TELKOM
NIKA
, Vol.14, No
.4, Dece
mbe
r
2016, pp. 13
97~140
7
ISSN: 1693-6
930,
accredited
A
by DIKTI, De
cree No: 58/DIK
T
I/Kep/2013
DOI
:
10.12928/TELKOMNIKA.v14i4.3056
1397
Re
cei
v
ed
No
vem
ber 1
7
, 2015; Re
vi
sed
Jun
e
2, 2016;
Accept
ed Ju
ne 16, 201
6
A Practical Coordinated Trajectory Tracking for A
Group of Mixed Wheeled Mobile Robots with
Communication Delays
Sisdarmanto
Adinandr
a*, D
w
i
Ana
Ratna
w
a
t
i
Dept. of Electri
c
al Eng
i
ne
eri
n
g,F
a
cult
y
of I
ndustria
l T
e
chnolo
g
y
, U
n
ivers
i
tas Islam Indon
esia,
Gedung K.H. Mas Mansur, K
a
mpus T
e
r
padu UII, Jl. Kaliurang Km 14,
5 Sleman,
Yog
y
akarta
55
584, Ind
ones
ia
*Corres
p
o
ndi
n
g
author, e-ma
i
l
: s.adina
ndr
a
@
uii.ac.i
d
A
b
st
r
a
ct
Coor
din
a
tio
n
b
e
tw
een a speci
f
ic mob
ile ro
bo
t ty
pe has bee
n w
i
dely inv
e
sti
gated, e.g co
or
din
a
tion
betw
een
unicy
cles. To exte
n
d
the
ap
p
lic
abi
lity of the
system,
a co
ordi
n
a
t
ed traj
ectory
tracking
of
mi
xe
d
type of mobi
le
robots is con
s
ider
ed.
W
e
pr
ove that if a c
e
rtain typ
e
of
w
heele
d
mob
i
l
e
robot is a
b
l
e
to
indiv
i
d
ual
ly tra
ck its ow
n refer
ence, th
en co
o
r
din
a
tion
in tra
cking w
i
th ot
he
r type of ro
bots
can b
e
ac
hiev
ed
simply
by s
har
ing
in
divi
du
al t
r
ackin
g
err
o
rs. Usi
ng
tw
o typ
e
s of w
h
eel
ed
mob
ile
ro
bots, na
mely
un
icy
c
l
e
types (a non
h
o
lo
no
mic
mo
bi
le robot) an
d omni w
hee
ls
type (a hol
on
o
m
ic
mo
bil
e
ro
bot), a coordi
n
a
ted
control
algorithm
can ac
hiev
e
a global asy
m
ptotically st
able
condition
of the error dynam
ic
s of the systems.
Und
e
r bid
i
recti
ona
l co
mmun
ic
ation b
e
tw
een
robots as a co
nstraint, the gr
oup is a
b
l
e
to ma
inta
in in
divi
dua
l
tracking
w
h
ile
coord
i
nati
ng
t
h
e move
ments w
i
th
other
r
o
b
o
ts reg
a
rdl
e
ss
occurri
ng
pert
u
rbati
ons
in th
e
system and d
e
lays in
co
mmu
n
ic
ation
c
h
ann
els.
Si
mu
l
a
tion r
e
sults
sugg
est that i
n
formatio
n
sh
arin
g
betw
een th
e ro
bots incr
eas
e t
he ro
bustn
ess
in co
ordi
nati
n
g
ind
i
vid
u
a
l
traj
e
c
tories. Res
u
lt
s also
show
th
at
del
ays caus
e d
r
op in p
e
rfor
ma
nce si
mil
a
r to the case of n
o
i
n
formatio
n
sha
r
ing.
Ke
y
w
ords
: co
ordi
nated tra
j
e
c
tory tracking, unicycl
es, o
m
n
i
w
heels rob
o
ts, Lyapun
ov fun
c
tion, time del
a
ys
Copy
right
©
2016 Un
ive
r
sita
s Ah
mad
Dah
l
an
. All rig
h
t
s r
ese
rved
.
1. Introduc
tion
The a
b
ility of a g
r
oup
of
whe
e
led
mo
bile robot
s in
coo
r
di
nating
com
p
lex tasks ha
s
dra
w
n attenti
on from scie
ntist, hobbyist and pra
c
ti
tioner. A gro
u
p
of mobile robots i
s
able
to
handl
e more
difficult and di
stribute
d
tasks. The gr
oup
can a
c
hieve
a certai
n form
ation as well as
coo
r
din
a
ting
movement be
tween the rob
o
ts so that ce
rtain tasks ca
n be com
p
let
ed.
One p
r
o
b
lem
in coordinati
on bet
wee
n
robots
i
s
th
e
probl
em of
coordi
nating i
n
dividual
trajecto
rie
s
so that the ov
erall traje
c
tori
es fo
rm
a
c
e
rtain s
p
atial patterns [1-7].
