TELKOM
NIKA
, Vol.13, No
.2, June 20
15
, pp. 528 ~ 5
3
8
ISSN: 1693-6
930,
accredited
A
by DIKTI, De
cree No: 58/DIK
T
I/Kep/2013
DOI
:
10.12928/TELKOMNIKA.v13i2.1111
528
Re
cei
v
ed
No
vem
ber 1
9
, 2014; Re
vi
sed
March 9, 201
5; Acce
pted
March 26, 20
15
Goal-Seeking Behavior
-Based Mobile Robot Using
Particle Swarm Fuzzy Controller
Andi Adria
n
s
y
ah*, Yudhi Gunardi, Ba
daruddin, Ek
o Ihsanto
Electrical E
ngi
neer
ing D
e
p
a
rtement, F
a
cult
y of Engine
eri
n
g
,
Universitas M
e
rcu Bua
n
a
Jl. Meru
ya Sel
a
tan, Kemb
ang
an, Jakarta Bar
a
t, 1165
0, Indo
nesi
a
*Corres
p
o
ndi
n
g
author, e-ma
i
l
: andi
@mercu
bua
na.ac.i
d
A
b
st
r
a
ct
Behav
ior-b
ase
d
co
ntrol archit
ecture has
s
u
c
c
essf
ully de
mo
nstrated
th
eir
c
o
mpete
n
ce in
mo
bil
e
robot dev
el
op
me
nt. Fu
z
z
y
l
ogic syste
m
character
i
stics are suita
b
le t
o
addr
ess the
behav
ior des
ig
n
prob
le
ms. How
e
ver, there ar
e
difficulties
enc
ounter
ed
w
h
e
n
setting fu
zz
y
p
a
ra
meters
ma
n
ually. T
her
efor
e,
m
o
st
of the works in the
field generate
c
e
rt
ain inter
e
st for the study of fu
z
z
y
system
s wit
h
added learning
capa
bil
i
ties. T
h
is
pap
er
pre
s
ents the
d
e
v
e
lo
p
m
ent
of
f
u
zz
y
b
ehav
ior-
base
d
c
ontrol
architectur
e
us
ing
Particle Sw
ar
m Optimi
z
a
tio
n
(
PSO). A goal-s
eeki
ng
beh
avi
o
rs base
d
o
n
P
a
rticle Sw
ar
m
F
u
zz
y
Contro
ll
er
(PSF
C) are de
velo
ped us
in
g the mod
i
fied P
S
O w
i
th
tw
o
stages of the PS
F
C
process. Severa
l simul
a
ti
ons
and
ex
peri
m
en
ts w
i
th Mage
ll
a
n
Pro
mobi
le
ro
bot h
a
ve
be
en
perfor
m
ed t
o
ana
ly
z
e
t
he
pe
rforma
nce
of t
h
e
alg
o
rith
m. T
h
e
pro
m
isi
ng res
u
lts hav
e prov
ed that the
pro
pos
e
d
contro
l
architectur
e
for
mo
bil
e
ro
bot h
a
s
better capa
bi
lit
y to accomplis
h useful task i
n
real office-l
ike
envir
on
me
nt.
Ke
y
w
ords
:
Behavior-Based Robot, Fu
z
z
y
Logic, PSO, PSFC
1. Introduc
tion
Develo
ping
a
mobile
robot
is
an i
n
tere
st
ing ta
sk.
Usually, the mo
bile robot
sh
o
u
ld fa
ce
unpredi
ctable
environ
ment
, perceive in
accurate sen
s
or and act with
un
sati
sfactory
a
c
tuat
or
in
high-speed response [1],[2]. Behav
ior-based control arch
itecture is an alternative method
approp
riate t
o
add
re
ss th
ese
problem
s [3]-[7]. The a
r
chite
c
tu
re i
s
able to
act
wi
th fast re
al-ti
m
e
respon
se,
provides for hig
her-l
evel
deli
beratio
n a
n
d
has de
mon
s
t
r
ated
its
relia
ble p
e
rfo
r
ma
nce
in stand
ard
robotic a
c
tivities. Ho
wever, a kind
of sof
t
computing i
s
nee
ded to complete two
key
issue
s
in
b
ehavior-b
a
se
d sy
stems,
su
ch
as
ge
neratin
g o
p
timal individ
u
al be
havior
and
coordinating
multiple behaviors.
