/j.ifacol.2016.10.394
2016 IFAC NECSYS
188
September 8-9, 2016. Tokyo, Japan Tadashi Nakamura et al. / IFAC-PapersOnLine 49-22 (2016) 187191
Fig. 1. Two-wheeled mobile robots.
polar and rectangular coordinates. This idea drastically
simplies the stability analysis; accordingly, we can pro-
vide a procedure for designing the parameters, and can
specify the radius. These analytical results are veried on
some numerical simulations.
2. CONTROL LAW FOR CIRCULAR FORMATIONS
Let us consider two-wheeled mobile robots sketched in
Fig. 1. We assume the kinematic model for the robots,
and the dynamics of robot i {1, . . . , N } is described by
[ ]
ri cos i 0 [ ]
ri i = sin i 0 vi , (1)
i vi = v i := kv {f (ri , ri ) cos i + ri g(i ) sin i } , (3a)
i 0 1
where ri > 0 denotes the distance between the robot i and
the origin. i R and i R are the angular velocity of
the robot i around the origin and the angle between the
heading of the robot i and its radial direction, respectively.
The angle between the radial direction of the robot i and
that of i + 1 is denoted by i R.
The robot i is controlled by the following inputs: the
heading-direction component of velocity (vi R) and
the angular velocity around the center of robot i (i
R). These control inputs are given by driving the two
wheels with appropriate rotational velocities. This paper
proposes a control law of these inputs which induce circular
formations. It should be emphasized that our previous
studies [Hara et al. (2013); Tsukiji et al. (2014)] employed
both of the polar coordinates and rectangular coordinates
to express the dynamics of the robots, while the present
paper employs only the polar coordinates.
Now we turn our attention to the reference dynamics of
radical direction and phases,
( )
ri 2
ri = f (ri , ri ) := ari 1 2 , (2a)
ri
i = g (i ) := + sin i . (2b)
This paper utilizes these reference dynamics to establish
the control law. Note that the dynamics (2a) depends only
on the i-th distance ri . On the other hand, the dynamics
(2b) depends on the i-th and (i + 1)-th phases, and is
one of the simplest models of well-known coupled phase
oscillators [Acebron et al. (2005); Scardovi et al. (2007)].
188
Fig. 2. Illustration of vector eld of dynamics of (2a), (2b)
and robot i.
Remark that N denotes the angle between the radial
direction of robot 1 and that of N ; thus, the dynamics (2b)
represents one-way ring coupled phase oscillators. ri > 0
denotes the stable equilibrium point of the dynamics (2a).
The parameters, a R and R, express the convergence
rate for ri and the angular velocity around the origin.
R is the coupling strength.
We establish the control law for vi and i such that the
two-wheeled mobile robots (1) behave in the reference
dynamics. In Fig. 2, the vector eld of the reference is
written over the robot i. The reference velocities in radial
direction (i.e., ri ) and in its vertical direction (i.e., ri i ) are
given by the reference dynamics (2). This paper proposes
the following control law: the velocity in heading direction
vi and the angular velocity i are respectively set to
be proportional to the reference velocities in the radial
direction and in the vertical direction,
i = i := k {ri g(i ) cos i f (ri , ri ) sin i } , (3b)
where kv , k R are feedback gains. In real situations, it is
dicult to realize large velocities vi and i due to practical
limitation of the wheel rotational velocity. For such cases,
kv and k have to be small to reduce the velocities.
Now let us restrict our attention to a situation where
the dynamics of robot i, which is described by Eq. (1),
is controlled by the law (3). In order for the robot i to
be autonomously controlled, the robot i has to get the
following three real-time measurements used in the control
law (3) (see Fig. 1).
(a) ri : distance to the origin.
(b) i : angle to the heading direction.
(c) i : angle to the robot i + 1.
While the requirement of these measurements comes from
the form of the control law we adopted, it is also feasible
in practice. This paper supposes that every robot has
sensors for measuring the distance to the origin ri , distance
to the next robot i + 1, the angle between its radial
and the heading directions, and the angle between its
radial and (i + 1)-th robot directions. Each robot gets
the measurements (a) and (b), but cannot (c) directly. It
is easy to see that the cosine formula allows the robot i
to compute (c) in real-time by using the measured data,
ri , the distance to the next robot i + 1, and the angle
between its own radial direction and the direction to the
robot i + 1. In consequence, our robots can be considered
as fully autonomous robots.
 2016 IFAC NECSYS
September 8-9, 2016. Tokyo, Japan Tadashi Nakamura et al. / IFAC-PapersOnLine 49-22 (2016) 187191 189