In this
particular
ca
se, a grou
p of unicycle
mobile ro
bots is coordi
nati
ng its memb
er’s traj
ecto
ri
es su
ch that
the
coo
r
din
a
tion
forms
sp
atial
pattern
s. T
he coordi
nati
on is
achiev
ed by excha
nging i
ndivid
ual
positio
n or tra
cki
ng erro
rs
within the gro
up.
Due to
the
no
nholom
omi
c
con
s
trai
nts th
at exist
in th
e
system
s, u
n
i
c
ycle
mobil
e
robots is
less flexible
comp
ared to
holon
omi
c
mobile
ro
b
o
ts, e.g. om
ni
whe
e
ls mo
bile robot
s. In a
compact
space, the ability of omni wheels to
move sideway
s
becom
e
s very important
Omni
whe
e
ls mobile
ro
b
o
ts h
a
ve be
en wi
dely u
s
ed
in footb
a
ll co
mpetiti
on [8-9].
Although it easily suffers
from slip, a
good
cont
rol
and mecha
n
ical d
e
sig
n
can o
p
timize
the
ability of an
omni
wheels mobile
robot. Example of trajectory
tracking or path followi
ng f
o
r
omniwheel m
obile ro
bots
can be
foun
d for exampl
e in
[10-12].
In [10], a
kin
e
matic
co
ntrol is built fo
r the om
nidire
ctional
ro
bot. A detaile
d
kinematic
model is imp
l
emented to
solve traje
c
to
ry trac
king p
r
oblem. The controlle
r take
s into accou
n
t
kinem
atic m
o
del of the
wh
eels to
redu
ce slip
e
ffect
s. In [11], a
co
mbination
of
kinem
atic
an
d
torque
contro
l is used to track re
fere
nce trajecto
rie
s
, while in [12], an omnidire
ctional robot
is
use
d
to help
patients during reh
abilita
t
ion pr
o
c
e
ss.
All algorithms are d
e
si
gned for a si
ngle
robot.
From the re
sults in [1-3] coordinatio
n
bet
wee
n
ro
bots can be
achieve
d
si
mply via
excha
nging
states of the robots,
i.e. the individual tracking e
r
ro
rs. The stability proof sugg
e
s
ts
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 16
93-6
9
30
TELKOM
NIKA
Vol. 14, No. 4, Dece
mb
er 201
6 : 1397 – 140
7
1398
that prop
er
choice of cont
ro
l st
ru
cture
can
simplify
the mathem
a
t
ical procedu
re, a
s
well a
s
extending the
system with
different type of robot.
On the othe
r hand, stu
d
ie
s of coordinati
ng traj
e
c
tori
e
s
of omni wh
eels in a
sim
p
le way
are
rare
subj
ects. F
r
om th
e re
sult
s in u
n
icycl
e
ca
se, it can b
e
said
that applying
simila
r con
c
ept
for om
ni wh
eels
mobil
e
robots will
be
a valua
b
le
result. Motivat
i
ng by the
fa
cts, thi
s
pa
p
e
r
addresse
s th
e problem
of coo
r
di
n
a
ting
a group
of omni wheel
mobile
rob
o
ts by me
an
s
of
excha
nging
i
ndividual t
r
a
c
king
erro
rs.
Furthe
rm
o
r
e,
the g
r
oup
is extended
to
be a
mixed
of
unicy
cle an
d
omni wh
eel
mobile ro
bo
ts. The st
abi
lity of the system is inv
e
stigate
d
usi
ng
Lyapun
ov theorem. Th
e al
gorithm i
s
st
ructured
su
ch
that delays in
comm
unication chann
el can
be ea
sily com
pen
sated.
The contrib
u
tions of thi
s
pape
r are a
s
follow:
i
)
a
trajecto
ry tracking
cont
roller for
unicy
cle an
d omni wheel
s
that able
to a
c
hieve
coo
r
di
nation with ei
t
her uni
cycl
es or omni
whe
e
ls
or both type
s in the pre
s
e
n
ce of del
ays in co
mmun
i
cation
cha
n
n
e
l iii) globally
asymptotical
ly
stable
erro
r dynamics of
the ove
r
all
mixed
whe
e
led m
obile
robot syste
m
s,
iv) simula
tion
validation of the overall
systems.
The re
st of the pap
er i
s
o
r
gani
ze
d as f
o
llo
ws. Secti
o
n 2 gives th
e kine
matic
model of
unicy
cle
s
and
omni wh
eel
s as well a
s
the
o
rie
s
for
stabi
lity. Section 3 pre
s
ent
s the
control de
sig
n
pro
c
e
ss;
start with the de
sign of
a traj
ectory trackin
g
for a si
ngl
e omni whee
ls mobil
e
ro
b
o
t,
followe
d by t
he exten
s
io
n
to
m
-mixed-unicy
cle-omni
wheel
s
m
o
bile robo
t syst
ems.