Curre
n
tly, se
veral meth
od
s that hybri
d
fu
zzy
syste
m
with evolut
ionary al
gorit
hms h
a
s
been proposed in be
havi
o
r-based
mobile robot, such
as
Genetic Algorithm [8],[9], Genetic
Programmin
g
[10] to overcome the beh
avior-ba
s
ed i
s
sue
s
. Ho
we
ver, the existing evolution
a
r
y
algorith
m
s
u
s
ed
have
several d
r
a
w
backs [11], su
ch a
s
not ea
sy to impleme
n
t and
comp
utationa
lly expensive
[12], require
much
p
r
o
c
e
ss
sho
u
ld b
e
compl
e
ted a
nd pa
ramete
rs
sho
u
ld b
e
a
d
juste
d
, hav
e sl
ow
co
nverge
nce
a
b
ili
ty to find n
ear-optimu
m
sol
u
tion, a
nd
depe
ndent h
e
u
risti
c
ally to genetic o
perators [13].
Fortun
ately, Kennedy a
n
d
Eberha
rt introdu
ced the P
a
rticle S
w
a
r
m
Optimizatio
n
(PSO) in 1
9
9
5
[14],[15]. PSO is one of
evolutiona
ry comp
utat
ion
techni
que to
find the optimal solutio
n
by
simulatin
g
su
ch soci
al be
havior of gro
ups
su
ch
a
s
fish schooli
ng or bi
rd flocki
ng. The
r
e
are
several a
d
va
ntage
s of the
PSO as co
mpared to
other
evolution
a
ry computati
on meth
od
s. The
PSO is ea
sy t
o
imple
m
ent
and i
s
co
mpu
t
ationally
inex
pen
sive
sin
c
e
its m
e
mo
ry a
nd
CPU spee
d
requi
rem
ents are lo
w. Additionally, the PSO r
equi
re
s only a few process
shoul
d be co
mplet
e
d
and pa
ram
e
ters to b
e
adj
usted. In an
other
si
de, th
e PSO has q
u
ick co
nverg
ence ability to find
optimum
or n
ear-optimu
m
solutio
n
. Ge
n
e
rally, PSO
has p
r
oved
to be
an
effici
ent meth
od f
o
r
nume
r
ou
s g
eneral optimi
z
ation p
r
o
b
le
ms, and in
some
ca
se
s it does not
suffer from
the
probl
em
s en
countered by o
t
her evol
ution
a
ry com
putati
on [11]-[13].
This
pap
er a
ddre
s
se
s the
pro
b
lem
s
of
developi
ng
control a
r
chitecture of m
o
bile robot
with beh
avior-ba
sed
syste
m
, espe
cially
in goal
-s
e
e
ki
ng beh
avior.
The proble
m
solving i
s
rel
a
ted
to the spe
c
ification of mo
bile rob
o
t tasks,
the dev
elopme
n
t of mobile robot
behaviors, the
interp
retation
of the environment a
nd t
he validat
ion
of the final system. Thi
s
pape
r u
s
e
s
and
develop
s soft
computin
g, makin
g
exten
s
ive use
of F
u
zzy Logic a
n
d
Particle Swarm Optimi
za
tion
(PSO) n
a
me
d as P
a
rticl
e
Swarm
Fu
zzy Co
ntrolle
r (PSFC). Th
e use of PSO is to tun
e
fuzzy
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
1693-6
930
Goal-S
eeking Behavior-B
as
ed Mobile Robot Using P
a
rticle S
w
arm
Fuzzy .... (Andi Adriansyah)
529
membe
r
ship
function
and
to learn fuzzy
rule
ba
se fo
r goal
-see
king
behavio
r. Th
is fuzzy tuni
n
g
and lea
r
nin
g
is perfo
rme
d
to accompli
sh
the best be
ha
vior-b
ased sy
stem.
2. Res
earc
h
Method
2.1. Goal-s
eekin
g
Behav
i
or
Model
Goal
see
k
in
g
behavio
r
ste
e
rs and
mov
e
s the
ro
bot to the ri
ght direction
and
re
ach th
e
goal effe
ctively. The mobil
e
ro
bot move
ment towa
rd
s the go
al is a
c
cordi
ng to t
he di
stan
ce
a
n
d
angle bet
ween the current position of the mobile
robot and the
goal position
[16],[17]. In this
work, Mag
e
ll
anPro m
obil
e
rob
o
t is u
s
ed fo
r verification a
nd p
e
rform
a
n
c
e
analysi
s
of the
prop
osed alg
o
rithm. The
Magella
nPro
is a circula
r
mobile rob
o
t from iRob
ot, Real Wo
rld
Interface (RWI), the
acknowl
edge
d in
dustry l
ead
er
in the
exciti
ng field
of cu
tting-edg
e m
obile
roboti
c
. The d
i
mensi
on of the rob
o
t is as follows:
D
= 40.64 cm,
H
= 25.4 cm,
r
= 5.7 cm,
W
=
36
cm a
nd
M
= 18.2 kg,
whe
r
e
D
i
s
diamet
er,
H
is heig
h
t,
r
is th
e radi
us
of wh
eel
s,
W
i
s
di
stan
ce
betwe
en
t
w
o whe
e
ls, and
M
is
wei
ght,
respe
c
tively. Figu
re 1
sh
ows the
physical
stru
ctu
r
e
of
Magella
nPro
mobile robot.