3. STABILITY ANALYSIS AND DESIGN

Lemma 1. The linear system (8) is stable if and only
This section will analyze the stability of circular forma- if both { ( )}
tions consisting of the N robots (1) with the control law l
(3), and will provide a procedure for designing parameters. k + sin 2 > 0, (15)
N
The dynamics of the N robots (1) with the control law (3) and
is described by the state space model with i := i+1 akv > 0, (16)
i , are satised.
{
ri = v i cos i
, (i = 1, . . . , N ) (4a)
i = i Proof. The stability of the linear system (8) is governed
1 1 by the characteristic polynomial,
i = v i+1 sin i+1 v i sin i , (4b)
ri+1 ri N

(i = 1, . . . , N 1). DA (s) := d(s, ri ), (17)
The equilibrium points of the state space model (4) are i=1
given by where d(s, ri ) is the characteristic polynomial of Ai ,
l ( ) ( )
ri = ri , i = , i = 2 , (5) l l
2 N d(s, ri ) := s2 +k ri g 2 s+2akv k ri g 2 . (18)
which correspond to a situation where the robot i N N
{1, . . . , N } runs on a circle with radius ri and all the As we have ri > 0, i {1, . . . , N }, the necessary and
robots keep the same angle i = 2 Nl . Here l {0, . . . , N } sucient conditions for the stability of d(s, ri ) are reduced
represents a type of circular formation patterns. Although to the conditions, (15) and (16): these conditions do not
some other equilibrium points coexist with points (5), the depend on the number i. 
present paper focuses only on points (5). Note that Eq.
(4b) does not include the dynamics of N . The reason Lemma 1 indicates that each robot can run on its own
is that, for a formation pattern l, the angles i for i = circle in a stable manner if and only if Eqs. (15) and (16)
1, . . . , N always satisfy hold. It should be noted that Lemma 1 cannot guarantee
N that all the robots run with the same angle 1 = 2 =

i = 2l. (6) = N .
i=1
This fact states that the dynamics of N is not needed Lemma 2. The linear system (9) is stable if and only
to be considered, since N is determined from the others if ( )
1 , . . . , N 1 . l
kv cos 2 > 0, (19)
In order to analyze the stability of the equilibrium points N
(5), we introduce errors, is satised.
l
ri := ri ri , i := i , i := i 2 . (7)
2 N Proof. The stability of the linear system (9) is governed
The linearized error dynamics around the equilibrium by the characteristic polynomial,
points (5) is described by { ( )}N 1
l
x = Ax, (8) DH (s) := s + kv cos 2 . (20)
N
y = Hy, (9)
where the state variables are It is clear that the stability of DH (s) can be described by
[ ]T T condition (19). 
x := xT1 , . . . , xTN , xi := [ri , i ] , (10)
T Lemma 2 indicates that all the robot can run with the
y := [1 , . . . , N 1 ] . (11)
and the system matrices are dened by same angle 1 = 2 = = N = 2l
N if and only if Eq.
(19) holds. From these lemmas, we can easily derive the
A := diag (A1 , . . . , AN ) , (12) following condition.
( )
l
0 kv ri g 2 N
Ai := ( ) , (13) Theorem 3. Formation pattern l (corresponding equi-
l librium point) is locally stable if and only if all of the
2ak k ri g 2
N conditions, (15) and (16) in Lemma 1 and condition
(19) in Lemma 2, hold.
1 1 0 0
( ) 0 1 1 0
l
H := 2kv cos 2 . . . . . . (14)
N .. .. .. . . .. Proof. From our arguments mentioned above, it is obvi-
0 0 0 1 ous that formation pattern l corresponding to the equi-
librium point (5) is locally stable if and only if both of
The necessary and sucient conditions for the linear the linear systems (8) (Lemma 1) and (9) (Lemma 2) are
systems, (8) and (9), to be stable are given below. stable. 

