Indonesi
an
Journa
l
of El
ect
ri
cal Engineer
ing
an
d
Comp
ut
er
Scie
nce
Vo
l.
13
,
No.
3
,
Ma
rch
201
9
, p
p.
11
10
~
11
1
6
IS
S
N: 25
02
-
4752, DO
I: 10
.11
591/ijeecs
.v1
3
.i
3
.pp
111
0
-
11
1
6
1110
Journ
al
h
om
e
page
:
http:
//
ia
es
core.c
om/j
ourn
als/i
ndex.
ph
p/ij
eecs
Solving
RFID
mobile
read
er
p
ath
probl
em
with
optimiz
ation al
go
rithms
M.
Z
ak
i Z
aka
ri
a
1
,
S
of
i
an
it
a Mu
ta
li
b
2
,
Shu
z
l
ina Abdul
R
ah
m
an
3
,
Sh
am
sul J E
li
as
4
,
A
Z
amb
ri
S
hahuddin
5
1
,2,3,5
Facul
t
y
of C
om
pute
r
and
Mathe
m
at
i
ca
l
Sc
ie
nc
es,
Univ
ersity
Technol
og
y
of
MA
RA,
Malay
s
ia
4
Facul
t
y
of
Com
pute
r and
Ma
them
at
ic
a
l
Sci
ences,
Univer
si
t
y
Tec
hnolog
i
MA
RA
Keda
h
,
Malay
si
a
Art
ic
le
In
f
o
ABSTR
A
CT
Art
ic
le
history:
Re
cei
ved
Oct
5
, 2
018
Re
vised Dec
6
,
2018
Accepte
d Dec
17
, 201
8
Radi
o
Freque
nc
y
Ide
nt
ifica
t
ion
(RFID
)
is
a
one
of
the
faste
st
growing
and
m
ost
bene
fic
i
al
te
chno
logi
es
being
adopt
ed
b
y
b
usinesses
today
.
One
of
the
important
issues
is
loc
aliz
at
ion
o
f
it
ems
in
a
war
ehouse
or
business
pre
m
ise
and
to
k
ee
p
trac
k
of
the
said
it
e
m
s,
it
req
u
ire
s
devi
c
es
which
a
re
costly
to
depl
o
y
.
Thi
s
is
bec
ause
m
an
y
r
ea
der
s
ne
ed
to
be
place
d
in
a
s
ea
rch
spa
ce.
In
det
e
ct
i
ng
an
obje
c
t,
a
re
ade
r
will
onl
y
r
epor
t
the
signa
l
streng
th
of
th
e
t
a
g
det
e
ct
ed
.
Once
t
he
signal
strength re
port
is obt
ai
n
ed,
the
s
y
st
em w
il
l
compute
the
coor
din
at
es
of
the
RF
ID
ta
gs
base
d
on
ea
ch
dat
a
grouping
.
In
thi
s
pape
r,
al
gorit
hm
s
using
gene
t
ic
al
gor
it
h
m
,
par
ticle
sw
arm,
ant
col
on
y
o
pti
m
iz
ation
are
proposed
to
ac
hi
eve
th
e
shortest
pat
h
f
or
an
RF
ID
m
obil
e
re
ade
r,
whil
e
cove
ring
fu
ll
se
arc
h
ar
ea.
In
co
m
par
ison,
for
p
at
h
opt
imiza
t
ion
,
the
m
obile
rea
der
t
rav
erse
s
from
one
nod
e
to
the
nex
t,
m
oving
aro
und
enc
o
unte
r
ed
obstac
l
es
in
it
s
p
at
h.
The
t
ag
reading
proc
ess
is
i
te
ra
ti
ve
,
in
whic
h
the
rea
d
er
arr
ive
s
at
it
s
star
t
point
at
the
en
d
of
ea
ch
round.
Based
on
the
sh
orte
st
pat
h
,
an
al
gori
thm
that
computes
the
l
oca
t
ion
of
it
ems
in
the
sea
r
ch
ar
ea
is
used
.
The
sim
ula
t
ion
result
s
show
that
the
ACO
m
et
hod
works
m
ore
eff
ectiv
e
l
y
and
eff
icientl
y
c
om
par
e
to
othe
rs
when
solving
sh
orte
st p
at
h
probl
ems
.
Ke
yw
or
ds:
ACO
GA
Path
o
ptim
iz
ati
on
PSO
RFID
Copyright
©
201
9
Instit
ut
e
o
f Ad
vanc
ed
Engi
n
ee
r
ing
and
S
cienc
e
.
Al
l
rights re
serv
ed.
Corres
pond
in
g
Aut
h
or
:
M. Zaki
Zaka
ri
a
,
Faculty
of Com
pu
te
r
an
d
Ma
them
a
ti
cal
Scie
nces,
Un
i
ver
sit
y Tec
hnology
of MARA,
Sh
a
h Alam
, M
al
ay
sia
.
Em
a
il
:
f
m
zaki
@yah
oo.co
m
1.
INTROD
U
CTION
W
it
h
the
gr
ow
t
h
of
t
he
I
nter
ne
t
of
Thi
ng
s
(
I
OT),
the
Ra
dio
Fr
e
qu
e
ncy
Id
e
ntific
at
ion
(RF
ID)
will
be
the
m
os
t
i
m
po
rtant
te
ch
no
l
ogy
to
be
a
dopte
d
an
d
it
will
be
seen
as
t
he
pre
-
requisi
te
for
the
I
OT.
T
he
RFI
D
us
es
ra
dio
wa
ve
s
to
i
den
ti
fy
people
or
ob
j
e
ct
s,
with
ap
plica
ti
on
s
i
n
c
on
t
act
le
ss
paym
e
nt,
asset
m
ana
gem
ent,
trusted
id
entifi
cat
ion
(
ID)
an
d
do
c
um
ent
trackin
g
[
1].
Ba
sed
on
the
m
ark
et
ob
ser
vatio
n
by
Fr
os
t
an
d
S
ulli
van
,
the
Ma
la
ysi
an
RFID
m
ark
et
i
n
2009
was
est
i
m
at
ed
to
be
a
ppr
ox
im
at
ely
RM
36
m
i
ll
ion
and
h
as
bee
n
exp
ect
e
d
to
gro
w
t
o
ab
ou
t
RM
120
m
il
li
on
in
2020.
The
RF
I
D
de
plo
ym
ent
will
gro
w
ra
pid
ly
du
e
to
th
e
ext
e
ns
i
ve
i
m
ple
m
entat
io
n
of
t
he
IO
T
,
t
hu
s
it
is
esse
ntial
fo
r
the
RF
I
D
te
ch
nolo
gy
to
em
bed
the
c
apab
il
it
y
to
tra
ce
any
obj
ect
easi
ly
on
ce
it
is
ta
gg
ed.
I
n
ge
ner
al
,
RFID
is
able
to
ta
g,
trace
an
d
chec
k
an
ob
je
ct
,
and
RFI
D
ta
gs
are
at
ta
ched
to
ob
je
ct
s.
The
ta
gs
captu
red
im
portant
in
form
at
i
on
a
bo
ut
an
obj
ect
,
s
uch
as
it
s
uniq
ue
I
D
num
ber
,
m
anu
fact
ur
e
dat
e and pr
oduc
t com
po
sit
ion
.
RFID
te
ch
nolog
y
us
es
r
adi
o
wa
ves
to
enab
le
the
co
m
m
un
ic
at
ion
betwee
n
reade
rs
an
d
ta
gs
.
The
RFI
D
ta
g
is
a
dev
ic
e
that
stores
a
n
obj
ect
’s
data
s
uch
as
it
s
I
D,
and
tra
ns
m
it
s
it
to
the
reade
r
in
a
con
ta
ct
le
ss
m
a
nn
e
r
us
i
ng
ra
di
o
wa
ves.
O
bj
ect
s
are
identif
ie
d
w
hen
a
ta
g
transm
it
s
this
inf
or
m
at
ion
ba
ck
to
the
rea
der
s
wit
hin
th
e
inter
rogati
on
ra
ng
e
.
The
r
eade
r
is
a
dev
ic
e
t
hat
c
an
rea
d
data
f
r
om
and
w
rite
data
to
Evaluation Warning : The document was created with Spire.PDF for Python.
