TELKOM
NIKA
, Vol.12, No
.2, June 20
14
, pp. 411~4
1
8
ISSN: 1693-6
930,
accredited
A
by DIKTI, De
cree No: 58/DIK
T
I/Kep/2013
DOI
:
10.12928/TELKOMNIKA.v12i2.1949
411
Re
cei
v
ed Ma
rch 5, 2
014;
Re
vised Ap
ril
7, 2014; Accepted April 2
6
, 2014
A New Clustering Algorithm Using Links' Weight to
Decrease Consumed Energy in MANETs
Abba
s Afsha
r
farnia
*
, Ab
b
as Karimi
Dep
a
rtment of Comp
uter Engi
neer
ing,
F
a
cult
y of Eng
i
ne
eri
n
g, Arak Branch
Islamic Azad U
n
iversit
y
, Arak,
Iran
*Corres
p
o
ndi
n
g
author, e-ma
i
l
: Abbas_
a
fsha
r67@
ya
ho
o.co
m; akar
imi@i
a
u-arak.ac.ir
A
b
st
r
a
ct
One of the most i
m
p
o
rtan
t proble
m
s of
clus
terin
g
al
gorith
m
s i
n
mobil
e
ad-
hoc
netw
o
rks
(MANETs) is
the rel
a
tive
ly l
o
w
stability
in
gen
erate
d
cl
usters w
h
ich
are res
u
lte
d
b
y
rapi
d cl
uste
rs
destructio
n
an
d hig
h
en
ergy
consu
m
ption
in
perfor
m
in
g the
re-clusteri
ng p
r
ocesses. Man
y
algor
ith
m
s ha
v
e
bee
n prov
id
ed
to incre
a
se the
clusters stabi
li
ty of w
h
ich the most si
gnific
a
nt are w
e
ig
ht-b
ased
alg
o
rith
ms.
In w
e
ight-bas
e
d
alg
o
rith
ms, o
n
ly li
mite
d info
rmati
on
of eac
h nod
e is use
d
to det
er
min
e
its w
e
ight and
it
causes
that th
e
best
possi
bl
e
optio
n for c
l
ust
e
r-he
ad
is
n
o
t selecte
d
.
T
he purp
o
se of
this
pa
per is
pr
ovi
d
in
g
one w
e
ight-b
a
s
ed a
l
g
o
rith
m
i
n
w
h
ich
eac
h
nod
e'
s w
e
ight
deter
mi
natio
n i
s
perfor
m
e
d
n
o
t only
by
usi
n
g its
nod
e inf
o
rmati
on b
u
t als
o
its
nei
ghb
or
’
s
nod
es infor
m
at
ion
and th
is w
o
rk i
s
perfor
m
e
d
by
deter
mi
nin
g
th
e
links'
w
e
ig
ht b
e
tw
een n
odes
that provi
de c
o
nnecti
ons
betw
een no
des.
Vi
a
this met
hod, the
best possi
bl
e
optio
ns can b
e
selected as cl
uster-he
ad. In simulati
ons
a
n
d
perfor
m
e
d
ex
peri
m
e
n
ts, it is reveal
ed that the
gen
erate
d
clus
ters by our pro
pose
d
al
gorith
m
s hav
e very h
i
gh stab
ility.
Ke
y
w
ords
: Cl
usterin
g
, Mobil
e
Netw
orks, Nodes'
Stabi
lity, MANET
s
, Consumed En
ergy
1. Introduc
tion
Clu
s
terin
g
is defined in
MANET a
s
follows
: natu
r
al arran
geme
n
t of mobile
node
s i
n
several different grou
ps [1,
2]. Figure 1 s
hows a MANET after the clusteri
ng p
r
o
c
esse
s.
Figure 1. Clus
tering in MANET [1]
Clu
s
terin
g
i
s
appli
ed to
increa
se
pe
rforma
n
c
e, fa
cilitate routin
g, de
cre
a
se
ene
rgy
con
s
um
ption,
increa
se
st
ability and n
e
twork
sp
rea
d
ability [3, 4]. Ho
weve
r, it has so
me
disa
dvantag
e
s
incl
udin
g
high ene
rgy co
nsum
pti
on to perfo
rm the re-cl
u
ste
r
in
g p
r
ocesse
s.