Stabilit
y
analysi
s
i
s
p
r
ese
n
ted i
n
thi
s
cha
p
ter. Se
ction
4 give
s
the sim
u
latio
n
validation
a
nd p
e
rfo
r
man
c
e
analysi
s
of th
e mixed
syst
ems
both d
e
l
a
yed an
d no
n
-
delaye
d
com
m
unicaton
ch
annel. Fin
a
lly, in
se
ction 5 th
e co
ncl
u
si
on
s of this work a
r
e give
n
and
sug
g
e
s
tions fo
r furth
e
r resea
r
ch
are
pre
s
ente
d
.
2. Kinematic
Model of Mixed Wh
eeled
Mobile Rob
o
ts
In this resea
r
ch
we con
s
i
der two type
s of mobile robots, nam
el
y unicycle type and
omniwheel
s
type. Unicyl
cle rep
r
e
s
ent
s a mobile
robot with n
o
n
-hol
ono
m
ic
con
s
trai
nt, i.e
.
pra
c
tically
ca
nnot move
si
deways. Om
ni wh
eel
ro
b
o
t rep
r
e
s
ent
s
mobile
ro
bot
that belo
ng
s to
holon
omic m
obile ro
bots.
The kin
e
mati
c model
s
of e
a
ch robot a
r
e
illustrated in
Figure 1.
Figure 1. Kinematic Mo
del
of an Unicy
c
l
e
(left); Omni
Whe
e
ls (rig
ht)
The kin
e
mati
c model of an
unicycl
e is gi
ven as follo
ws:
cos
0
sin
0
01
(
1
)
With
and
are the forward
and stee
rin
g
velocitie
s
of
the ro
bots, wh
ile the kine
m
a
tic mod
e
l of
an omni whe
e
ls is:
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
1693-6
930
A Practical Coordi
nated T
r
ajecto
ry T
r
a
c
king fo
r A G
r
oup of Mixe
d
…
(Sisd
a
rm
a
n
to Adinand
ra)
1399
∅
cos
cos
1
sin
s
i
n
0
1/
1
/
1/
(
2
)
Whe
r
e
,
and
are the indivi
dual wh
eel speed that
dri
v
e the robot and are determined by
the multiplicat
ion of whe
e
l rotation and ra
dii.
It is to be no
ted that the two mo
del
s u
s
e
different i
nput state
s
.
At the one h
and, the
unicy
cle u
s
e
s
forwa
r
d an
d stee
ring ve
locitie
s
a
s
in
puts. On th
e
other h
and,
the omni
wh
eels
use
s
i
ndividu
al wheel
spee
ds
as inp
u
ts.
Athough
see
m
contra
dicti
v
e, the
choi
ces
help
in
pro
v
ing
the con
c
ept t
hat coordinati
on
c
an
be
achieved o
n
ly b
y
excha
nging
individual
st
ate informati
on
rega
rdl
e
ss th
e model of the robot
s.
3. Contr
o
l Design
In this
se
ctio
n the
cont
rol
desi
gn i
s
ex
plai
ne
d in
de
tail. For the
delay-free
ca
se, the
controlle
rs for the pure uni
cycle
s
in the
grou
p are obt
ained fro
m
[1-2].
If referen
c
e
s
trackin
g
for a
grou
p of mixed whe
e
led m
obile ro
bots,
i
is etiher an
unicy
cle
or omni whe
e
ls, are given as
,
,…,
and the actual position
s
are
given as
,
,…,
, and the tracking e
rro
rs of the group are
,
,…,
,
,
….,
, the problem
of coordinatin
g the trajecto
ry is defined a
s
to make:
→
as
→
0
(
3
)
3.1. Coordin
a
tion
w
i
th
ou
t Dela
y
s
in Communicati
on Chan
nels
The problem
if minimizing trackin
g
errors of
the grou
p can
solved
in the followin
g
step
s.
Firs
tly for the unic
y
c
l
e mobile robot, the trajec
tory
tracking
cont
ro
llers i
s
obtai
n
ed from [2], i.e.
for each uni
cycle
i,
the con
t
rollers are:
cos
∑
,
∑
,
(4)
Whe
r
e
,
,
are
control gai
ns for individu
al trackin
g
,
,
are the
cou
p
ling gai
ns,
,
are referen
c
e forwa
r
d an
d steeri
ng velocitie
s
of each unicycl
es,
are
individual tra
cki
ng errors,
are individua
l trackin
g
erro
rs from othe
r robot
s that can be
either an u
n
icycle or om
ni whe
e
ls.