Ind
on
esi
a
n
J
E
le
c Eng &
Co
m
p
Sci
IS
S
N:
25
02
-
4752
So
lv
in
g RFI
D mobil
e re
ad
er
pa
t
h pr
ob
le
m w
it
h
opti
miza
ti
on a
l
gorit
hm
s
(
M. Z
aki
Za
k
ari
a)
1111
com
patible
RF
ID
ta
gs.
RFID
read
ers
ca
n
be
fixed
(stat
ic
)
,
or
m
ob
il
e
(w
it
hin
a
pr
ede
fi
ned
ph
ysi
cal
area)
.
Objects a
re ide
ntifie
d w
hen tags tra
ns
m
it
this
infor
m
at
ion
ba
ck
to
the
rea
de
rs wit
hin
t
he
i
nterro
gation ra
ng
e
.
Ba
sed
on
the
analy
sis
and
t
he
f
or
ecast
c
onduct
ed
by
IDTec
hE
x
(a
re
s
earch
c
om
pan
y
based
in
the
Un
it
ed
King
dom
wh
ic
h
ha
s
tr
acked
t
he
RFI
D
m
ark
et
since
1999),
RF
ID
will
con
ti
nue
to
be
a
dopted
i
n
retai
l
industry.
I
n
20
16
al
one,
the
dem
and
foreca
ste
d
for
ap
parel
ta
gg
ing
was
4.
6
bill
io
n
RFID
la
bels
whil
e
the
RFID
in
the
form
of
ti
cket
s
us
e
d
f
or
tra
ns
it
w
ould
de
m
and
800
m
i
l
li
on
ta
gs
ready
for
c
onsu
m
ptio
n.
The
ta
gg
i
ng
of
ani
m
al
s
(su
ch
as
pig
s,
sh
ee
p
and
pets)
is
su
bs
ta
ntial
as
it
con
ti
nues
to
be
a
le
gal
req
uir
e
m
ent
in
m
any
m
or
e
te
rr
it
o
ries,
with
42
0
m
i
ll
ion
ta
gs
bei
ng
us
e
d
for
this
sect
or
in
2016.
In
t
otal,
ID
Tec
hE
x
e
xp
ect
s
tho
se
8.9
bill
ion
ta
gs
to
be
s
old
in
2015
a
nd
10.4
bill
ion
i
n
2016.
Ba
se
d
on
the
f
or
ecast
,
RFID
de
plo
ym
ent
in
Ma
la
ysi
a
wil
l
be
the
biggest
po
te
ntial
con
tri
bu
t
or
to
t
he
re
vital
iz
ing
of
th
e
el
ect
ro
nics
and
el
ect
rical
sect
or
a
s
aim
ed
unde
r
th
e
Eco
no
m
ic
Tran
sf
orm
ation
P
la
n.
T
o
ac
hiev
e
this,
it
is
im
p
or
ta
nt
to
ide
ntify
the
key
fact
or
s
i
n
ens
ur
in
g
t
he
a
f
ford
a
bili
ty
f
or
the m
ass i
m
ple
m
entat
ion
.
In
t
his
pap
e
r,
we
pro
po
se
G
en
et
ic
Algo
rithm
(G
A
),
Part
ic
le
Sw
arm
O
pti
m
iz
at
ion
(PSO)
an
d
A
nt
Colo
ny
O
ptim
iz
at
ion
(
ACO
)
te
chn
i
qu
e
s
th
at
inco
rpor
at
e
the
nea
rest
neighb
or
f
or
t
he
m
ov
e
m
ent
of
RFI
D
m
ob
il
e read
ers
. Th
e
opti
m
u
m
p
at
h (i.e.
, sh
ort
est
p
at
h) in
for
m
at
ion
obtai
ne
d
is
us
e
d
to c
om
pu
t
e the loca
ti
on
of
obj
ect
s
within
the g
i
ven area
us
in
g
t
he
tria
ngulati
on tec
hn
i
qu
e
.
It
has
al
ways
be
en
ou
r
intenti
on
t
o
disc
over
the
best
te
ch
ni
qu
e
i
n
locat
in
g
the
RFI
D
rea
de
rs
(
nodes
)
within
a
gi
ven
area
to
guara
nt
ee
100%
co
ve
r
age
by
em
plo
yi
ng
a
m
ini
m
u
m
nu
m
ber
of
r
eaders
.
W
e
de
velo
p
a
prototype
of
t
he
softwa
re
sim
ula
ti
on
to
de
te
rm
ine
how
m
any
RFID
fi
xed
read
e
rs
ar
e
require
d
to
cov
e
r
a
giv
e
n
area
bas
ed
on
he
xago
na
l
pack
in
g
in
our
first
phase
.
On
e
RFI
D
re
ader
is
able
to
cov
e
r
eve
ry
he
xago
n
area
a
nd
t
he
c
ov
e
ra
ge
of
a
wireless
se
nso
r
net
wor
k
m
ay
be
a
ppr
ox
im
a
te
d
by
a
dis
k
of
a
presc
ribe
d
ra
dius
(sen
si
ng
ra
ng
e
).
Co
ve
rag
e
ca
n
be
com
pu
te
d
by
ta
king
the
un
i
on
of
in
div
i
du
al
c
overa
ge
areas
of
al
l
se
ns
or
s
(r
ea
der
s
)
i
n
th
e
netw
ork.
T
his
can
be
pe
rfo
rm
ed
by
assu
m
ing
that
eac
h
ci
rcle
is
a
he
xago
n
wh
ic
h
a
tt
aches
with a
no
t
her.
We ass
um
e all
the r
ea
ders
have the sam
e sen
sing an
d
tra
nsm
issi
on
r
a
nge.
In
t
he
sec
ond
ph
a
se,
the
loc
at
ion
of
t
he
R
FI
D
rea
der
th
at
we
cal
culat
ed
pr
e
viously
is
us
e
d
as
a
ref
e
ren
ce
poin
t
the
m
ob
il
e
r
eader
visit
s
i
n
a
certai
n
per
i
od
of
ti
m
e.
We
pro
posed
to
us
e
t
he
m
ob
il
e
RFI
D
read
e
r
to
re
duce
hardw
a
re
(rea
der)
costs,
a
nd
el
im
inate
read
ers
’
inter
fe
ren
ce
.
The
m
ob
il
e
read
er
is
m
ov
ed
from
on
e
po
i
nt
(n
ode)
t
o
an
oth
e
r
point
ba
sed
on
the
s
hortest
path
com
pu
te
d
us
i
ng
P
SO
an
d
AC
O
for
the
giv
e
n
dim
ension
are
a.
The
ta
g
read
i
ng
proc
ess
by
the
m
ob
il
e
RFID
read
e
r
is
rep
eat
ed
w
hen
the
rea
de
r
arr
ive
s
to it
s start
node
in
the
lo
op. T
hi
s p
r
ocess
r
e
pe
at
ed
in a
n
it
era
ti
ve
fas
hion.
In
the
t
hir
d
phase,
we
sta
rt
t
he
al
go
rithm
with
a
n
i
niti
al
i
zat
ion
process
to
dete
rm
ine
the
nu
m
ber
of
ta
gs
to
be
us
e
d
i
n
the
giv
e
n
area.
T
he
process
co
ntin
ue
s
by
placi
ng
t
ags
at
ra
ndom
in
t
he
sam
e
area.
The
proces
s
c
on
ti
nues
w
hic
h
ta
gs
detect
in
g
by
m
ob
il
e
RFID
rea
de
r
3
t
i
m
es
at
diff
e
re
nt
locat
io
n
or
po
i
nt
in
the
gi
ven
dim
e
ns
io
n
a
rea.
T
he
process
co
ntinu
e
s
by d
et
ect
ing
a
rea
de
r
a
nd
it
is
do
ne
un
t
il
al
l
three
rea
de
rs
ar
e
detect
ed
the
ta
gs
.
The
n
we
con
ti
nue
with
the
cal
culat
io
n
to
de
te
rm
ine
the
posit
io
n
of
ta
gs
by
us
in
g
th
e
equ
at
io
ns
disc
us
se
d
in
Sect
io
n
3
.
T
he
e
quat
ion
is
ba
sed
on
the
data
of
the
receive
d
sig
nal
stren
gth
in
dicat
ion
(RSSI)
t
hat
is
ob
ta
ine
d
duri
ng
the
detect
io
n
process
.