Figure 1. MagellanP
ro Mo
bile Ro
bot
Figure 2 illust
rated a mo
de
l of Magellan
P
ro m
obile
ro
bot for simul
a
tion exerci
se
s for the
prop
osed
alg
o
rithm.
The
mobile
ro
bot i
s
lo
ca
te
d on a
two
dim
e
n
s
ional Ca
rtesi
an wo
rksp
ace,
in
whi
c
h a glo
b
a
l coo
r
din
a
te
{
X,O,Y
} is def
ined. The
rob
o
t has three
degree
s of p
a
ram
e
ter p
o
si
tion
that are
represe
n
ted
by a
po
sture
p
c
=
(
x
c
, y
c
,
θ
c
)
, where
(x
c
, y
c
)
indi
cate the
spatial
po
sitio
n
of
the rob
o
t gui
de poi
nt in the
glob
al co
ordinate
syste
m
and
θ
c
is t
he he
ading
a
ngle of the
ro
bot
cou
n
t
e
r
-
cl
oc
k
w
is
e
from the x-axis
.
Figure 2. Model of Magell
anPro mo
bile
robot
X
x
c
C
D
2
r
y
c
O
Y
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 16
93-6
930
TELKOM
NIKA
Vol. 13, No. 2, June 20
15 : 528 – 53
8
530
The mathe
m
atical mo
del
for the ro
bot
movement
can b
e
obtai
ned with diff
erentially
steered d
r
ive system o
r
kn
own a
s
differential drive system [18].
Base
d on this
system, the robo
t
can m
o
ve to
different po
sit
i
ons
and
orie
ntations
as
a
function
of time. The d
e
rivatives of
x
,
y
and
θ
can b
e
obtaine
d as
c
c
v
d
t
dx
cos
(1)
c
c
v
dt
dy
sin
(2)
c
dt
d
(3)
whe
r
e
ω
c
is t
he ang
ular ve
locity of the robot and
whe
r
e
v
c
is the lin
ear velo
city of the robot.
By applying the current p
o
sition of the
robot,
p
c
=
(
x
c
,
y
c
,
θ
c
), th
e next positi
on of the
robot
is su
ch as
follo
ws:
t
v
x
x
c
c
c
c
*
cos
1
(4)
t
v
y
y
c
c
c
c
*
sin
1
(5)
t
c
c
c
*
1
(6)
Then, a
s
a
s
suming the val
ue of
∆
t
is
a unit time step
, the next position of the ro
bot,
p
c+
1
= (
x
c+
1
,
y
c+
1
,
θ
c+
1
), in
s
i
mple form is
:
c
c
c
c
v
x
x
cos
1
(7)
c
c
c
c
v
y
y
sin
1
(8)
c
c
c
1
(9)
In maintaini
n
g a
co
urse
to
a g
oal l
o
cation, o
r
sea
r
ch
ing, an
effecti
v
e strategy
called
as
aiming navig
ation [19] is used. The mo
b
ile robot ai
mi
ng at a goal has to orient its body axis such
that the go
al
is in f
r
ont of i
t. The g
oal
must b
e
a
s
so
ciated
with
some
salie
nt cue. In thi
s
work,
the goal is
pri
o
r spe
c
ified b
y
the human
use
r
. Furt
h
e
rmore, by mea
n
s of od
ometry strategy, bo
th
the dire
ction and the dista
n
ce to the go
al are
acqui
re
d. Finally, the
goal can be
approa
che
d
from
variou
s directions, as ill
u
s
trated i
n
Fi
gure
3.
This techniq
ue i
s
sim
p
le, fast, and ha
s no
cumul
a
tive error
reroute to the goal.