As for th
e om
ni wh
eel
s typ
e
, by re
writin
g (2
) into
,
wher
e
,
the
prop
osed traj
ectory tra
c
kin
g
controlle
rs
are given by:
∑
,
∑
,
(5)
Whe
r
e
is the
individual tra
cki
ng control gain vecto
r
,
,
are the
cou
p
li
ng gain
s
,
is
the
referen
c
e traj
ectory vecto
r
of each omni
whe
e
ls,
are individual tra
c
king e
rro
rs
(omni wh
eel
s)
are individ
ual tra
c
king e
rro
rs from
other
rob
o
ts th
at can
be eit
her a
n
unicy
cle or o
m
ni whe
e
ls.
For th
e given
controlle
rs
an
d coordinate
d
tracki
n
g
p
r
ob
lems fo
r a
g
r
o
up of mixe
d
whe
e
ld
mobile robot
s, the following
theorem h
o
ld
s.
Theorem
1.
There exi
s
t
m-mobil
e
rob
o
ts,
which
ca
n be
eith
er u
n
icycl
e
o
r
om
ni wheel
s
mobile robot
s. The rob
o
ts
are tra
c
king t
he given re
fe
ren
c
e
s
traje
c
t
o
rie
s
that in
overall
cre
a
te
a
spatial fo
rmat
ion patte
rn. If the co
ntrol
p
a
ram
e
ters a
r
e ch
osen
so t
hat and
k
k
,k
k
0
,
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 16
93-6
930
TELKOM
NIKA
Vol. 14, No. 4, Dece
mb
er 201
6 : 1397 – 140
7
1400
k
,k
,k
,k
,k
0
and it is assume
d that
the shared
informatio
n is received by
the
corre
s
p
ondin
g
robots with
out delay, th
en the cont
roller given in
(5) and (6
) rend
ers origi
n
of
,
,…,
globally asy
m
ptotically st
able (GAS
).
Proof
of
The
orem 1.
The
stability is proven usi
ng L
y
apunov fun
c
tions. To si
m
p
lify the
analysi
s
, the
system i
s
categori
z
e
d
into
three
sub
s
yst
e
ms. Th
e sub
s
ystem
s
represe
n
t the types
of robot
s exist in the syste
m
. Us
in
g the
theory that st
ates if a
sub
system is GAS
using
Lyapu
nov
function, the
n
the overall
system i
s
al
so GAS
if there i
s
no
switching in
systems, the error
dynamics a
r
e
analyze
d
. The sub
s
ystem
are a
s
follow:
1.
A subsy
s
tem S1, where
∀
,
∈
is a unicycl
e type. In this
con
d
ition the syst
em become
s
a
homog
eny u
n
icycl
e
sy
ste
m
s, which
is
simila
r
a
s
in
[2] and its e
r
ror
dynami
c
s is GAS. All
stability proof
follows the one given in [2]. Thus
, stabil
i
ty proof follows the
results in [2].
2
.
A s
u
bs
ys
te
m S2
, w
h
er
e
∀
,
∈
is a om
ni
whe
e
ls type.
In this cond
ition the sy
stem
become
s
a h
o
moge
ny om
ni whe
e
ls
system.
3.
A subsy
s
tem
S3, where
,
∈
is either uni
cy
cle or om
ni whe
e
ls., i.e. the ca
se of mixed
mobile robot
system
s.
a. Stability proof for S2
Con
s
id
er a Lyapunof funct
i
on
. For S2,
the errors are
all coming fro
m
omni
whe
e
ls mo
bil
e
robot
s. Thu
s
, the derivati
v
e of
can be
expresse
d a
s
:
∑
∑
∑
,
∑
,
0
∑
(6)
1
,
1
,
0
= 0
, becau
se
of the coupli
n
g gain choi
ce
s
(7)
Thus,
∑
0
(
8
)
Whi
c
h i
s
neg
ative definite.
This sho
w
s
that the
controller i
n
(6) re
nders th
e o
r
i
g
in of th
e e
r
ror
dynamics
of
sub
s
ystem
S
2
, i.e. a g
r
ou
p of om
ni
wh
eels mobil
e
robots,
globall
y
asymptotically
stable (GAS)
becau
se of the choi
ce of
,
0
.
b. Stability proof for S3
For S3, errors are
comin
g
from either unicy
cl
e or o
m
ni whe
e
ls. Suppo
se, tha
t
i
is a
unicy
cle and
j is an omni whe
e
ls. Co
nsider a Lyapu
nof function
V
e
e
.
In this
c
a
s
e
,
the
e
e
e
. Using the given controll
er in (5) and (6
), the derivative of
the Lyap
unov functio
n
contai
ns the
unicy
cle pa
rt and the omni
whe
e
ls p
a
rt. For the om
ni whe
e
ls, the result
s are
si
milar
to the results
in the
su
bsy
s
tem S2 d
ue t
o
the
sp
ecifi
c
stru
ctu
r
e of
the controller and
th
e coupl
ing
gain choi
ce
s.