T
his
pap
e
r
pr
e
sents
an
al
gorithm
for
the
RFI
D
m
ob
ile
read
e
r
f
or
pat
h
op
ti
m
iz
ation
usi
ng
GA,
PS
O
and
A
C
O
te
ch
niques
inc
orp
or
at
ing
t
he
nea
r
est
neig
hbor
w
hich
is
the
al
go
rit
hm
is
able
wo
r
k
with
any
desi
gn
area
,
w
hether
it
is
asymme
tric
al
or
sy
m
m
e
tric
al
.
In
add
it
io
n,
the alg
or
it
hm
has t
he
a
bili
ty
to
de
fine
t
he po
si
ti
on
of the
obsta
cl
es b
ase
d on
work
i
ng e
nv
ir
onm
ent.
2.
RELATE
D
W
ORK
S
RFID
te
ch
no
l
ogy
is
us
ed
in
m
any
app
li
cat
ion
s
f
or
e
xam
ple
in
trac
king
or
ide
ntifyi
ng
obj
ect
s
[2
]
m
easur
e
data
of
pro
xim
it
y,
R
ecei
ved
Si
gn
al
Stren
gth
(R
SS)
and
rea
d
rate
of
passive
RF
I
D
ta
gs
by
sim
u
la
ti
on
m
od
el
li
ng
.
T
he
com
par
iso
n
of
local
iz
at
ion
ac
cur
acy
us
in
g
di
ff
ere
nt
al
gorit
hm
s
sh
own
t
ha
t
resu
lt
s
on
ave
rag
e
is
8%.
A
local
iz
at
ion
te
chn
i
que
has
bee
n
pr
opos
e
d
us
i
ng
c
om
bin
at
ion
of
wireless
sens
or
network
(
WSN)
an
d
rad
i
o
f
re
qu
e
nc
y
identific
at
ion
te
chnolo
gy
to
so
lve
the
l
ocali
zat
ion
in
WSN
[3
]
.
U
HF
RFI
D
tra
nspo
nders
a
nd
Directi
on
of
A
rr
ival
(
DoA
)
e
stim
ation
te
c
hniq
ue
has
bee
n
us
e
d
t
o
detect
ta
gs
local
iz
at
io
n
by
filt
erin
g
out
the
carrier
fr
e
quen
cy
and
t
he
sig
nal
is
us
e
d
for
est
i
m
ating
th
e
directi
on
of
arr
ival
[4
]
.
A
local
iz
at
ion
m
et
hods
us
in
g
P
hase
D
iffer
e
nt
of
A
r
rival
(PD
oA)
ha
s
bee
n
pr
opos
e
d
[
5].
Mo
re
over,
var
i
ou
s
ga
ps
in
the
a
ppr
oa
ches
base
d
on
P
D
oA
are
identifie
d.
RFI
D
-
base
d
centrali
zed
co
op
e
rati
ve
local
iz
at
ion
has
bee
n
sugg
e
ste
d
in
indoor
env
i
ronm
ent
[6
]
.
The
m
et
ho
d
is
to
i
m
pr
ov
e
the
posit
ion
i
ng
acc
ur
acy
of
f
ou
r
us
e
rs
m
ov
i
ng
i
n
a
buil
din
g,
by
us
i
ng
the
r
ecei
ved
sig
nal
stren
gth
(RS
S)
m
easur
em
e
nts
f
ro
m
a
nu
m
ber
of
know
n
locat
io
n
RF
I
D
ta
gs
(an
c
hors
)
a
nd
com
bin
ing
it
w
it
h
RSS m
easur
em
ents o
f
m
ob
il
e tags ca
rr
ie
d by them
sel
ves.
The
pri
nci
ple
of
the
La
ndm
a
rc
ap
proa
c
h
w
as
us
e
d
to
intr
oduce
the
c
oncept
of
pro
xim
it
y
m
ap
[7
]
,
wh
ic
h
is
the
w
ho
le
se
ns
in
g
a
rea
is
di
vid
e
d
into
re
gions
,
wh
e
re
eac
h
re
gion
co
rresp
on
ds
to
a
refe
re
nc
e
ta
g.
A
pro
ba
bili
stic
local
iz
at
ion
te
chn
iq
ue
ha
s
been
s
ugge
ste
d
to
detect
the
obj
ect
usi
ng
th
re
e
ste
ps
[
8]
.
First,
the
pro
pa
gation
va
riabl
es
are
cal
ib
rated
usi
ng
on
-
sit
e
ref
e
ren
ce
ta
gs,
an
d
the
n
the
distance
betwe
en
the
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
S
N
:
2502
-
4752
Ind
on
esi
a
n
J
E
le
c Eng &
Co
m
p
Sci,
Vo
l.
13
, N
o.
3
,
Ma
rc
h 201
9
:
111
0
–
11
1
6
1112
obj
ect
(tar
gete
d
ta
g)
a
nd
the
read
e
rs
is
est
im
at
ed
with
a
pro
ba
bili
sti
c
RS
S
m
od
el
.
Finall
y,
the
locat
ion
of
th
e
ta
g
is
determ
ined
by apply
ing a Ba
ye
sia
n
i
nference
tech
nique.
RFID
te
c
hnol
og
y
has
al
so
been
us
e
d
to
gu
i
de
blin
d
pe
op
le
to
fin
d
the
sho
rtest
route
from
their
current
l
ocati
on
to
a
de
sti
na
ti
on
.
This
te
c
hnology
al
so
helps
t
hem
wh
en
they
get
lost
by
a
uto
m
at
ic
al
l
y
detect
ing
the
r
ou
te
,
a
nd
recal
culat
ing
a
ne
w
ro
ute
to
the
sa
m
e
destinat
ion
[9
]
.
The
syst
em
e
m
bed
s
RFID
ta
gs
into
a
f
oo
t
path
that
can
be
rea
d
by
an
RFID
rea
der
with
a
cane
a
nten
na.
The
s
yst
e
m
is
al
so
us
e
d
f
or
nav
i
gation
i
n
r
escue
op
e
rati
ons
in
vo
l
ving
ha
zardo
us
en
vir
on
m
en
ts,
w
here
it
is
diff
ic
ult
to
fin
d
a
n
em
e
rg
e
ncy
exit.
The
i
ndoor
guida
nce
sy
stem
[1
0]
use
s
RFID
te
c
hnol
og
y
to
fin
d
a
gu
i
dan
ce
pat
h
for
im
po
rtant
places,
su
c
h
as m
us
eu
m
s,
ho
sp
it
al
s, a
irp
or
t t
erm
inals an
d
e
xh
i
biti
on h
al
ls.
The
m
os
t
chall
eng
in
g
ta
sk
is
to
achi
eve
th
e
sh
ort
est
path
fo
r
a
netw
ork
by
m
ini
m
iz
i
ng
t
he
cos
t
(d
ist
ance
)
i
ns
ide
a
giv
e
n
a
rea
.
T
he
s
hortest
path
has
bee
n
us
e
d
e
xtensiv
e
ly
fo
r
ot
her
pur
po
s
es,
s
uc
h
as
veh
ic
l
e
routin
g
in
the
t
ran
s
portat
io
n
s
yst
e
m
[1
1],
pat
h
pla
nn
i
ng
i
n
r
obotics
[
12
]
a
nd
traf
fic
r
ou
ti
ng
in
c
omm
un
icati
on
netw
orks
[13].
An
a
ppli
cat
ion
f
or
c
om
plete
cov
e
ra
ge
with
in
a
prede
fine
d
per
i
od
of
ti
m
e
us
in
g
m
ob
il
e
RFID
read
e
rs
in
a
s
ymm
et
rical
ar
ea
has
bee
n
intr
oduce
d
[
14
]
.
The
syst
em
pro
po
se
d
by
[
14
]
only
cat
ered
f
or
sy
m
m
e
tric
al
area,
an
d
wi
thin
the
area
co
ver
e
d
by
RFI
D
m
ob
il
e
read
e
rs,
s
he
assum
es
the
areas
only
loc
at
e
the
pro
per
sta
cke
d
sh
el
ves
.
The
giv
en
a
rea
is
div
i
ded
int
o
m
any
sect
or
s
,
an
d
each
sect
or
is
cat
ered
by
on
e
m
ob
il
e
read
e
r.
T
he
pat
h
show
n
in
[
14
]
is
s
i
m
i
la
r
fo
r
ever
y
sect
or.