Each
nod
e'
s
energy is limi
t
ed in MA
NE
Ts
and
if
this
energy is fini
she
d
, the
nod
e will
be
destroyed. Now, if this no
de is a
clu
s
te
r-h
ead, n
o
t only the node i
t
self will be
d
e
stroye
d but i
t
will
also
re
sult to its clu
s
ter d
e
s
tru
c
tion
whe
r
eby
after d
e
s
troying
ea
ch
cluste
r, one
node i
s
sel
e
cted
as the cl
uster-head among remain
i
ng nodes and a new cl
uster will
be built fo
r remaining nodes.
These proce
s
ses a
r
e
call
ed re
-cl
u
ste
r
i
ng. Re-clu
ste
r
ing is a
ste
p
whi
c
h cau
s
e
s
high e
n
e
r
gy
con
s
um
ption
and con
s
ide
r
ing the
re
stricted en
ergy
; it increa
se
s the numbe
r of re-clu
stering
results in
ra
pid n
e
two
r
k
destruct.
On
e pu
rpo
s
e
o
f
clu
s
teri
ng
algorith
m
s is de
crea
sing
the
numbe
rs of this step o
c
currences a
nd its so
lution in g
enerating mo
re stabl
e clu
s
ters.
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 16
93-6
930
TELKOM
NIKA
Vol. 12, No. 2, June 20
14: 411 – 41
8
412
Many algo
rith
ms rel
a
ting to clu
s
terin
g
i
n
MANETs
were
sugg
este
d of which th
e most
signifi
cant
are weight-ba
s
ed al
gorith
m
s. To
dete
r
mine the
no
des'
weight
in weight-ba
s
ed
algorith
m
s, o
n
ly limited informatio
n of e
a
ch n
ode i
s
u
s
ed fo
r dete
r
mining its o
w
n weig
ht and
the
best po
ssible
option will be
sele
cted to b
e
com
e
the cl
uster-h
ead.
In this
pap
er,
one
cl
uste
ri
ng al
gorithm
in MANET
s
i
s
sug
g
e
s
ted i
n
which ea
ch
nod
e's
weig
ht is
pe
rforme
d n
o
t only by u
s
in
g its o
w
n
n
ode info
rmati
on but
also
its nei
ghb
o
r
s'
informatio
n a
nd this work
is pe
rform
ed
by det
ermini
n
g
the links' weight betwee
n
node
s which
provide
co
nn
ection
s b
e
tween no
de
s. This m
e
thod
is the be
st
way to mea
s
ure the m
obi
lity
predi
ction of
node
s, ene
rg
y perce
ntage,
neighbo
r sh
are
an
d other factors. In this algo
rithm, the
clu
s
ter-he
ad
s whi
c
h h
a
ve
the be
st fact
ors amo
ng t
heir n
e
igh
b
o
r
node
s a
r
e
selecte
d
. In this
algorith
m
first
l
y, the links' weig
ht betwe
en both
nod
e
s
are d
e
termi
ned on the b
a
si
s of four basi
c
factors
and
then, the
final
wei
ght of
ea
ch
nod
e i
s
d
e
termin
ed
according
to th
e lin
ks'
obtai
ne
d
weig
ht. Finally, the manners of each nod
e are
defin
ed
according to their obtai
ned
weig
hts.
In the next
section, th
e p
aper revie
w
s pr
evio
us me
ntioned
wo
rks. The
n
, the
prop
osed
algorithm is
provided. In the next secti
on, the
sim
u
l
a
tion result
s
will be eval
uated. In the l
a
st
se
ction, co
ncl
u
sio
n
s of the
menti
one
d di
scussio
n
s
will
be pre
s
ente
d
.
2. Related Work
Many factors
are offered to
sele
ct the cl
us
ter-h
ead. Clusteri
ng alg
o
r
ithms a
r
e
cla
ssifie
d
according to
these facto
r
s. The first offer
ed facto
r
is the neig
hborhoo
d de
gree (neig
h
b
o
rs'
numbe
r).
Am
ong
algo
rith
ms
whi
c
h
are
built o
n
the
basi
s
of it we
ca
n m
ention
the
HD (hig
hest
degree
) by G
e
rla [1
-3] an
d
LLC
(lea
st cl
uster
ch
ang
e) by Chian
g
[1
-3] whi
c
h
attempted to sele
ct
the clu
s
ter-he
ad usi
ng the
neigh
borhoo
d
degre
e
value
s
of node
s.