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
1693-6
930
A Practical Coordi
nated T
r
ajecto
ry T
r
a
c
king fo
r A G
r
oup of Mixe
d
…
(Sisd
a
rm
a
n
to Adinand
ra)
1401
Similar condit
i
on ap
plies fo
r the u
n
icycl
e
part. As p
r
e
s
ente
d
in [2], the additio
n
of omni
whe
e
ls d
o
e
s
will not affect
the Lyapuno
v condition
d
ue to the spe
c
ific st
ru
cture
of the propo
se
d
controlle
r and
coupli
ng gai
n choi
ce
s. Th
us, the de
rivative of the Lyapun
ov functi
on is sim
p
ly:
0
(
9
)
Whi
c
h i
s
ne
g
a
tive definite. This im
plie
s
that t
he co
ntroller give
n in
(5)
and
(6
) re
nders th
e ori
g
in
of the erro
r d
y
namics of t
he mixed syste
m
globally asymptotically stable (GAS).
Since i
n
all
subsy
s
tems S
1
, S2, and
S
3
the
co
ntroll
ers in
(5
) a
n
d
(6
)
rend
er th
e ori
g
in
o
the error dyn
a
mics
GAS, it can be co
nclu
ded that the co
ntrolle
r in (5) an
d (6
) rend
ers the
compl
e
te mixed uni
cycle a
nd omni whe
e
ls mobil
e
ro
bot grou
p glo
bally asympto
t
ically stable.
3.2. Coordin
a
tion
w
i
th
Dela
y
s
in Communicatio
n Chann
e
ls
The se
co
nd
ca
se con
s
ide
r
s the p
r
e
s
en
ce of delay
s in comm
uni
cation ch
ann
el
. In this
ca
se the sh
a
r
ed me
ssage
s that is to be sha
r
ed i
s
delayed.
The sha
r
ed sign
al
that
is
origin
ally
,
∈
,
(for simpli
city is written as
as used in th
e previou
s
se
ction), is d
e
la
yed for
perio
de of ti
me. Thu
s
, th
e sh
ared info
rmation i
s
di
stributed
as
,
written a
s
,
for
∈
,
.
As con
s
e
que
nce
s
of the e
x
isting delay
s,
the controlle
rs a
r
e modifie
d
as follo
ws:
cos
∑
,
∑
,
(10
)
∑
,
∑
,
0
(11
)
It is to be
not
ed that the
problem fo
rmul
ation remain
s simila
r to th
e
ca
se
witho
u
t delays.
For th
e
given
co
ntroll
ers a
nd
coo
r
din
a
te
d tra
c
king
pro
b
lems for a
g
r
oup
of
mixed
wh
eeld
mo
bi
le
robot
s, the followin
g
theore
m
holds.
Theorem
2.
There exi
s
t
m-mobil
e
rob
o
ts,
which
ca
n be
eith
er u
n
icycl
e
o
r
om
ni wheel
s
mobile robot
s. The rob
o
ts
are tra
c
king t
he given re
fe
ren
c
e
s
traje
c
t
o
rie
s
that in
overall
cre
a
te
a
spatial fo
rmat
ion patte
rn. If the co
ntrol
p
a
ram
e
ters a
r
e ch
osen
so t
hat and
k
k
,k
k
0
,
k
,k
,k
,k
,k
0
and it is
assumed that th
e sh
are
d
inf
o
rmatio
n is
d
e
layed
uni
fo
rmly
,
then the cont
rolle
r given in (10) a
nd (11) re
nde
rs origin of
,
,…,
globally asy
m
ptotically st
able (GAS
).
Proof of
The
orem 2.
Con
s
ide
r
ing th
e delay-free ca
se, the mo
st
important p
a
rt is to
che
c
k the va
lue of
φ
. In the delay-free
ca
se,
φ0
in all possibl
e su
b
s
ystem. Sin
c
e delay
occurs in the
communi
cati
on cha
nnel b
e
twee
n the robots, only
φ
cha
nge
s, the rest (pa
r
t for
individual tra
c
king
) is si
mila
r, rega
rdle
ss
the type of the robot in the system.
Usi
ng S2, i.e. all omni wh
e
e
ls, as
ca
se
study, the an
alysis of
φ
is given as follo
ws. If
φ
in delayed system is denot
ed
φτ
. Using si
milar Lyapun
ov function
,
the derivative
of the functio
n
is given a
s
follows:
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 16
93-6
930
TELKOM
NIKA
Vol. 14, No. 4, Dece
mb
er 201
6 : 1397 – 140
7
1402
∑
∑
∑
,
∑
,
0
(12
)
∑
∑
,
∑
,
0
(
1
3
)
In the d
e
lay-f
r
ee
ca
se, th
e
value of e
qua
tion (1
3) is
ze
ro b
e
cau
s
e
of
the
stru
cture
of the
c
h
os
en
c
o
up
lin
g
p
a
r
ame
t
er
s
.