[
1
5]
prese
nts
the
inv
est
igati
on
of
PS
O
to
so
l
ve
the
sh
ort
est
pat
h
r
ou
ti
ng
pro
blem
us
ing
a
m
od
ifie
d
pr
i
or
it
y
base
d
in
direct
encodin
g,
i
nc
orp
or
at
in
g
a
he
ur
ist
ic
op
e
rato
r for
re
du
ci
ng the
pos
sibil
it
y of
lo
op
form
ation
i
n
th
e p
at
h co
ns
tr
uc
ti
on
process
.
3.
PATH
OP
TI
MIZ
ATION
The
sho
rtest
pa
th
pro
blem
is
find
i
ng
a
pat
h
between
node
s
in
a
search
s
pace
or
area
s
uch
that
the
su
m
of
the
wei
gh
ts
or
distanc
e
of
e
dg
e
s
is
m
ini
m
iz
ed.
The
pro
blem
of
fin
ding
the
m
ini
m
u
m
distance
for
this
case
can
be
m
od
el
e
d
by
ass
um
ing
that
eac
h
node
i
n
a
search
s
pace
(s
uch
that
the
s
um
of
distanc
e
of
it
s
const
it
uen
t
e
dges)
is
m
ini
m
iz
e
d.
Th
e
re
a
re
m
any
resea
rc
hers
that hav
e
bee
n
done
i
n
t
he
s
hortest
pat
h
prob
le
m
bu
t
no
t
in
pat
h
opti
m
iz
at
ion
for
m
ob
il
e
RFID
rea
der
[16].
A
sim
ulatio
n
is
c
reated
to
te
st
for
the
pat
h
op
ti
m
iz
ation
s
uch as
ACO
, PSO a
nd
GA. Si
m
ula
ti
on
s a
re c
onduct
ed
t
o
c
om
par
e resu
lt
s
f
or all
techn
i
ques.
3.1.
Genetic
Algor
ithm
I
n
this
w
ork,
a
basic
G
A
is
em
plo
ye
d
to
so
lve
the
be
nc
hm
ark
pr
ob
le
m
fo
r
path
optim
iz
ation
of
m
ob
il
e
RFID
read
er
.
The
GA
re
presents
the
design
va
riable
with
nodes
num
ber
that
ref
err
e
d
to
as
chrom
os
om
es. Th
e GA
wor
ks for a co
m
bin
at
ion
of
discrete
par
am
et
ers
and also for
ces t
he
d
esi
gn v
a
riabl
es to
on
ly
ta
ke
val
ues
with
no
duplica
te
value
s
i.e.
no
so
l
ution
is
ever
pr
oduc
e
the
du
pl
ic
at
e
no
de
in
new
chrom
os
om
es.
On
e
popula
ti
on
from
so
luti
on
s
are
us
e
d
to
form
a
new
popu
la
ti
on
with
the
hope
that
th
e
ne
w
popula
ti
on
m
ay
be
bette
r
th
an
the
old
one.
The
popula
ti
on
w
hich
is
sel
ect
ed
to
f
or
m
new
s
olut
ion
s
(chro
m
os
om
e
@
offs
pr
i
ng)
m
us
t
m
eet
the
fitness
crit
eria
in
w
hich,
t
he
fitt
er
they
are,
the
m
or
e
cha
nces
to
reprod
uce.
Thi
s
proces
s
re
pe
at
s
con
ti
nuous
ly
un
ti
l
the
c
onditi
on
is
sat
isfie
d
s
uch
as
achievi
ng
the
total
nu
m
ber
of it
er
at
ion
.
The
process
sta
rts
f
ro
m
a
po
pu
la
ti
on
of
ra
ndom
ly
gen
erat
ed
in
div
i
du
al
a
nd
each
it
erati
on
is
cal
le
d
a
gen
e
rati
on.
I
n
ever
y
ge
ne
rati
on,
the
fitne
ss
of
eac
h
ch
r
omoso
m
e
is
cal
culat
ed
base
d
on
obj
ect
ive
funct
ion
in
path
op
ti
m
iz
a
t
ion
(m
ini
m
u
m
path)
for
a
m
ob
il
e
RFI
D
re
ader.
The
new
gen
e
rati
on
of
chrom
os
om
e
si
m
ply
cro
ss
ed
to
pro
du
ce
offs
pr
in
g
tou
r
wh
ic
h
re
su
lt
s
in
il
le
gally
tou
r
i.e.
to
urs
wh
ic
h
visit
so
m
e
no
des
m
or
e
than
on
e
tim
e.
To
de
al
with
this
prob
le
m
,
the
offs
pr
i
ng
t
ours
a
re
correct
ed
s
o
t
ha
t
the
duplica
t
e
nodes
are
repl
aced
by
un
visit
ed
node
s
at
ra
ndom
.
The
proces
s
te
rm
inate
s
wh
e
n
a
m
axim
u
m
nu
m
ber
of
it
erati
on
ha
s
be
e
n
pro
du
ce
d or t
he
r
es
ult o
f
fit
ne
ss fun
ct
i
on is s
at
isfact
or
y.
In
t
he
fi
rst
ca
se,
the
al
gorithm
ran
dom
ly
init
ia
li
zes
popu
la
ti
on
of
c
hrom
os
om
e.
For
exam
ple,
ei
gh
t
(
8)
locat
ion
s
(
nodes
)
of
a
searc
hed
sp
a
ce
are
ide
ntifie
d
in
w
hich
th
e
m
ob
il
e
RFID
r
eader
will
be
vi
sit
ed
no
t
m
or
e
than
on
ce
an
d
the
nu
m
ber
of
popula
ti
on
a
re
five.
Nex
t
ste
p
is
to
app
ly
the
near
est
ne
ighbou
r
const
raint
in
w
hich
t
o
e
ns
ure
that
the
nex
t
node
to
go
is
in
the
s
urrou
nd
i
ng
li
st
of
cu
rr
e
nt
no
des.
With
thi
s
app
li
cat
io
n,
use
r
can
ch
oose
a
le
vel
of
the
ne
arest
neig
hbour
.
For
le
vel
1,
su
r
rou
nd
i
ng
node
s
is
within
rad
i
us
of
c
urre
nt
no
de
co
ver
a
ge
(m
ob
il
e
read
e
r
c
ov
e
ra
ge)
but
f
or
le
vel
2
s
urr
oundin
g
node
is
double
of
r
eader
cov
e
ra
ge
(no
de
).
The
reas
on
to
im
ple
m
e
nt
the
le
vel
of
the
ne
arest
neig
hbour
is
t
o
c
on
t
ro
l
t
he
ne
w
chrom
os
om
e w
it
h
the
valid
pa
th r
es
ults.
Ba
sed
on
the
s
olu
ti
ons
s
hows
on
Ta
ble
4,
there
a
re
f
our
(
4)
valid
po
pu
l
at
ion
s
w
hic
h
a
re
al
l
node
s
visit
ed
only
on
ce
and
t
her
e
is
on
e
s
ol
ution
t
hat
is
inv
al
id
because
the
s
urr
oundin
g
no
de
s
(n
ei
ghbour
nodes
)
hav
e
bee
n
visit
ed
m
or
e
than
on
ce
.
Chrom
os
om
e
2
and
3
ha
ve
the
best
va
lue
for
fitness
functi
on
(m
ini
m
u
m
distance)
an
d
t
hese
c
hrom
os
om
es
are
then
be
ing
treat
e
d
as
par
e
nt
f
or
the
nex
t
it
erati
on.
The
nex
t
ste
p
is
to
cro
ss
over
by
c
reati
ng
ne
w
po
pu
la
ti
on
w
hic
h
us
es
on
e
-
po
i
nt
cro
ss
over
te
c
hn
i
qu
e
.
I
n
this
ph
a
se,
the
off
sp
ri
ng
Evaluation Warning : The document was created with Spire.PDF for Python.
Ind
on
esi
a
n
J
E
le
c Eng &
Co
m
p
Sci
IS
S
N:
25
02
-
4752
So
lv
in
g RFI
D mobil
e re
ad
er
pa
t
h pr
ob
le
m w
it
h
opti
miza
ti
on a
l
gorit
hm
s
(
M. Z
aki
Za
k
ari
a)
1113
will
be
check
e
d
so
that
no
duplica
te
value
in
the
op
e
rato
r
s.
Duplic
at
e
value
will
be
rep
la
ced
by
un
visit
ed
nodes
at
rand
om
.