189
2016 IFAC NECSYS
190
September 8-9, 2016. Tokyo, Japan Tadashi Nakamura et al. / IFAC-PapersOnLine 49-22 (2016) 187191

Table 1. Conditions for stabilizing control

law (3).
1 1,2,3,4,5
l
[ 1
) l
(1 1
] l
[1 3
) l
(3 ] 4
0, , , ,1
N 4 N 4 2 N 2 4 N 4 3
k > 0
0 >0 <0
2 5
kv > 0 kv < 0 kv > 0 kv < 0
a>0 a<0 a>0 a<0

Even though this theorem provides the necessary and

sucient condition for formation pattern l to be stable, it
is not easy to design the parameters in the control law (3).
Now, we show an easy way to design them in the following
corollary.

Corollary 4. Let total number of robots (N ), forma-

tion pattern (l), and angular velocity () be given.
If the parameters in the control law (3) (i.e., , a,
kv , k ) are designed in accordance with Table 1, then Fig. 3. Time series data of 1,...,5 (t) with = 0.5 for
formation pattern l is locally stable t [0, 100) and = 0.5 for t [100, 200].
come the drawbacks: (i) diculty of analyzing formation
Proof. The condition (15) holds if and only if either stability: (ii) impossibility of controlling radius of forma-
( ) tions.
l
> sin 2 for k > 0 (21) First, let us make formation patterns l = 2, 5 for ve robots
N
or (N = 5) with angular velocity = +1 and radius r1,...,5 =
( )
l 60.0. These formations l = 2, 5 represent the following
< sin 2 for k < 0 (22) situations: all the robots run respectively with the angular
N
dierences, 1,...,5 = 2/5 and 1,...,5 = 0, on a circle in a
is satised. Further, the condition (19) holds if and only if counterclockwise direction. Based on Corollary 4, we can
either [ ) ( ] design the parameters , a, kv , k ,
l 1 3
0, , 0 for kv > 0 (23) > 0, a < 0, kv < 0, k > 0 for l = 2,
N 4 4 (25)
< 0, a < 0, kv < 0, k > 0 for l = 5.
or ( )
l 1 3 Equation (25) implies that formation l = 2 can be switched
, for kv < 0 (24) to l = 5 only by changing the sign of from positive to
N 4 4
negative. The parameters except for are xed to
is satised. These inequalities and the condition (16) can
a = 0.3, kv = 0.5, k = 0.5,
be summarized in Table 1 with careful consideration of
signs of the parameters. Remark that Table 1 indicates in accordance with Eq. (25). Figure 3 shows time series
sucient conditions for the inequalities (15), (16), and (19) data of 1,...,5 (t). For t [0, 100), it is observed that the
to be satised.  robots with = 0.5 converge to the formation l = 2. At
t = 100, is changed from 0.5 to 0.5. After that, the
This corollary provides us only information about the sign robots converge to the formation l = 5, which is equivalent
of parameters. As a consequence, it is easy to see that if to the formation l = 0. As can be seen from Fig. 3,
the absolute values of these parameters are set to be large, our stability analysis and design procedure for overcoming
then we obtain fast-convergence for creating formations. drawback (i) lead to these successful results. It should be
However, such large parameters require large control sig- emphasized that the formation l = 5 (i.e., l = 0) cannot
nals. Note that other set of signs that satisfy (15), (16), be experimentally realized owing to collision of robots (i.e.,
and (19), dierent from those in Table 1, can be also drawback (ii)).
possible; the resulting control performance dierence has
Second, we make formation l = 0 for ve robots (N = 5)
not been fully investigated. Furthermore, we should pay
with angular velocity = +1 and dierent radii
attention to the fact that the control parameters designed
by this corollary may cause the coexistence of multiple r1 = 10.0, r2 = 30.0, r3 = 50.0, r4 = 70.0, r5 = 80.0.
formation patterns. One of the coexisting formations ap- The robots must run with the same phase 1,...,5 = 0
pears depending on initial conditions. Note that coexisting on ve circles in a counterclockwise direction. Corollary
in synchronization pattern has been observed in coupled 4 suggests that the parameters have to satisfy
phase oscillators. > 0, a > 0, kv > 0, k > 0. (26)
Then we x them to
4. NUMERICAL EXAMPLES
= 0.5, a = 0.3, kv = 0.5, k = 0.5.
This section will show some numerical examples to verify The trajectories of robots running from initial positions
that our stability analysis and design procedure can over- () at t = 0 to steady positions () at t = 200 are shown in