Simila
r
l
y,
for the delaye
d
case, the value of
0
. This, again,
is be
ca
use of
the stru
ctu
r
e
of the
co
ntro
ller. For
cla
r
ifi
c
ation, ta
ke
a
look fo
r the t
r
ackin
g
e
rro
r
in
x
dire
ction.
Define
1
, the
comp
one
nt of
in
x
dire
ction is given a
s
follows
:
∑
,
(
1
4
)
∑
,
0
Equation (14
)
is ze
ro be
cause
the
u
n
iform delay
in the
system
makes th
e
value of
or
in all robot remains th
e same. Com
b
ined wi
th the spe
c
ific co
nt
rol stru
cture, the
comp
ut
at
ion r
e
sult
s in ze
ro
.
Thus,
∑
0
(
1
5
)
Whi
c
h is n
e
g
a
tive definite and prove t
hat the traj
e
c
tory of the origin of the error dyn
a
mics is
globally a
s
ymptotically stab
le (GAS).
Usi
ng
similar approa
ch a
nd metho
dol
ogy to prove
the stability of S1 and
S3, the
resulting
0
, i.e. the derivativ
e of t
he Lyap
unov functio
n
s
are simil
a
r t
o
the ca
se i
n
S1 and
S3 without de
lay.
Since all sub
s
ystem
s
are GAS, the the over
all syste
m
is also GA
S under the
con
d
ition
of bidire
ction
a
l informatio
n
and uniform delay in the system.
Some rem
a
r
k
s reg
a
r
d
ing th
e res
u
lts:
1.
It seem
s tha
t
there
is n
o
differe
nce b
e
tw
ee
n the
delayed
an
d
non
-d
elayed
sy
stems.
However, it is only applied
for the stabilit
y proof.
2.
From the pe
rsp
e
ctive of the perfo
rma
n
ce,
the del
ayed ca
se repre
s
e
n
ts th
e worst-ca
se
scena
rio, i.e. coo
r
din
a
tion
only by mean
s of individual
trajecto
ry tra
cki
ng.
3.
In the propo
sed controll
er,
the delaye
d
sign
al mea
n
s that
the syst
em “can
not”
excha
nge
informatio
n so that coo
r
din
a
tion is a
c
hie
v
ed only individually.
4.
The
assum
p
tion of
unifo
rm del
ay is a
c
ceptabl
e in
pra
c
tice
since in
a
spe
c
if
ic n
e
two
r
k
typically transmissi
on del
ay is simila
r in a
ll directio
n.
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
1693-6
930
A Practical Coordi
nated T
r
ajecto
ry T
r
a
c
king fo
r A G
r
oup of Mixe
d
…
(Sisd
a
rm
a
n
to Adinand
ra)
1403
4. Simulation results a
n
d analy
s
is
4.1. Control Parameters, Simulation
Scenarios a
n
d Performan
ce Indicato
r
s
The
controller is validate
d
by mean
s of
si
mulatio
n
. T
he follo
wing
control
param
eters are
use
d
:
0
.
4
,
100,
0
.
5
,
2
,
1
,
1
,
0
.06,
1
0
A group of 4 robot
s is give
n a task to move in an-8
-
shape like traj
ectory a
s
dep
icted in
Figure 3. Different d
e
si
red
formation
shape,
type o
f
robots in th
e gro
up an
d
commu
nication
topologi
es a
r
e investigate
d
. The sum
m
a
ry of
param
eters
choi
ce
s is given in Table 1.
Table 1. Simulation sce
n
a
r
ios
Forma
tio
n
sh
ap
e
(FS)
R
o
bo
t
t
y
pe
(
ty
p
e
)
The sequence in
dicates the type of the robo
ts in the for
m
ation f
r
o
m
robot 1 to
4 resp
ectively; ‘
u
’
is for
unicycle, ‘
o
’
is for om
ni w
heels
ID 1: o-o
-
o-o;
ID
2: u-u-u
-
u; ID
3:
u-o-o
-
o; ID 4:
u-o
-
o-u;
ID 5: u-o
-
u-o;
ID
6: o-u-o
-
u; ID
7:
o-u-u
-
o
C
o
mm
un
i
c
a
t
io
n t
o
po
l
o
g
i
e
s
(
co
m to
p
)
The bidirectional arros indicates t
hat the robo
ts are
comm
unicating.
ID 1: all robots communicates to each other
ID 2: 1
2
3
4; I
D
3:
1
2
3
4
1
ID 4: 1
2, 1
3, 1
4; I
D
5: 1
2, 3
4
ID 6: 1
3, 2
4; I
D
7: no c
o
mmunication betw
e
en ro
bots
Figure 2. The
refere
nce tra
j
ectori
es of th
e robot
s
To comp
are
the perfo
rmance of
the cont
rollers in each
sc
ena
rio, a
RMS
-
like
perfo
rman
ce i
ndicator is u
s
ed [1, 2]:
-1
-0
.