Figure 3. Aiming navigatio
n
targ
et
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
1693-6
930
Goal-S
eeking Behavior-B
as
ed Mobile Robot Using P
a
rticle S
w
arm
Fuzzy .... (Andi Adriansyah)
531
In the directi
on of movin
g
to a g
oal
point, mobil
e
rob
o
ts n
eed
to kn
ow it
s relative
positio
n. By
some m
odification, integrat
ing
and ap
plying the initial positio
n of the robot a
s
x
(0)
=
x
0
,
y
(0
) =
y
0
, and
θ
(0
) =
θ
0
yields:
0
0
0
sin
)
(
sin
)
(
2
)
(
)
(
W
t
v
v
v
v
v
v
W
x
t
x
l
r
l
r
l
r
(10
)
0
0
0
cos
)
(
cos
)
(
2
)
(
)
(
W
t
v
v
v
v
v
v
W
y
t
y
l
r
l
r
l
r
(11
)
W
t
v
v
t
l
r
)
(
)
(
0
(12
)
whe
r
e
v
r
an
d
v
l
are the rig
h
t
and the left
whe
e
l velociti
es alon
g the grou
nd, re
sp
ectively, and the
positio
n of the robot at time
t
at the coordinate is
x(t
)
,
y(t
)
,
θ
(t)
.
Based
on
th
e ro
bot p
o
sit
i
on a
nd h
e
a
d
ing, the
rel
a
tive po
sition
to the
goal
point i
s
cal
c
ulate
d
. The relative po
sition
s are
kn
own a
s
targ
et distan
ce (
d
)
and targ
et an
gle
(
δ
)
,
w
h
er
e:
2
2
)
)
(
(
)
)
(
(
g
g
y
t
y
x
t
x
d
(13
)
)
(
)
(
(
)
)
(
(
tan
t
x
t
x
y
t
y
arc
g
g
(14
)
whe
r
e
p
g
= (
x
g
,
y
g
,
θ
g
) is th
e goal p
o
sitio
n
. The targ
et distan
ce
(
d
)
and target an
gle
(
δ
)
ar
e
use
d
as th
e in
puts
for g
oal
se
eki
ng b
ehavio
r.
Fig. 4 ill
ustrates th
e
relativ
e
po
sition
bet
wee
n
the
rob
o
t
and the go
al point.
Figure 4. Rel
a
tive position
s
between th
e robot an
d the goal p
o
int
2.2.
Fuzz
y
Goal-seeking
Beh
a
v
i
or Structure
FLC
structu
r
e
ba
sed
on M
a
mdani te
ch
ni
que i
s
u
s
e
d
i
n
this
syste
m
. The
r
e a
r
e
two input
s
requi
re
d, na
med
as targ
et dista
n
ce
(
d
)
and
targ
et angl
e
(
δ
).
T
hese in
puts
are
obtain
e
d
by
cal
c
ulatio
n of the relative
posit
io
n between the
cu
rre
nt position
of
robot by me
ans of o
dom
ete
r
and th
e g
oal
positio
n, a
s
stated in Eq
ua
tion (1
3)
an
d
Equation
(1
4). Also, the
r
e
are t
w
o
outp
u
t
resulted, nam
ed as lin
ear v
e
locity,
v
, and
angula
r
velo
city,
ω
.
Trap
ezoid a
n
d
triang
ula
r
sha
pe a
r
e u
s
ed
as i
nput
membe
r
ship
function
s a
n
d
output
membe
r
ship f
unctio
n
s, for f
u
zzificatio
n a
nd defu
zzi
fi
cation process, resp
ectively
. The relatio
nal
O
X
Y
d
δ
p
g
p(
t)
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 16
93-6
930
TELKOM
NIKA
Vol. 13, No. 2, June 20
15 : 528 – 53
8
532
function
between in
put a
n
d
control
of
a fuzzy be
ha
vior a
r
e d
e
scribed
by me
a
n
s
of fuzzy rule
base. Each rule is con
c
ate
nated a
s
:
RB
i
: if
X
1
is
A
1
and
X
2
is
A
2
and …
X
n
is
A
n
then
Y
1
is
B
1
and
Y
2
is
B
2
(15
)
Then, the o
u
tput is obta
i
ned by appl
ying the fu
zzy rule ba
se
inferen
c
e a
nd the cent
roid
defuzzificatio
n
scheme, a
s
l
i
ox
i
l
i
ox
ox
i
o
D
D
C
y
.
.