Nex
t
ph
ase
is
m
utati
on
pro
cess
in
wh
ic
h
the
ope
rato
r
is
r
andom
ly
picked
.
T
he
m
ai
n
reason
for
m
utatio
n
is
to
pr
eser
ve
th
e
gen
et
ic
div
e
r
sit
y
(ex
plo
it
at
ion)
of
the
popula
ti
on
.
T
he
pr
ocess
co
ntin
ue
s
un
ti
l
the
te
rm
inati
on
crit
eria
are
sat
isfie
d
an
d
wh
e
n
s
pecific
nu
m
ber
of
it
erati
on
has
been
exceede
d.
Alth
ou
gh
so
m
et
i
m
es
the
pe
rfor
m
ance
of
an
al
gorith
m
is
ob
ser
ve
d
to
be
sat
isfac
tory,
sti
ll
it
enco
unte
rs
s
om
e
fa
ults,
su
c
h
as t
he
loc
al
search abil
it
y and t
he ge
net
ic
algorit
hm
d
oe
s not ass
ur
e
to
f
in
d
a
g
l
ob
al
optim
u
m
.
3.2.
Part
ic
le
Sw
am
Op
timi
z
at
ion
Partic
le
swarm
op
tim
iz
at
io
n
(PSO
)
is
a
popula
ti
on
-
bas
ed
stoc
hastic
op
ti
m
iz
ation
te
chn
i
qu
e
t
ha
t
or
i
gin
at
es from
n
at
ure an
d
e
voluti
onary c
ompu
ta
ti
ons, de
ve
lop
e
d
by
[17].
The
P
SO
is a
m
et
ho
d
th
at
is able to
identify
an
op
tim
a
l
so
luti
on
within
a
po
pula
ti
on
(i.e.
,
a
swar
m
).
T
he
PSO
al
gorith
m
flow
co
ns
is
ts
of
a
popula
ti
on
of
i
nd
i
viduals
refe
rr
e
d
to
as
“pa
rt
ic
le
s”.
Ever
y
pa
rtic
le
is
a
po
te
ntial
so
luti
on
to
an
n
-
dim
ens
ion
al
pro
blem
.
The
gro
up
ca
n
achi
eve
the
s
olu
ti
on
ef
fec
ti
vely
by
us
ing
c
omm
on
i
nfor
m
at
ion
sh
are
d
by
the
gro
up,
and
the
i
nfor
m
at
ion
is
owne
d
by
the
par
ti
cl
e
it
sel
f.
The
pa
rtic
le
s
change
their
sta
te
by
“fly
ing
”
ar
ound
in
an
n
-
dim
ension
al
searc
h
s
pace
base
d
on
t
he
velocit
y,
upda
te
d
unti
l
a
re
la
ti
vely
un
cha
ng
i
ng
s
ta
te
ha
s
bee
n
encou
ntere
d,
or
un
ti
l com
pu
ta
ti
on
al
li
m
it
ation
s a
re e
xceede
d.
Fo
r
al
l
it
erati
ons,
any
pa
rtic
le
is
up
dated
by
fo
ll
owin
g
t
w
o
"
best"
value
s.
T
he
first
one
is
the
best
so
luti
on
(
fitne
ss)
it
has
ac
hieve
d
s
o
fa
r,
ref
e
rr
e
d
to
as
pbest
,
an
d
the
fitn
ess
va
lue
is
al
so
s
tore
d.
Anothe
r
"
best"
value
t
hat
is
tracke
d
by
the
par
ti
cl
e
swa
rm
op
ti
m
iz
er
is
t
he
be
st
value
obta
ined
so
far
by
any
par
ti
cl
e in t
he pop
ulati
on
,
wh
ic
h
is cal
le
d gl
ob
al
best
(gbes
t).
In
P
SO
path
pl
ann
in
g
op
ti
m
i
zat
ion
,
eac
h
pa
rtic
le
of
a
pa
rtic
le
swar
m
rep
rese
nts
a
sin
gle
possibl
e
path
that
c
ou
l
d
be
ta
ke
n
by
th
e
m
ob
il
e
RFID
read
e
r.
Hen
ce
,
the
pa
rtic
le
swar
m
at
tem
pts
to
fin
d
the
optim
al
path
for
th
e
m
ob
il
e
RF
ID
re
ader
t
o
m
ov
e
from
on
e
no
de
to
an
othe
r.
T
his
te
ch
nique
r
andom
ly
gen
er
at
es
a
pop
ulati
on
of
par
ti
cl
es
withi
n
the
giv
e
n
di
m
ension
a
rea.
Ther
e
f
or
e,
t
he
swar
m
at
tem
pt
s
to
fin
d
the
optim
a
l
path
t
hat w
il
l
be
u
se
d for t
he m
ob
il
e RFID
r
eader
to
m
ov
e
to ev
e
ry sin
gle
node
i
n
the
g
i
ve
n
a
rea.
The pse
udo
c
ode
of the
PS
O al
gorithm
can
be
s
umm
arised as
f
ollo
ws:
Step
1
:
In
it
ia
li
ze par
a
m
et
ers
c
1
,
c
2
an
d
ω
for t
he pa
rtic
le
s
.
Step
2
:
In
it
ia
li
ze rand
om
p
os
it
ion
s
for
all
d
im
ension
each
par
ti
cl
e a
nd their
ass
ociat
ed
vel
ociti
es.
Step
3
:
Evaluate t
he fit
ness funct
io
n f
or each
p
a
rtic
le
.
Step
4
:
C
heck crit
erio
n
te
rm
inati
on
ba
sed o
n nu
m
ber
of
it
erati
on.
Step
5
:
Update
velocit
ie
s and
posit
ion base
d o
n near
est
n
ei
gh
bour c
on
st
raint.
Step
6
:
Update
l
ocal
be
st
(
pb
est
)
,
w
hi
ch
com
par
es
the
c
urren
t
val
ue
of
fitness
functi
on
wit
h
th
e
previ
ous
best
value of
th
e p
a
rtic
le
s.
Step
7
:
Update
gl
ob
al
best
(
gb
e
st)
,
wh
ic
h
determ
ines
the
cu
rr
e
nt
gl
ob
al
m
ini
m
u
m
fitness
value
a
m
on
g
the
current
posit
io
ns
of the
pa
rtic
le
s.
Step
8
:
Apply
sta
te
tra
ns
it
ion
bas
ed
on
near
est
nei
ghbo
ur
co
ns
trai
nt
an
d
the
proc
ess
will
c
on
ti
nu
e
ti
ll
en
d
conditi
on whe
n
m
axi
m
u
m
n
um
ber
o
f
it
erati
on
s
is re
ache
d.
3.3.
An
t
Colo
ny O
pt
im
iz
at
ion
The
a
nt
c
olony
opti
m
iz
at
i
on
(
ACO
)
is
a
m
et
a
-
heu
r
ist
ic
op
ti
m
iz
a
ti
on
te
ch
ni
que
f
or
ha
r
d
com
bin
at
or
ia
l
op
ti
m
iz
ation
pro
blem
s.
We
e
xam
ine
the
ant
colo
ny
syst
e
m
(A
CS)
as
a
represe
ntati
ve
of
the
ACO
te
c
hniq
ue
. T
her
e a
re t
hree (
3)
key as
pe
ct
s p
ertai
ni
ng t
o
the
A
C
S that
shou
l
d be
unde
rstood:
a)
the
sta
te
trans
it
ion
r
ule
pro
vid
es
a
direct
way
to
ba
la
nce
bet
wee
n
exp
l
or
at
io
n
of
new
e
dges
a
nd
exp
l
oitat
ion
of
a priori a
nd ac
cum
ulate
d
knowle
dge a
bout t
he pr
oble
m
;
b)
the g
l
ob
al
up
da
ti
ng
is a
ppli
ed
only
to
e
dg
e
s
wh
ic
h belo
ng t
o
the
b
e
st ant t
our; a
nd
c)
wh
il
e a
nts c
onstruct a
so
l
utio
n,
a
local
pher
om
on
e up
datin
g ru
le
s a
ppli
ed.
Ba
sic
al
ly
,
fo
r
the
AC
S
proc
ess,
m
ants
ar
e
init
ia
ll
y
pu
t
on
n
node
s
c
ho
s
en
acco
r
din
g
to
so
m
e
init
ia
li
zation
r
ule
(e.
g.,
ra
nd
om
l
y).