Another offered fa
ctor i
s
th
e nod
e-m
obili
ty
and we
ca
n mentio
n
so
me alg
o
rithm
s
b
u
ilt on
the basi
s
of i
t
such as th
e
MOBIC (mo
b
ility-
based
metric fo
r clu
s
terin
g
) by K
han [1-3], DDVC
(dynami
c
Do
ppler velo
city cluste
ring
) a
nd DL
DC
(dynamic lin
k du
ration cl
uste
ri
ng) by Sakh
a
ee
et al. [5]. In this alg
o
rithm,
one nod
e is sele
cted a
s
the clu
s
ter-h
e
ad whi
c
h ha
s the least mo
bility
among its n
e
i
ghbo
rs o
r
ha
s similar mo
bili
ty to its neigh
bors.
Another fact
or i
s
e
n
e
r
gy
and
the
alg
o
rith
m
s
b
a
se
d on
which
are
built fro
m
ene
rgy
efficient cl
ust
e
ring
algo
rith
ms such a
s
the po
we
r-a
ware
con
n
e
c
te
d domin
ant set by Wu [1-3],
SCA (stable
clu
s
ter
algo
rithm) by Sh
eu
[1-3]
an
d DEECF (di
s
tri
b
uted en
ergy
efficient cl
ust
e
r
formation
)
by
Kim et al. [6]
.
In these
alg
o
rithm
s
, on
e
node i
s
sel
e
cted a
s
the
clu
s
ter-he
ad
whi
c
h
has the mo
st
energy amon
g their neig
h
b
o
rs.
Another fa
cto
r
is in fa
ct a compou
nd of
pr
eviou
s
fa
ctors: n
e
igh
borhood d
e
g
r
ee,
mobility,
energy and l
oad-bala
n
ci
n
g
calle
d the
weig
ht factor.
Th
is fa
ctor i
s
the be
st cri
t
erion in
sele
cting
the clu
s
ter-h
ead
s be
cau
s
e it covers a
ll the pr
eviou
s
facto
r
s. M
any
algorith
m
s are provided
according
to this fa
ctor
whi
c
h a
r
e i
n
cl
ud
ed in th
e com
b
ined
-metri
cs-ba
sed
cl
uste
ring
cla
s
s such
as
th
e WCA
(weighte
d
clusteri
ng alg
o
rithm)
by
Chatterjee
et
al. [7], DSCAM (Distri
but
ed
Scena
rio
-
ba
sed Clu
s
teri
n
g
Algorithm
for MANE
T
s
) by Anitha et al. [8], EWCA (enh
an
ced
weig
hted clu
s
terin
g
algo
rithm) by
Chan
g
Li et
al. [9]
,
KCMBC (k-hop
co
mpou
nd m
e
tric ba
se
d
clu
s
terin
g
) by
Leng et al. [10] and one
algorithm
p
r
opo
sed by M
u
thura
m
aling
a
m et al. [11].
Ho
wever, thi
s
gro
up of a
l
gorithm
s ha
s som
e
disa
dvantage
s i.e. the information whi
c
h
are
colle
cted to o
b
tain facto
r
value
s
for e
a
ch nod
e is the
same limite
d
informatio
n of the node it
sel
f
and it
ca
uses the
best
opti
on to
be
not
often sele
cte
d
to
be
come
the
clu
s
ter-he
ad. Thi
s
pro
b
l
e
m
has b
een rem
o
ved from the
propo
se
d alg
o
rithm in this
pape
r.
3. The Propo
sed Algori
t
h
m
In our propo
sed algo
rithm
firstly a weig
ht for each n
ode is d
e
termined a
c
cord
ing to its
links' wei
ght and finally, the clu
s
ter-he
ads a
r
e sel
e
cted a
c
cordi
ng to the obtained wei
ght
of
node
s a
nd m
e
mbe
r
s
are
a
d
mitted and
the man
n
e
r
s
of each no
de
are
determin
ed. By usin
g
the
links' weight,
we
can o
b
tai
n
the final we
ight
of each
node b
a
sed
on the obtai
n
ed informatio
n o
f
neigh
bor no
d
e
s
wh
ere
a
s i
n
p
r
eviou
s
m
e
thod
s, the fi
nal
weig
ht of
ea
ch
nod
e i
s
o
b
taine
d
b
y
using its own limited information only. We
can in
crease the
clust
e
r-heads
stability signifi
cantly
usin
g our p
r
o
posed metho
d
.