(16
)
whe
r
e
C
ox
an
d
D
ox
are the
param
eters
of cente
r
and
widt
h of output membe
r
ship functio
n
s
at
rule
i
,
α
i
is the prod
uct of
the degree of
me
mbershi
p
of each inp
u
t
s at rule
i
, and
l
is
the total
numbe
r of rul
e
s fired. The
sele
cted te
ch
nique
s we
re
cho
s
e
n
due to their linea
rity, computatio
nal
simpli
city, and easy to und
ersta
nd.
Every input of fuzzy has three lingui
stic te
rm
s, which are CLOSE, MEDIUM an
d
FAR for
distan
ce
s an
d RIG
H
T, FO
RWA
RD
and
LEFT for a
ngl
e,
as de
picte
d
gene
rally in
Figure 5 Th
ree
lingui
stic term
s i
s
chosen on
behalf of the minimal
number fo
r fuzzy system. The value
of
x
i
an
d
y
i
are tun
ed a
u
tomatically a
s
de
scribe
d in the next se
ction
s
.
0
0.
2
0.
4
0.
6
0.
8
1
di
s
t
anc
e (
m
)
degree of
m
e
m
bers
h
i
p
(
u
)
0
0.
2
0.
4
0.
6
0.
8
1
angl
e (
r
ad)
degree of
m
e
m
b
e
r
s
h
i
p
(
u
)
C
L
O
SE
M
E
D
I
UM
F
A
R
RI
G
H
T
FO
R
W
A
R
D
L
E
F
T
a1
a2
a3
b1 b2
b3
Figure 5. The
membershi
p
f
unction
s of d
i
stan
ce
s and
angle
In this wo
rk, li
near velo
city,
v
, and ang
ul
ar velocity,
ω
, are ap
plied a
s
output
s of a
ll fuzzy
behavio
r mo
dule
s
. The
lingui
stic te
rm
s u
s
ed
a
r
e
L
O
W, ME
DIUM and
HIG
H
for lin
ear vel
o
city,
and, RIG
H
T, FORWARD, and LEFT for angula
r
velo
city. The fixe
d membe
r
shi
p
function
s of
v
and
ω
is
sho
w
n in Figu
re
6.
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
1693-6
930
Goal-S
eeking Behavior-B
as
ed Mobile Robot Using P
a
rticle S
w
arm
Fuzzy .... (Andi Adriansyah)
533
0
0.
2
0.
4
0.
6
0.
8
1
v (
m
/
s
)
de
gree of
m
e
m
bers
h
i
p
(u)
0
0.
2
0.
4
0.
6
0.
8
1
w
(
r
ad/
s
)
deg
ree of
m
e
m
bers
h
i
p
(u)
LO
W
M
E
DI
U
M
HI
G
H
RI
G
H
T
F
O
R
W
AR
D
L
EF
T
0.
05
0.
15
0.
25
-pi
/
4 0
pi
/
4
Figure 6. The
membershi
p
function of
lin
ear velo
city and ang
ular ve
locity
2.3. PSFC
Desig
n
Basically, PSFC is an F
L
C augmente
d
by a tuning o
r
learning p
r
o
c
e
ss b
a
sed o
n
PSO.
In PSFC, PS
O is ap
plied
i
n
o
r
de
r to
se
arch fo
r a
n
a
ppro
p
ri
ate Kn
owle
dge
Base (KB) of a
fu
zzy
system
for a
particula
r p
r
o
b
lem
and
to
ensure
tho
s
e
pa
ramete
r v
a
lue
s
a
r
e
opti
m
al
with
re
sp
ect
to the desig
n crite
r
ia. Th
e KB param
eters
co
nsti
t
u
te the optimization
spa
c
e, whi
c
h i
s
then
transfo
rme
d
i
n
to suitable
positio
n o
n
which
the
se
arch
pro
c
e
s
s o
perate
s
. Fi
gu
re 7
sho
w
s the
con
c
e
p
t of a PSFC syste
m
whe
r
e PS
O de
sign a
n
d
fuzzy pro
c
essing a
r
e th
e two fund
a
m
ental
con
s
tituent
s.
Figure 7. The
con
c
ept of a PSFC
At the beginn
ing of the proce
s
s,
the initial populatio
ns comp
rise a set of parti
cle
s
that
are scatte
red
all over the sea
r
ch sp
ace
.
The initial populatio
n ma
y be randoml
y
generate
d
or
may be partl
y supplied b
y
the user. Howeve
r, in
this wo
rks, the
population
s
are rando
mized
initially.