Each
a
nt
buil
ds
a
path
b
y
rep
eat
e
dly
app
ly
ing
t
he
sta
te
transiti
on
r
ule
.
Durin
g
co
ns
tr
uc
ti
ng
it
s
path
,
an
ant
m
akes
so
m
e
chan
ges
t
o
the
am
ou
nt
of
ph
e
r
om
on
es
on
t
he
visit
ed
edg
e
s
by
ap
plyi
ng
th
e
local
up
datin
g
r
ule.
O
nce
a
ll
ants
ha
ve
finish
e
d
t
heir
pa
th,
the
am
ou
nt
of
phe
ro
m
on
e
s
on
edg
e
s
is
m
od
if
ie
d
ag
ai
n
by
a
pp
ly
in
g
t
he
global
up
datin
g
r
ul
e.
The
pat
h
wi
th
a
high
am
ount
of
pher
om
on
es
is
pr
e
ferred
b
y t
he
an
ts,
h
e
nce,
t
hey w
il
l ch
oos
e the s
hort
path
.
The
m
ai
n
goal
of
th
e
loc
al
update
is
to
di
ver
si
fy
th
e
searc
h
by
decr
easi
ng
the
ph
e
r
om
on
e
con
ce
ntrati
on
on the tra
ve
rse
d
e
dg
es
. T
hus, t
he
ant
w
ou
l
d
c
hoos
e a
nothe
r route
to p
rod
uc
e d
if
fer
e
nt s
olut
ion
s.
This
w
ou
l
d
pr
even
t
se
ver
al
ants
to
pro
duc
e
identic
al
so
luti
ons
duri
ng
an
it
erati
on.
The
gl
ob
al
phe
r
om
on
e
update
is
ap
plied
at
the
en
d
of
ea
ch
it
erati
on
to
on
e
a
nt,
w
hi
ch
can
be
ei
th
er
the
it
erati
on
-
best
a
nt
or
th
e
best
-
so
-
far
a
nt.
T
he
ph
e
ro
m
on
e
updatin
g
ru
le
s
a
re
desi
gn
e
d
s
o
that
they
te
nd
to
giv
e
m
or
e
ph
e
r
om
on
e
to
edg
e
s
wh
ic
h
s
houl
d b
e v
isi
te
d by a
nts.
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
S
N
:
2502
-
4752
Ind
on
esi
a
n
J
E
le
c Eng &
Co
m
p
Sci,
Vo
l.
13
, N
o.
3
,
Ma
rc
h 201
9
:
111
0
–
11
1
6
1114
The pse
udo
c
ode
of the
ACO
is as f
ollo
ws:
Step
1
:
In
it
ia
li
ze par
a
m
et
er t,
α
, ρ, q
0,
nk, a
nd calc
ulate
τo
Step
2
:
Fo
r
e
ver
y l
in
k (r
,s
), pe
rfo
rm
t
he ph
e
r
om
on
e
init
ia
li
zation
τ
(
r,
s
)
=
τ
0
;
Step
3
:
In
it
ia
li
ze nu
m
ber
of
it
erati
ons
and ap
ply st
at
e transiti
on
ru
le
base
d on nei
ghbor
c
onstrai
nt
Step
4
:
Apply l
ocal
phero
m
on
e
updat
e unti
l al
l ants
hav
e
buil
t a c
om
ple
te
so
luti
on
Step
5
:
Ca
lc
ulate
f
it
ne
ss
Step
6
:
Check
crite
rio
n
te
rm
inati
on
ba
sed o
n nu
m
ber
of
it
erati
on
Step
7
:
Apply gl
ob
al
pher
om
on
e
upda
ti
ng
for
the
be
st solutio
n p
rodu
ce
d by a
nts
Step
8
:
Apply
sta
te
tra
ns
it
ion
bas
ed
on
near
est
nei
ghbo
ur
co
ns
trai
nt
an
d
the
proc
ess
will
c
on
ti
nue
ti
ll
en
d
conditi
on whe
n
m
axi
m
u
m
n
um
ber
o
f
it
erati
on
s
is re
ache
d.
4.
SIMULATI
O
N
SET
UP
Nu
m
erous
re
s
earch
act
i
viti
es
ha
ve
been
pro
posed
in
loc
al
iz
at
ion
an
d
po
s
it
io
ning
a
ppli
cat
ion
s
of
RFID
rea
der
s
and
ta
gs
[18].
In
ge
ner
al
,
on
e
of
t
he
m
os
t
fun
dam
ental
is
su
es
in
wireles
s
sens
or
net
work
s
i
s
read
e
r
co
ve
rage.
Eve
ry
po
i
nt
of
the
sel
ect
e
d
area
m
us
t
be
within
the
s
ensin
g
ra
nge
of
at
le
ast
on
e
sens
or
netw
ork.
I
n
th
e
li
t
eratur
e,
so
l
utions
to
this
pr
oble
m
hav
e
be
en
pro
posed
i
n
m
any
diff
eren
t
ways,
e.g.,
usi
ng
a
m
ob
il
e
read
er
[7
]
.
T
he
co
verage
of
a
wirele
ss
sens
or
netw
ork
m
ay
be
app
r
oxim
a
te
d
by
a
disk
of
a
pre
scribe
d
rad
i
us
(se
ns
in
g
ra
ng
e
).
C
overag
e
ca
n
be
c
om
pu
te
d
by
ta
kin
g
the
unio
n
of
in
div
i
du
al
cov
e
ra
ge
areas
of
al
l
sens
or
s
(r
ea
ders)
in
the
netw
ork.
In
this
e
xperim
ent,
the
nu
m
ber
of
no
de
s
dep
e
nds
on
the
dim
ension
of
the
area
an
d
interr
og
at
io
n
ra
nge
that
we
set
for
the
m
ob
il
e
RFID
rea
der
.
If
th
e
di
m
ension
of
area
is
la
rg
e
then
th
e
nu
m
ber
of
no
de
s
that
a
m
ob
il
e
RFID
rea
de
r
trave
rses
is
increase
d.
In
oth
er
words,
the
interrogati
on
range
influ
e
nces t
he nu
m
ber
of no
de
s the m
ob
il
e
RFID rea
der vi
sit
s.
The
siz
e
of
a
hex
a
gon
is
bas
ed
the
inte
rrogat
ion
of
a
rea
de
r.
T
o
ad
dress
this
pro
blem
,
t
he
softwa
re
si
m
ulati
on
that
is
dev
el
op
e
d
can
def
i
ne
th
e
read
ra
ng
e
of
the
RFI
D
r
eader
a
nd
ob
st
acl
es
inside
the
area
cov
e
re
d,
ba
sed
on
use
r
re
quir
e
m
ent.
In
this
exp
e
rim
e
nt,
we
sta
rt
the
al
gorithm
with
an
i
niti
al
iz
at
ion
process
to
de
te
rm
ine
the
nu
m
ber
of
read
e
rs
t
o
be
us
e
d
in
a
giv
e
n
a
rea,
an
d
t
he
syst
em
op
ti
m
iz
es
the
nu
m
ber
of
read
e
rs
base
d
on
t
he
are
a.
U
sing
G
A,
P
SO
and
AC
O
te
ch
niques,
t
he
RF
ID
m
ob
il
e
rea
der
sca
ns
e
ve
r
y
sing
le
node
insi
de
th
e
area.
G
A,
P
S
O
an
d
AC
O
at
tem
pt
to
find
a
path
to
c
om
pl
et
e
a
ci
rcle,
in
wh
ic
h
the
m
ini
m
u
m
path
(d
ist
anc
e)
is
cho
se
n
as
t
he
op
ti
m
al
path.
W
e
r
un
eve
ry
al
gorithm
su
ccessi
vely
for
10
0
an
d
500
it
era
ti
on
s
unde
r
the
sam
e
init
ia
li
zat
ion
co
ndi
ti
ons,
t
hen rec
ord
the
m
ini
m
u
m
an
d
m
axi
m
u
m
p
at
hs
fo
r
PSO a
nd A
C
O
.
5.
RESU
LT
A
N
D DIS
CUSSI
ON
We
us
ed
t
he
f
ollow
i
ng
co
nf
i
gurati
on
f
or
the
PS
O
a
nd
ACO
al
gorith
m
.