The ste
p
s of
building
clu
s
ters in the p
r
o
posed alg
o
rit
h
m are a
s
foll
ows:
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
1693-6
9
30
A New Clus
tering Algorithm Us
ing Links
'
Weight to Dec
r
ease ....
(Abbas
Afs
h
arfarnia)
413
1.
Determinin
g
the weight of
links whi
c
h are
between
two n
ode
s u
s
ing
the follo
wing f
a
cto
r
s:
links' n
e
igh
b
o
rho
od
sh
are
,
spe
ed
of th
e lin
ks,
co
nsumed
ene
rgy
between
two conn
ecte
d
node
s to the link an
d dista
n
ce b
e
twe
en
two co
nne
cte
d
node
s to the link.
2.
Determinin
g each nod
es'
weig
ht based
on its links' weight
3.
Selecting
clu
s
ter-he
ad
s
a
c
cordi
n
g
to
t
he
o
b
taine
d
weig
hts and
determi
ning
each clu
s
ter'
s
members.
3.1. Determining the links'
w
e
ight
The first fa
ctor is the
sha
r
e of links
nei
ghbo
rho
od. T
he more neig
hbors that on
e node
has, the
bett
e
r
can
d
id
wi
ll be to b
e
come the
clu
s
ter-he
ad. T
he ste
p
s of
determi
ning t
he
neigh
borhoo
d
share of link
are a
s
follows:
First, e
a
ch n
ode
s' n
e
igh
b
o
rho
od
deg
re
e is dete
r
min
e
d
whi
c
h
are
also
calle
d "
d
" an
d
sho
w
the nu
mber of no
de
s whi
c
h exi
s
t in the transaction rang
e [7]:
v
v
v
v
ra
ng
e
v
tx
v
v
dist
d
'
,
'
)
'
,
(
(1)
W
h
er
e
d
i
s
t
(
v
,v'
)
is
d
i
s
t
an
c
e
be
tw
e
en
two neighb
orho
od no
de
s. No
w, factor (1)
[percentag
e o
f
that links' ne
ighbo
rho
od share]
is
cal
c
ul
ated acco
rdin
g to equation
(2):
v
v
d
P
1
(2)
The se
co
nd factor is the
spe
ed of the links.
If c
l
uster-heads
move fas
t
er than their
membe
r
s, th
ey have to
outsid
e
form
their
me
mb
ers
rang
e co
nstantly
[5]. Therefore, m
o
re
attention sh
o
u
ld be paid t
o
sele
ct clu
s
t
e
r-hea
ds, be
cau
s
e the
cl
uster-h
ead
s
sho
u
ld have
th
e
least spee
d [12
]
,[13].
In this step, we obtain the speed
of the li
nks as e
quati
o
n (3
):
2
E
V
L
S
S
N
(3)
Whe
r
e V and
E are two con
necte
d nod
es to link. SV a
n
d SE are the
i
r spe
e
d.
In Figure 2, t
he nei
ghb
orh
ood
sha
r
e a
n
d
the spee
d
of the links
related to the
node
s V
and E are d
e
termin
ed.
Figure 2. Det
e
rmini
ng the
neigh
bor
hoo
d
share and th
e spe
ed of lin
ks
The third fact
or is t
he
con
s
umed e
n
e
r
gy
of the links.
Becau
s
e th
e
clu
s
ter-he
ad
has th
e
most re
sp
on
sibilitie
s amo
ng other
nod
es, they
con
s
ume m
o
re
energy. Hen
c
e, the remai
n
ing
energy sho
u
l
d
be con
s
id
ered in sele
ctin
g clu
s
ters
[14
-
16]. In this step, the energ
y
of each link is
estimated a
c
cording to its
node
s' en
erg
y
. Equat
ion (4) is u
s
ed to
determi
ne factor (EL):
2
E
v
L
E
E
E
(4)
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 16
93-6
9
30
TELKOM
NIKA
Vol. 12, No. 2, June 20
14: 411 – 41
8
414
In which EV and EE are the
con
s
ume
d
e
nergy of nod
e
s
V and E re
spectively.