190
2016 IFAC NECSYS

September 8-9, 2016. Tokyo, Japan Tadashi Nakamura et al. / IFAC-PapersOnLine 49-22 (2016) 187191 191

Hara, N., Takarada, Y., and Konishi, K. (2013). Coupled

nonlinear oscillators with application to mobile robot
formation control. Proc. of SICE Annual Conference,
21742177.
5 In, V., Longhini, P., and Palacios, A. (2009). Applications
of Nonlinear Dynamics. Springer.
4 4
Kumar, V., Leonard, N., and Morse, A. (2005). Coopera-
3 tive Control. Springer.
5 2
1 Mesbahi, M. and Egerstedt, M. (2010). Graph Theoretic
Methods in Multiagent Networks. Princeton University
3
2 Press.
1
Okuda, T., Konishi, K., and Hara, N. (2011). Experi-
mental verification of synchronization in pulse-coupled
oscillators with a refractory period and frequency dis-
tribution. Chaos, 21(2), 023105.
Pikovsky, A., Rosenblum, M., and Kurths, J. (2001).
Synchronization. Cambridge University Press.
Scardovi, L., Sarlette, A., and Sepulchre, R. (2007). Syn-
chronization and balancing on the N-torus. Systems &
Control Letters, 56(5), 335341.
Fig. 4. Trajectories of robots running from initial positions Strogatz, S.H. (2003). Sync : the emerging science of
at t = 0 to positions at t = 200.0. spontaneous order. Hyperion Books, New York.
Fig. 4. Each robot converges to its own circle with the same Tsukiji, M., Hara, N., and Konishi, K. (2014). Experi-
phase. This result indicates that we successfully overcome mental evaluation of formation control of mobile robot
the drawback (ii). group by coupled oscillators. Proc. of SICE Annual
Conference, 301306.
5. CONCLUSION Yamada, K., Hara, N., and Konishi, K. (2011). Circular
motion generation for mobile robots using limit cycle
This paper analyzed stability of mobile robot circular systems application to a circular formation control.
formations to overcome the drawbacks of our previous Proc. of SICE Annual Conference, 342345.
studies. The stability analysis allowed us to provide a Zhou, C. and Low, K. (2012). Design and locomotion
simple procedure for design a control law. This law makes control of a biomimetic underwater vehicle with fin
a desired formation stable. Some numerical examples ver- propulsion. IEEE/ASME Transactions on Mechatron-
ified our analytical results. ics, 17(1), 2535.

ACKNOWLEDGMENT

The present study was supported in part by JSPS KAK-

ENHI (26289131).

REFERENCES
Acebron, J.A., Bonilla, L.L., Perez Vicente, C.J., Ritort,
F., and Spigler, R. (2005). The Kuramoto model: A sim-
ple paradigm for synchronization phenomena. Reviews
of Modern Physics, 77(1), 137185.
Arenas, A., Diaz-Guilera, A., Kurths, J., Moreno, Y., and
Zhou, C. (2008). Synchronization in complex networks.
Physics Reports, 469(3), 93153.
Boccaletti, S., Latora, V., Moreno, Y., Chavez, M., and
Hwang, D.U. (2006). Complex networks: Structure and
dynamics. Physics Reports, 424(4-5), 175308.
Dunbabin, M. and Marques, L. (2012). Robotics for en-
vironmental monitoring. IEEE Robotics & Automation
Magazine, 19(1), 2439.
Fukunaga, T., Imasaka, T., Ito, A., Sugitani, Y., Konishi,
K., and Hara, N. (2016). Dynamical behavior and peak
power reduction in a pair of energy storage oscillators
coupled by delayed power price. Physical Review E,
93(2), 022220.
Hara, N., Kokame, H., and Konishi, K. (2010). Circular
periodic motion generation for mobile robots using limit
cycle systems. Proc. of the 2010 American Control
Conference, 42714276.

191