Fo
r
the
P
S
O
al
gorithm
,
the
num
ber
of
par
ti
cl
e
is
10,
t
he
c
ogniti
on
fa
ct
or
c
1
a
nd
s
oc
ia
l
factor
c
2
a
re
1.4
,
an
d
t
he
inerti
a
wei
gh
t
w
is
0.4
to
0.9.
Wh
ereas
f
or
the
A
CO
al
gorithm
,
the
po
pu
la
ti
on
(i.e.,
num
ber
of
ants
)
is
10,
ρ
and
ζ
are
0.1,
β
is
2
and q0 is
0.9.
All t
he
al
go
rith
m
s w
ere run 1
00 a
nd
500 t
i
m
es, as
s
how
n
i
n Table
1
a
nd
T
able
2.
Table
1.
T
he
Re
su
lt
of
Op
ti
m
i
zat
ion
Pat
h
us
i
ng GA,
PSO a
nd A
C
O for
150
Nodes
an
d 5
Ob
sta
cl
es
Metho
d
GA
PSO
ACO
#
Ru
n
n
in
g
100
500
100
500
100
500
100
500
100
500
100
500
#
iter
atio
n
100
100
500
500
100
100
500
500
100
100
500
500
Max p
ath
7
9
5
0
.3
7
9
6
6
.5
7
5
3
3
.1
7
5
1
3
.6
7
9
3
8
.3
7
9
5
6
.4
7
5
3
3
.1
7
5
3
.6
6
9
9
9
.9
6
9
6
6
.9
6
5
0
9
.9
6
3
1
4
.0
Min Path
4
8
3
3
.0
4
7
9
1
.3
4
7
1
3
.2
4
6
6
1
.8
4
7
2
1
.2
4
6
1
1
.8
4
7
0
0
.2
4
6
6
1
.8
4
6
1
1
.8
4
6
1
1
.8
4
6
6
1
.8
4
6
6
1
.8
Av
g
Path
5
8
3
1
.6
5
8
6
1
.9
5
5
2
4
.7
5
5
2
1
.4
5
8
3
1
.6
5
8
6
1
.9
5
5
2
4
.7
5
5
2
1
.4
5
6
1
3
.1
5
5
4
3
.4
5
4
2
5
.0
5
2
2
1
.4
Std
Dev
3
3
1
.7
3
2
0
.8
2
8
0
.8
2
5
1
.3
3
3
1
.7
3
2
0
.8
2
8
0
.8
2
5
1
.3
3
1
8
.2
3
1
1
.1
2
6
8
.8
2
1
9
.5
Table
2.
T
he
P
ercenta
ge
of
O
pti
m
iz
at
ion
Path
us
in
g GA f
or 150
N
odes
a
nd
5
Ob
sta
cl
es
#
Ru
n
n
in
g
100
500
100
500
#
I
teration
100
100
500
500
%
%
%
%
Un
co
m
p
let
ed
5831
5
8
.31
2
8
9
8
2
5
7
.96
2
8
4
9
3
5
6
.99
1
4
0
9
3
2
5
6
.37
Co
m
p
leted
FALSE
2648
2
6
.48
1
2
8
9
3
2
5
.79
1
2
0
9
3
2
4
.19
6
0
2
8
3
2
4
.11
TRUE
1521
8125
1
0
2
9
8
1
6
.25
9414
1
8
.83
4
8
7
8
5
1
9
.51
Total
1
0
0
0
0
1
0
0
.00
5
0
0
0
0
100
5
0
0
0
0
100
2
5
0
0
0
0
100
Max Co
n
v
at
#
100
100
100%
0
.2
489
0
.98
491
0
.20
Min Co
n
v
at
#
70
0
.70
75
0
.15
150
0
.30
145
0
.60
Av
g
Co
n
v
9
6
.43
0
.96
9
2
.32
0
.18
3
9
8
.92
0
.80
3
8
2
.32
0
.15
Std
Dev Co
n
v
2
0
.91
1
9
.29
2
9
0
.21
2
8
0
.25
Evaluation Warning : The document was created with Spire.PDF for Python.
Ind
on
esi
a
n
J
E
le
c Eng &
Co
m
p
Sci
IS
S
N:
25
02
-
4752
So
lv
in
g RFI
D mobil
e re
ad
er
pa
t
h pr
ob
le
m w
it
h
opti
miza
ti
on a
l
gorit
hm
s
(
M. Z
aki
Za
k
ari
a)
1115
The
resu
lt
s
of
the
tw
o
al
gorith
m
s
sh
ow
that
t
he
ACO
al
gori
thm
sign
ific
antly
ou
tpe
rfo
rm
e
d
the
PS
O
al
gorithm
.
This
dem
on
strat
es
that
the
ACO
al
go
rit
hm
has
faster
co
nver
gen
ce
a
nd
ada
ptabili
ty
.
Si
m
ulati
on
resu
lt
s
s
how
that
f
or
the
AC
O
al
go
rithm
,
t
he
m
ini
m
u
m
path
(
best
value
)
is
4611.
82
f
or
100
runs
(a
nd
10
0
it
erati
on
s
f
or
e
ach
r
un).
The
com
plete
d
and
true
path
for
ACO
is
higher
than
t
he
PS
O
as
show
n
i
n
T
able
3
and Ta
ble 4.
Table
3.
T
he
P
ercenta
ge
of
O
pti
m
iz
at
ion
Path
us
in
g PSO
fo
r 150
Nodes
an
d 5
Ob
sta
cl
es
#
Ru
n
n
in
g
100
500
100
500
#
I
teration
100
100
500
500
%
%
%
%
Un
co
m
p
let
ed
5552
5
5
.52
%
2
6
7
8
1
5
3
.56
%
2
6
1
3
1
5
2
.26
%
1
2
0
0
2
3
4
8
.01
%
Co
m
p
leted
FALSE
2401
2
4
.01
%
1
2
9
2
1
2
5
.84
%
1
3
3
9
3
2
6
.79
%
6
0
2
8
3
2
4
.11
%
TRUE
2047
2
0
.47
%
1
0
2
9
8
2
0
.60
%
1
0
4
7
6
2
0
.95
%
6
9
6
9
4
2
7
.88
%
Total
1
0
0
0
0
1
0
0
.00
%
5
0
0
0
0
1
0
0
.00
%
5
0
0
0
0
1
0
0
.00
%
2
5
0
0
0
0
1
0
0
.00
%
Max Co
n
v
at
#
100
100%
100
100%
489
98%
491
98%
Min Co
n
v
at
#
60
60%
60
60%
125
25%
147
29%
Av
g
Co
n
v
9
0
.21
9
0
.21
%
8
2
.32
8
2
.32
%
3
4
5
.32
6
9
.06
%
3
2
0
.26
6
4
.05
%
Std
Dev Co
n
v
1
5
.41
1
5
.34
2
5
0
.77
2
5
0
.45
Table
4.
T
he
P
ercenta
ge
of
O
pti
m
iz
at
ion
Path
us
in
g ACO
for 1
50
N
odes
a
nd 5
O
bs
ta
cl
es
#
Ru
n
n
in
g
100
500
100
500
#
I
teration
100
100
500
500
%
%
%
%
Un
co
m
p
let
ed
4712
4
7
.12
%
2
2
0
3
9
4
4
.08
%
2
1
2
3
4
4
2
.47
%
1
0
2
9
3
2
4
1
.17
%
Co
m
p
leted
FALSE
2375
2
3
.75
%
1
3
2
8
1
2
6
.56
%
1
2
7
9
3
2
5
.59
%
6
3
2
8
3
2
5
.31
%
TRUE
2913
2
9
.13
%
1
4
6
8
0
2
9
.36
%
1
5
9
7
3
3
1
.95
%
8
3
7
8
5
3
3
.51
%
Total
1
0
0
0
0
1
0
0
.00
%
5
0
0
0
0
1
0
0
.00
%
5
0
0
0
0
1
0
0
.00
%
2
5
0
0
0
0
1
0
0
.00
%
Max Co
n
v
at
#
100
100%
100
100%
489
98%
491
98%
Min Co
n
v
at
#
58
58%
56
56%
102
20%
90
18%
Av
g
Co
n
v
8
6
.43
8
6
.43
%
8
2
.32
8
2
.32
%
2
5
0
.3
5
0
.06
%
2
3
2
.01
4
6
.40
%
Std
Dev Co
n
v
1
5
.15
1
3
.44
2
3
1
.11
2
1
3
.44
6.