8
-0
.
6
-0
.
4
-0
.
2
0
0.
2
0.
4
0.
6
0.
8
1
-1
-0
.
8
-0
.
6
-0
.
4
-0
.
2
0
0.
2
0.
4
0.
6
0.
8
1
x [
m
]
y [
m
]
1
2
3
4
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 16
93-6
930
TELKOM
NIKA
Vol. 14, No. 4, Dece
mb
er 201
6 : 1397 – 140
7
1404
∑∑
∑
(
8
)
Whe
r
e
m
is the numb
e
r
of rob
o
ts in the systems,
l
is t
he num
ber
of data in the
simulatio
n
/experim
ents, a
nd
∆
∆
are erro
rs in keepi
n
g
the relative time varying
distan
ce b
e
tween the ro
bot
s.
If
=0, it means that all robots maintai
n
t
he desire
d
relative distance
s
, i.e. the
formation is kept. It is to b
e
noted that
can indi
cate
a good forma
t
ion shap
e but the one
that equal
s to a rotation mirror of the de
si
red formation
shap
e.
To de
mon
s
trate the
coo
r
d
i
nation, du
rin
g
the ex
p
e
ri
ments, at
different tim
e
s,
a ro
bot i
s
simulate
d to
drive a
w
ay f
r
om its
cu
rren
t positio
n.
Th
us, the
effect
of ad
ding
co
upling
gain
s
can
be investig
ated
4.2. Simulation Res
u
lts a
nd Analy
s
is: Dela
y
-free Cases
Figure 3
sho
w
s th
e exam
ple of ro
bot
movem
ent
s i
n
one
scena
ri
o. The top fig
u
re
sho
w
s
the re
sulting
movement
s whe
r
e the
r
e i
s
no
co
mm
u
n
icatio
n between the robot
s. It can be
seen
that there i
s
n
o
rea
c
tion f
r
o
m
other
rob
o
ts, i.e.
formati
on is
achi
eve
d
only by me
ans
of traje
c
tory
tracking. On t
he other h
a
n
d
, in Figure 4
it can be
ob
served that on
e a robot is o
ff the trajecto
ry,
other
rob
o
ts react
s
to the
pertu
rbatio
n in ord
e
r
to
ke
ep the ove
r
all
formation
as the refe
ren
c
es.
Less
com
m
u
n
icatio
n sha
r
i
ng me
an
s th
e communi
ca
tion between
robot
s
can
n
o
t
be do
ne
pe
er-
to-pee
r, whi
c
h mean
s re
actions to pertu
rbati
on maybe
delayed or e
v
en can
not b
e
execute
d
.
Figure 3. The resulting movement
s for platoon group ‘
o
-u-u-o’. left:
no inform
ation sharing; ri
ght:
fully couple
d
Figure 4.
from simulation, Left: “triangle” fo
rmation; ri
ght: “platoon” formation
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
1693-6
930
A Practical Coordi
nated T
r
ajecto
ry T
r
a
c
king fo
r A G
r
oup of Mixe
d
…
(Sisd
a
rm
a
n
to Adinand
ra)
1405
Figure 4
sho
w
the valu
es of
the pe
rformance indi
ca
tors fr
om the
experim
ental
re
sul
t
s
usin
g the “tria
ngle” a
nd “pla
toon” formation sh
ape.
The re
sult
s in Figure 4 sh
ows that for varyi
ng ro
bot
type in the
grou
ps, the
controlle
rs
also p
e
rfo
r
m
s
differe
ntly althoug
h the effect can
not
be justified
clea
rly. The simulatio
n
re
sults
sug
g
e
s
that
mixed robot
s tend
s to p
e
rform b
e
tter
compa
r
e
s
to a
ll unicy
cle
or
all omni
whe
e
ls
robot
s.
Figure 5. Slices of
from s
i
mulation us
ing “triangle” form
ation. Left: “o-u-o-u”; right: “u-o-o-
o”
Figure 6. Slices of
_
^
from s
i
mulation us
ing “platoon” fo
rmation. Left: “o-u-o-u”; right: “u-
o-o
-
o”
On the oth
e
r ha
nd, re
gardl
ess the
r
obot type
s an
d form
ation shape
s, th
e
comm
uni
cati
on top
o
logi
es have
more i
n
fluen
ce to
the p
e
rfo
r
man
c
e. Amo
ng
a
ll com
b
inatio
ns,
when all robots
c
o
mmunic
a
te,
_
^
has th
e smalle
st values. At the oppo
site si
de
, whe
n
less rob
o
ts share in
dividu
al tracking e
r
rors,
_
^
tends to gro
w
.