Afterwa
r
d, on
e particl
e is take
n and de
cod
ed to the
actual value
of the goal-see
kin
g
fuz
z
y
parameter. Thes
e sets
of fuz
z
y
c
o
ntrolle
r pa
ra
meters are th
en used to control the fu
zzy
behavio
r whe
r
e it un
de
rgo
e
s a
serie
s
of trackin
g
resp
onse of multi
s
tep
referen
c
e set
point. T
he
use
of
a m
u
ltistep
refe
ren
c
e si
gnal
is to
excite
the
diff
erent
state
s
of the
syste
m
, to en
able
th
e
evaluation
to
cove
r
a
wid
e
r
system
op
eration
ran
g
e
.
Base
d o
n
t
he va
riou
s
st
ate of the
co
ntrol
system, the
perfo
rman
ce
of the control
l
er i
s
ca
lculat
ed by u
s
in
g
a predefin
ed
fitness functi
on.
PSO is then
use
d
to
tune
the fuzzy
co
n
t
roller pa
rameters to
minimiz
e
the fitnes
s func
tion. The
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 16
93-6
930
TELKOM
NIKA
Vol. 13, No. 2, June 20
15 : 528 – 53
8
534
assignm
ent o
f
the fitness f
unctio
n
serve
s
a
s
a
gui
da
nce
to lea
d
t
he
sea
r
ch to
ward the
opti
m
al
s
o
lution.
Furthe
rmo
r
e,
for the pu
rposed of tun
i
ng fuzzy m
e
mbe
r
ship fu
nction
s the f
o
llowin
g
equatio
ns we
re
define
d
:
C
x
= C
x
+
k
i
(17
)
D
x
= D
x
+ j
i
(18
)
whe
r
e
k
i
a
nd
j
i
are adju
s
tm
ent coeffici
en
ts,
C
x
, and
D
x
are set of ce
ntre and
widt
h of each fu
zzy
membe
r
ship
function,
re
spectively. T
he a
d
ju
st
me
nt co
efficient
s ta
ke
any
real
po
sitive o
r
negative valu
e. Therefo
r
e
,
k
i
make
s ea
ch cente
r
of membe
r
ship
function m
o
ve to the right
or
left
and
the membe
r
ship
function
s sh
ri
nks
o
r
expa
nds
throug
h
j
i
,
as shown
in Fig.
8.
T
he
shifting codin
g
strate
gy
wi
ll simplify se
arching
com
putation,
be
cause there is no ne
ce
ssit
y
to
sort the value
of membersh
ip function
s in
ascendi
ng m
anne
r.
Figure 8. Prin
ciple in tuni
ng
of membersh
ip function
The PSO
p
r
ocess
sta
r
ts
with rand
omly
gen
era
t
ed
initial p
opulatio
ns. Then,
all
popul
ations o
f
particle
s
are
eval
uated
an
d asso
ciate
d
based o
n
fi
tness fun
c
tion t
o
determine t
he
pbe
st
and
g
best
. Based
on seve
ral
initial investi
gation
s
and
trials a
nd e
r
rors, the fitness
function
s for
goal see
k
ing
can b
e
obtain
ed as
I
i
K
k
d
goal
k
v
k
e
k
e
f
00
2
2
)
)
(
/
100
)
(
)
(
100
(
(19
)
w
h
er
e
I
is th
e total
numb
e
r
of sta
r
t p
o
s
ition,
K
i
s
t
h
e nu
mbe
r
of
step
sim
u
lati
on fo
r e
a
ch
star
t
positio
n,
e
θ
is the angle error,
e
d
is the distan
ce erro
r,
ω
(k
)
, and
v(k
)
a
r
e the
linear velocity
a
nd
angul
ar velo
city
at
k
, respec
tively.
In this work,
a Sigmoid De
cre
a
si
ng Inertia
Weight
(SDIW) is u
s
ed to provid
e faster
spe
ed of con
v
ergen
ce
and
better a
c
curacy of optimi
z
ed valu
e [20
]. Conse
que
ntly, PSFC would
gene
rate opti
m
al and relia
ble goal
-see
king beh
avior
of the mobile robot.
3. Resul
t
s
and
Discus
s
ion
Several expe
riment
s exercise
s have be
en per
fo
rme
d
.
Some steps of experime
n
ts have
been
de
sig
n
e
d
. Fi
rstly, a
PSFC o
p
timization p
r
o
c
e
s
ses i
s
co
ndu
ct
ed to
find th
e
optimize
d
val
u
e
of fuzzy pa
ra
meters. T
h
e
n
, sim
u
lation
s of the
mobil
e
ro
bot b
a
sed
on the
PSFC
are
analy
z
ed
to
investigate th
e co
ntrol
beh
avior of PSF
C. Resu
lts of
fuzzy
beh
avior that a
r
e
ob
tained m
anua
lly,
obtaine
d by GA, called
a
s
Ge
netic F
u
zzy
Contro
lle
r (GF
C
) from
previou
s
wo
rks a
r
e u
s
ed
as
comp
ari
s
o
n
[21]. Finally, a real
rob
o
t, Magella
nPro, movement
s are teste
d
to go to so
me
loc
a
tion.