CONCL
US
I
O
N
This p
a
per
has pr
ese
nted
So
l
vi
ng
RFI
D
Mo
bi
le
Reader
Path
Pr
oble
m
w
it
h
Op
ti
m
iz
ation
A
lg
or
it
hm
s
.
W
e
pr
opos
e
d
a
lgorit
hm
s
fo
r
a
n
RFI
D
m
ob
il
e
read
e
r
to
opti
m
iz
e
the
path
a
nd
t
o
c
ov
e
r
a
gi
ven
a
rea
by
putt
ing
on
the
G
A,
P
S
O
a
nd
AC
O
te
chn
i
qu
e
s.
T
he
s
e
te
chn
i
qu
e
s
s
hould
m
eet
the
obj
ect
ive
w
he
n
we
set
the
n
to
the
near
est
neig
hbor
in
w
hich
th
e
m
ob
il
e
read
e
r
will
no
t
be
di
ver
te
d
to
a
l
onge
r
pat
h
in
or
der
to
ac
hieve
bette
r
resu
lt
s.
A
fter
dev
el
op
m
ent
of
a
sim
ulati
on
prototype
,
a
nd
evaluati
ve
ex
per
im
entat
ion
of
t
he
prototy
pe,
t
he
feasibil
it
y
and
eff
ect
iven
ess
of
the
ACO
a
lgorit
hm
s
ign
ific
antly
ou
tpe
r
form
the
GA
and
P
SO
al
go
rithm
.
The
re
su
lt
s
s
how
that
the A
C
O
al
gorithm
achieve p
r
om
isi
ng
res
ults,
a
nd
has
com
petit
ive
po
te
ntial
for
sol
vin
g
discrete
opti
m
i
zat
ion
pr
ob
le
m
s
li
ke
path
optim
iz
ation
com
par
e
d
to
t
he
P
SO
al
go
rithm
.
As
rec
omm
en
datio
n
for
f
uture
w
or
k,
it
w
ou
l
d
be
worthy
to
f
ur
t
her
in
vestig
at
e
wh
et
he
r
th
e
u
se
of
bo
t
h
the
PS
O
an
d
ACO
al
gorithm
s
in
com
bin
at
ion
w
ould
achie
ve
be
tt
er
accuracy.
We
plan
to
f
ur
ther
in
vestigat
e
the
fu
nc
ti
on
a
li
ty
of
this com
bin
at
ion ap
proac
h
a
nd
pr
ese
nt
our
e
xp
e
rim
ental
r
esults i
n
a
s
ub
se
qu
e
nt
pap
e
r.
REFERE
NCE
S
[1]
P.
M.
D.
,
“
Ec
o
nom
ic
Tra
nsfor
m
at
ion
Program
m
e
(ET
P)
Annual
Report
,”
Putr
aj
a
y
a
,
Jab
at
an
P
erd
ana
Men
te
ri
,
2012
.
[2]
Y
.
B.
Gim
pil
evic
and
D.
A.
Savochki
n,
“
Sim
ula
ti
on
of
m
ea
suri
ng
dat
a
obtaine
d
from
R
FID
-
ta
gs
in
sy
stems
of
spati
al l
o
ca
l
ization of
obj
ects
,”
S
pringer
Journal Com
pute
r
,
2016
.
[3]
J
.
Shen,
et
al
.
,
“
A
R
FID
Bas
ed
Local
i
za
t
ion
Algorit
hm
for
W
ire
le
ss
Sensor
Networks
,”
Springer
Journal
Computer
,
2016
.
[4]
A
.
As
che
r,
et
al
.
,
“
Loc
a
li
z
at
ion
of
UH
F
RF
ID
Tra
nsponders
reg
ard
ing
Industr
y
4.
0
Scena
rios
using
Dire
ction
of
Arriva
l
(DoA
) E
stim
at
i
on
Te
chn
i
ques
,”
IEEE
j
ou
rnal
,
2016
.
[5]
O
.
Friede
w
al
d
a
nd
M
.
La
nge
,
“
L
oca
l
iz
a
ti
on
m
et
h
ods b
y
using ph
a
se
of a
rrival
,”
I
E
EE
journal
,
201
6
.
[6]
F
.
Seco
and
A
.
R.
Jim
´
ene
z,
“
RF
ID
-
base
d
ce
n
tra
lized
coopera
ti
ve
loca
lization
in
indoor
envi
r
onm
ent
s
,”
IE
EE
journal
,
2016
.
[7]
Y.
Zha
o
,
et al
.
,
“
VIRE:
Act
ive R
FID
-
base
d
localiza
t
ion
using v
irtual
r
efe
r
ence
e
li
m
ina
ti
on
,”
Proc. of
IC
PP
,
2007
.
[8]
X
.
Huang,
e
t
al
.
,
“
Outdoor
localiza
t
ion
using
acti
ve
RF
ID t
e
chnolog
y
,”
Proc. of B
ROADNETS
,
20
06.
[9]
S
.
Chum
kamon,
et
a
l.
,
Proc
ee
d
ings
of
ECTI
-
C
ON
,
2008
.
A
B
li
nd
Navig
at
ion
S
y
stem
Us
ing
RF
ID
for
Indoor
Envi
ronm
ent
s
.
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
S
N
:
2502
-
4752
Ind
on
esi
a
n
J
E
le
c Eng &
Co
m
p
Sci,
Vo
l.
13
, N
o.
3
,
Ma
rc
h 201
9
:
111
0
–
11
1
6
1116
[10]
Chou
L.
D.,
e
t
al.
,
“
Requi
re
m
ent
Anal
y
sis
andI
m
ple
m
ent
ati
on
of
Palm
-
bas
ed
Multi
m
edi
a
Mus
eum
Guide
S
y
stems
,”
I
EEE
18th
Int
ernati
on
al
,
2004
.
[11]
F.
B.
Za
hn
and
C.
E.
Noon,
“
Sh
orte
st
pat
h
al
gor
it
hm
s:
An
eva
lua
ti
on
using
rea
l
roa
d
net
works
,
”
Tr
anspo
rt.
Sci
.
,
vol.
32
,
pp
.
65
–
73
,
1998
.
[12]
G.
Desaul
nie
rs
and
F.
Soum
is,
“
An
eff
ic
ie
nt
algorithm
to
find
a
shortest
pat
h
f
or
a
ca
r
-
l
ike
rob
ot,
”
I
EE
E
Tr
ans.
Robot
Aut
omat
.
,
vol/
issue:
11(6)
,
pp.
819
–
828
,
19
95
.
[13]
J.
Mo
y
,
“
Open
S
horte
st Pa
th
Firs
t
Version
2
.
RF
Q 1583,
”
Inte
rne
t
Eng
ine
ering
Tas
k
Force
,
1994
.
[14]
S
.
Anus
ha
and
S
.
I
y
er,
“
A
Covera
ge
Planni
ng
Tool
for
RF
ID
Networks
W
it
h
Mobile
Re
ade
rs:
EUC
works
hops,
”
LNCS 3823
,
200
5
.
[15]
A
.
W.
Mohem
m
ed
,
et
al
.
,
“
Solving
shortest
pat
h
prob
le
m
u
s
ing
par
ticl
e
sw
arm
opti
m
iz
a
ti
o
n
,”
Appl
i
ed
Sof
t
Computing
,
vo
l.
8
,
pp
.
1643
-
165
3
,
2008
.
[16]
J.
B.
Orlin,
et
al.
,
“
A
faste
r
algorithm
for
the
single
source
shortest
pat
h
pro
ble
m
with
few
disti
nct
posit
ive
len
gths
,”
Journal
of
Discret
e
A
lg
orithms
,
vol/is
sue:
8(2)
,
pp
.
189
-
198
,
2010
.
[17]
J.
Kenne
d
y
and
R.
C.
Ebe
rh
art
,
“
Parti
cl
e
sw
arm
opti
m
iz
ation
,
”
P
roce
edi
ngs
of
th
e
IEEE
Int
ernational
Confe
ren
ce
on
Neural
N
et
w
orks
,
pp.
1942
–
1
948
,
1995
.
[18]
D.
E.
Goldb
erg
,
“
Gene
ti
c
Algor
ithm
s
in
Sear
ch,
Optimiza
ti
o
n
an
d
Mac
hine
Lear
ning
,”
Addison
-
W
esley
,
Re
adi
n
g,
MA
,
1989
.
Evaluation Warning : The document was created with Spire.PDF for Python.