Figure 5
and
6 sho
w
so
me sli
c
e
s
of
the re
sult
s
sho
w
n i
n
Fig
u
re
4. The
s
e
figure
s
indicate that no informatio
n sha
r
ing do
es not
always mean
s the worst pe
rformance. Althoug
h
addin
g
more
information
can i
n
crea
se
robu
stne
ss
, it has to do
n
e
co
rrectly a
c
cordi
ng to t
he
desi
r
ed
coo
r
dination a
nd
type of robot
in the grou
p
.
Thus, to ha
ve a stron
g
coordi
nation, it
is
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 16
93-6
930
TELKOM
NIKA
Vol. 14, No. 4, Dece
mb
er 201
6 : 1397 – 140
7
1406
sug
g
e
s
ted to
have complet
e
com
m
uni
ca
tion between
the rob
o
ts. Howeve
r, this
requires
a larg
e
comm
uni
cati
on ban
dwidt
h
.
4.3. Simulation Res
u
lts a
nd Analy
s
is: Dela
y
e
d Cases
Figure 7 sh
o
w
s the
summ
ary of
_
^
from simulation u
s
i
ng “tria
ngle
”
and plato
o
n
formation
with the delay
0.5
_
. The value of
_
rep
r
e
s
ent
s the sampli
ng time.
Figure 7.
from simulation using delay
s in comm
uni
cation channel,
Left: “triangle”
formation; rig
h
t: “platoon
” formatio
n
The sim
u
latio
n
results in Fi
gure 7, both l
e
ft and right, sho
w
that the
perform
an
ce
of the
system
is cl
o
s
ed
to
perfo
rmance
of the
sy
stem
s wh
en the
r
e
is n
o
info
rmation
sh
ari
ng
between
the ro
bots. A
s
me
ntione
d
in re
marks of
se
ction
3,
th
e spe
c
ific
structure of th
e
co
ntrolle
rs a
nd
assumptio
n
s
of uniform
de
lay within the
system
a
llows
the
s
y
tems
to be
sta
b
l
e
. Ho
wev
e
r, with
the con
s
e
q
u
ence of having the wo
rst
case pe
rfor
mance. The
simulatio
n
re
sults
confi
r
m
this
con
c
lu
sio
n
s.
Whe
n
comm
unication i
s
d
e
layed, the in
cre
a
si
ng n
u
m
ber
of sh
are
d
informatio
n d
oes
not
increa
se the
perfo
rman
ce.
The results i
n
Figure
7 in
dicate
s that regardl
e
ss the
formation, type
of ro
bots, fo
r
any commu
ni
cation
topol
o
g
ies, th
e
p
e
rf
orma
nce i
s
si
milar to
the
si
tuation
whe
n
no
informatio
n is shared.
It is to be noted that the probl
em form
ulation
in this research all
o
ws the coordinati
o
n
betwe
en the
robot
s to be
achi
eved onl
y by means o
f
individual trajecto
ry tracking. The ad
di
tion
of information
from other ro
bots is exp
e
ct
ed to
increa
se robu
stne
ss againt
s pertu
rbation
s
.
The re
sults,
both for dela
y
-free and d
e
layed
ca
se
s, show that the pro
p
o
s
ed
control
algorith
m
s
work well for a group
of
mixed wh
e
e
ld mo
bile robots. T
h
u
s
, the proble
m
of
coo
r
din
a
ting individual traj
ectory tra
c
kin
g
is achieved.
5. Conclusio
n
In this paper we pre
s
ent
controlle
rs th
at ac
hieve gl
obally asymp
t
otically stabl
e of the
tracking
error dynamics of
the mi
xed group of uni
cycle and om
ni
whe
e
ls m
obil
e
rob
o
ts both
in
the ab
se
nt o
r
presen
ce
of
delays in
the
co
mm
uni
cati
on cha
nnel. The co
ordi
na
tion
bet
wee
n
the
robot
s
can
b
e
achieved
b
y
sha
r
ing in
d
i
vidual tr
a
cki
ng erro
rs bet
wee
n
the
rob
o
ts. The
rob
o
ts
requi
re h
a
ving a bidire
ctio
nal com
m
uni
cation, i.e. if
i
shares m
e
ssage
s to
j,
then
j
has to sh
a
r
es
messag
es to
i.
Simulation results
sug
g
e
s
t that more i
n
formatio
n sh
aring, for
del
ayed-free case,
rega
rdl
e
ss th
e formation
shape a
nd type of robots in
the grou
p ten
d
s to increa
se the robu
stn
e
ss
in coordinati
ng the
move
ments und
er pe
rturb
a
tion
s. Fo
r the
d
e
layed
ca
se,
re
gardle
s
s the
formation
sh
ape
and
type of
robot
s,
the pe
rfor
m
a
nce
in simila
r
to situation
wh
en no ro
bots
comm
uni
cat
e
.
Evaluation Warning : The document was created with Spire.PDF for Python.