PSO and G
A
pro
c
esse
s for goal
-see
king b
ehav
io
r are sh
own
in Figure 9,
whe
r
e,
evolution
s
of
the be
st fitness value
ag
ainst g
ene
rati
on are illu
strated. At the begin
n
ing of
the
run, the process tended t
o
have
more global
search ability beca
use of l
a
rge i
nertia weight.
It
wa
s sho
w
n t
hat the fitne
ss val
ue ove
r
all ge
ne
rati
ons i
s
conve
r
ging
qui
ckly.
After that, the
process tended to have more local
search ability caused by the small value of inertial weight
.
D
i
x
µ
C
i
C
i+1
D
i+1
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
1693-6
930
Goal-S
eeking Behavior-B
as
ed Mobile Robot Using P
a
rticle S
w
arm
Fuzzy .... (Andi Adriansyah)
535
(a)
(b)
Figure 9. Co
mpari
s
o
n
of PSO vs. GA
pro
c
e
ss fo
r g
oal se
eki
ng b
ehavior
(a)
Rule b
a
se lea
r
ning, (b) Me
mbershi
p
fun
c
tion tunin
g
Hen
c
e,
a
go
al was pla
c
e
d
in
certain
p
o
sition,
(5,
8, 0) an
d
depi
cted
as a
sm
all bla
ck
squ
a
re.
Initia
lly, robot
are
positio
ned
at
(2, 2,
π
/2
),
(5
, 2, 0),
and
(8
, 2,
π
). Sim
u
l
a
tions of m
o
b
ile
robot m
o
vem
ents a
r
e
depi
cted at Fi
gure 10. It is
no
ted that the
PSFC was
a
b
le to ap
ply goal
see
k
in
g beh
a
v
ior more effe
ctive than GA
, in three ca
ses.
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 16
93-6
930
TELKOM
NIKA
Vol. 13, No. 2, June 20
15 : 528 – 53
8
536
(a)
(b)
(c
)
Figure 10. Simulation of m
obile ro
bot m
o
vements
with GA (left) an
d PSO (right
) comp
ari
s
o
n
in
differential ini
t
ial positions
Finally, this experime
n
t wa
s perfo
rme
d
to inve
stigate
the moveme
nt of the real mobile
robot. Th
e a
c
tual rob
o
t mo
vement wa
s
depi
cted in Fi
gure
11. Th
e
distan
ce
and
angle
of targ
et
from
the cu
rrent
po
sition a
s
cal
c
ulated
from
odom
ete
r
in
side th
e
robot. The
fig
u
re
sh
owed t
h
e
perfo
rman
ce
of goal se
e
k
ing b
ehavio
r, whe
r
e
the
mobile robo
t progressive
ly redu
ced t
h
e
distan
ce
and
the a
ngle
be
tween
the ta
rget an
d
the
current p
o
sitio
n
. The
figure dem
on
strat
e
d
the mobile ro
bot wa
s able
to reach the target effe
ctively.
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
1693-6
930
Goal-S
eeking Behavior-B
as
ed Mobile Robot Using P
a
rticle S
w
arm
Fuzzy .... (Andi Adriansyah)
537
Figure 11. Re
al robot move
ment
4. Conclu
sion
Goal-se
e
ki
ng
behavior-ba
sed
control a
r
chite
c
tu
re h
a
s succe
s
sful
ly demonst
r
at
ed their
comp
eten
ce i
n
mobil
e
rob
o
t develo
p
m
ent. Fu
zzy
L
ogic System
s app
ear to b
e
very u
s
eful
to
develop the h
i
gh relia
ble a
nd effective behavior-b
a
se
d system. Ho
wever, there are difficultie
s to
set memb
ership function
and fuzzy rul
e
base in
Fu
zzy System manually. Thi
s
pape
r presents
the devel
op
ment of fu
zzy goal-se
e
ki
ng b
ehavio
r-based
co
ntro
l archite
c
ture
usi
ng PSO
for
Magella
nPro
mobile ro
bot. The work ha
s been d
one
in some tasks: behavio
ral desi
gning of the
mobile
rob
o
t, desig
ning
new fu
zzy
behavio
r
co
ordin
a
tion, a
nd finally, i
m
pleme
n
ting
the
prop
osed alg
o
rithm in re
al environ
ment.