The fou
r
th fa
ctor i
s
the
distan
ce b
e
twe
en two
co
nn
ected
nod
es
to links. The
clu
s
ter-
head
whi
c
h h
a
s
clo
s
er
nei
ghbo
rs
nee
d
s
le
ss e
n
e
r
gy
to interchan
ge inform
atio
n. This i
s
du
e to
its energy that
does
not finish very
soon
and the
cluster
stability’s increase. A
c
cording to thi
s
factor, parall
e
l
k-me
an
s cl
usteri
ng algo
rithm wa
s
su
gge
sted
by L
.
Thoma
s
a
n
d
B.Annapp
a [
17,
18] in which node
s with short dista
n
ce
are locate
d in the same cluster [19]. To determin
e
the
distan
ce fa
ctor, we pe
rform step
s as fo
llows:
A messa
ge i
s
se
nt from th
e source
(first)
nod
e to th
e
target
(seco
n
d)
me
ssage
and th
e
se
con
d
nod
e
returns
a re
spo
n
se to th
e first no
de
as
soo
n
as i
t
receive
s
th
e messa
ge.
The
sou
r
ce (first)
node
cal
c
ulat
es the return
-time and
co
n
s
ide
r
s it as th
e distan
ce fa
ctor (IL
)
.
After obtainin
g
four fa
ctors PV, NL, EL and IL
the fin
a
l link'
s
weig
ht (CL
)
i
s
e
s
timate
d
according to them. We u
s
e
a weight-equ
ation to estim
a
te it:
)
*
(
)
*
(
)
*
(
)
*
(
4
3
2
1
L
L
L
V
LV
I
W
E
W
N
W
P
W
C
(5)
In whi
c
h
W1,
W2,
W3 a
nd
W4 a
r
e
wei
g
ht factors th
at determi
ne th
e rel
a
ted fa
ctor effe
ct
amount. The
followin
g
crite
r
ia is con
s
ide
r
ed in dete
r
mi
ning them.
1
4
3
2
1
W
W
W
W
Figure 3. Det
e
rmini
ng final
weight of ea
ch no
de
3.2. Dete
rmining the Fin
a
l Weight o
f
Each Nod
e
In this ste
p
, the final
weig
ht of node
s a
r
e e
s
timated
con
s
id
erin
g the lin
k weigh
t
of ea
ch
node a
nd loa
d
-bal
an
cing.
The work is p
e
rform
ed a
s
follows: ea
ch
node a
d
d
s
up
its links' wei
ght
and divide
s in
to its neighb
o
r
hoo
d deg
re
e
(d) [20]:
V
d
i
LVi
V
d
C
C
V
1
1
(6)
No
w, the pri
m
ary wei
ght (CV1)
sho
u
ld
be reve
rsed:
In
th
is
s
t
ep
, th
e
lo
ad
b
a
l
an
c
i
ng
fa
c
t
or
s
h
ou
ld
be
in
vo
lve
d
su
ch
tha
t
e
a
ch
c
l
u
s
te
r
-
h
e
a
d
sup
port
s
ju
st
a limited
num
ber
of n
ode
s
in orde
r
for th
e loa
d
di
strib
u
tion to
be j
u
stified b
e
twe
e
n
clu
s
ter-he
ad
s [21]. First, a
crite
r
ion
calle
d
δ
i
s
dete
r
mi
ned. T
h
is criterion
indi
cate
s the
nu
mbe
r
of
optimum no
d
e
s which sh
o
u
ld be involve
d
in a clu
s
ter
to achieve lo
ad-b
a
lan
c
in
g. In the followi
ng,
the differen
c
e betwe
en n
ode nei
ghb
orhood d
egree
(d
V) an
d sui
t
able criteria
of neighb
o
rh
ood
degree (
δ
) wil
l
be cal
c
ulate
d
and later
kn
own a
s
∆
V:
V
V
V
d
(8)
At the en
d, th
e final
wei
ght
of ea
ch
nod
e
(CV)
is e
s
tim
a
ted by
exerti
ng the
loa
d
fa
ctor a
s
equatio
n (9
):
V
e
C
C
V
V
*
1
(9)
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
1693-6
930
A New Clus
tering Algorithm Us
ing Links
'
We
ight to Dec
r
ease ....
(Abbas
Afs
h
arfarnia)
415
3.3. Selectin
g Cluster-He
ads
After estimati
ng the final
weight of ea
ch
node,
the
no
des t
r
an
smit
their weight
s
to their
neigh
bors.
When tran
smi
s
sion
ha
s
fini
shed,
all nod
e
s
kno
w
th
eir weig
hts and
t
heir neig
hbo
rs'.