Based o
n
the experime
n
t result
s, the mobile ro
bot
is able to deal with goal
-se
e
ki
n
g
behavio
rs.
G
enerally, it is
noted th
at th
e p
r
opo
se
d
control
archite
c
ture
h
a
s the
good
ability t
o
be
applie
d in Ma
gellanP
ro mo
bile.
Ackn
o
w
l
e
dg
ement
The autho
r wishe
s
to express his
sin
c
e
r
e t
han
ks to t
he Directo
r
at
e Gene
ral of
High
er
Educatio
n, Mi
nistry
of Edu
c
ation
an
d
Culture
which
awa
r
de
d the
grant
fundi
ng
for
Fun
dame
n
tal
Re
sea
r
ch (Project No. 201
3 0263/E5/20
14).
Referen
ces
[1]
W
i
dod
o NS, R
ahma
n
A. Visi
on Bas
ed S
e
lf
Loca
lizati
on for
Huma
noi
d R
o
bot Socc
er.
TEL
K
OMNIKA
T
e
leco
mmunic
a
tion C
o
mputi
n
g Electron
ics a
nd Co
ntrol
. 20
12; 10(4): 6
37-
644.
[2]
W
i
cakson
o
H,
Khos
w
a
nto H
,
Kus
w
a
d
i
S. T
e
leauton
omo
u
s Co
ntrol o
n
Rescu
e Ro
b
o
t Protot
ype.
T
E
LKOMNIKA T
e
leco
mmunic
a
tion C
o
mputi
n
g Electron
ics a
nd Co
ntrol
. 20
12; 10(4): 6
21-
628.
[3]
Don
g
shu
W
,
Yushe
ng Z
,
W
enji
e
S.
B
eha
vior-Bas
ed H
i
e
r
archica
l
F
u
zzy Contro
l for
Mobil
e
R
o
b
o
t
N
a
vi
ga
ti
on
in
D
y
n
a
m
i
c En
vi
ro
nm
en
t.
Chi
n
e
s
e Co
ntrol
and
Decissi
on
Co
ntrol (CC
DC
2
011). C
h
in
a.
201
1: 241
9-24
24.
[4]
Parasur
a
man
S, Ganap
ath
y
V, Shiri
n
zad
e
h
B.
Beh
a
vio
u
r
Based
Mob
i
l
e
Rob
o
t Nav
i
g
a
ti
on T
e
chni
qu
e
AI S
y
stem: E
x
perime
n
tal
Inv
e
stigati
on
on
Active Me
dia
Pion
eer
Ro
bot
.
IIUM Engi
ne
erin
g Jo
urn
a
l
.
200
5; 6(2): 13-
25.
[5]
Bao QY, Li
SM, Shang
W
Y
, An MJ.
A F
u
zz
y
B
e
h
a
vior-b
ase
d
A
r
chitecture for
Mobil
e
R
o
b
o
t
Navig
a
tio
n
in
Unknow
n E
n
viro
nments
. Internati
o
n
a
l C
onfere
n
ce o
n
Artificial Inte
llig
enc
e an
d
Comp
utation
a
l
Intelli
genc
e (AICI 2009)
. Sh
an
gha
i. 200
9: 25
7-26
1.
[6]
Mo H, T
ang Q
,
Meng
L. Be
h
a
vior-b
ase
d
F
u
zz
y
Contro
l for
Mobi
le
Rob
o
t
Navig
a
tio
n
.
Ma
thematica
l
Probl
e
m
s in En
gin
eeri
n
g
. 20
1
3
: 1-10.
[7]
Khatoo
n S, Ibrahe
em. Auto
nomo
u
s Mob
i
l
e
r Rob
o
t Navi
gatio
n b
y
Co
mbini
ng L
o
ca
l
and Glo
b
a
l
T
e
chniques.
In
ternatio
nal J
o
u
r
nal of Co
mput
er Appl
icatio
ns
. 2012; 37(
3): 1-10.
[8]
Sim KB, B
y
u
n
KS, Lee DW
. Desig
n
of F
u
zz
y
C
ont
rol
l
er
usin
g Schem
a
Coev
oluti
o
n
a
ry Al
gorit
hm.
IEEE Transaction of Fu
z
z
y
Sy
stem
. 20
04; 12
(4): 565-5
70.
Evaluation Warning : The document was created with Spire.PDF for Python.