In co
ntinuatio
n, the n
ode
which
ha
s the
most
weig
ht
among
its
nei
ghbo
rs will
int
r
odu
ce
itself
as
the candi
date
of cl
uste
r-he
ad. C
andi
dat
es
will
be
col
l
ected
to g
e
n
e
rate
a d
o
ma
in set. Clu
s
te
r-
head
s will be
sele
cted a
nd
clu
s
ters will b
e
gene
rated.
4. Simulation
To stu
d
y the
efficien
cy of
pro
p
o
s
ed
al
gorithm,
som
e
differe
nt ex
perim
ents ha
ve bee
n
perfo
rmed
by the simul
a
to
r NS-2.34 [2
2]. In
these experim
ents,
one
mo
bile ad-h
o
c network
(MANET
) in the enviro
n
me
nt with dimen
s
ion
s
of 2000
*2000 m who
s
e data is
sh
own in Tabl
e 1
has b
een
sim
u
lated.
Table 1. Simulating Para
meters in Simulator NS
2
Value
Meaning
Parameter
100-500
Number of
Node
s
N
2000*2000 m
Size of netw
o
rk
X*
Y
0-5 m/s - Ra
ndo
mly
Speed Range
S_Range
100-200 m
Transmission Range
T_Range
600 s
Time of Simulation
Run Time
0.21, 0.37, 0.
21,
0.21
Weights
W
1
, W
2
, W
3
, W
4
In all experi
m
ents, the p
r
opo
se
d alg
o
r
ithm
was
co
mpared to t
w
o al
gorith
m
s: WCA
(Weighted
Cl
usteri
ng Al
go
rithm)
by Ch
atterjee
et al
. [7] and
LEACH
(lo
w
-en
e
rgy a
daptiv
e
clu
s
terin
g
hie
r
archy)
by Heinzelman
et al. [23].
The results in
dicated that ou
r algorithm i
s
the
most sta
b
le a
l
gorithm am
o
ng them.
4.1. Cluste
rs
' Life-Time (LT)
Whe
n
a
give
n nod
e i
s
se
lected
as the
clu
s
ter-he
ad
and
a
clust
e
r i
s
ge
ne
rat
ed, this
perio
d until the clu
s
ter i
s
destroyed is
calle
d the
clu
s
ter'
s life-tim
e
and it indicates the net
work
clu
s
ters' sta
b
i
lity and it is expresse
d in p
r
opo
rtion to seco
nd
s.
In this
expe
ri
ment, the
nu
mber of n
ode
wa
s
co
nsi
d
e
r
ed
a
s
1
00-5
00 a
nd
Figu
re 4
wa
s
obtaine
d wh
e
n
the experim
ent is finish
ed
.
Figure 4. Life-time parame
t
er with incre
a
sin
g
numb
e
r of node in ne
twork
It is shown in
Figure 4 that
the propo
se
d algor
ith
m
h
a
s lon
ger life
-
time than alg
o
rithm
s
WCA and LEACH and this life-time
will be incr
eased by the increasi
ng number of network
node
s. Its rea
s
on i
s
the p
r
o
per
sele
ction
of cluste
r-hea
ds
such that they
have the
highe
st sco
r
e
s
in their g
r
ou
p
.
Adding up
score
s
in
co
ntrast
with al
gorithm
WCA has
been
p
e
rform
ed by l
i
nk
informatio
n b
e
twee
n nod
e
s
and thi
s
in
clud
es oth
e
r node
s information. Mea
n
whil
e, only the
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 16
93-6
930
TELKOM
NIKA
Vol. 12, No. 2, June 20
14: 411 – 41
8
416
node'
s limited
informatio
n i
s
ap
plied to e
s
timate
its
score in
algo
rith
m WCA. The
load-bala
n
ci
n
g
factor which is also applie
d in the prop
ose
d
algo
rith
m enable
s
th
e algorithm t
o
transmit less
informatio
n for cross-net
wo
rk
comm
uni
cation and
co
n
t
ributes to m
a
intain stability
and de
crea
se
the netwo
rk li
fe-time.
In Table 2, a
comp
ari
s
on
has
been
do
ne in two
pa
rts of diag
ram
and the diff
eren
ce
s
betwe
en life-t
i
mes have b
e
en investig
ated.
Table 2. Co
m
parin
g Prop
o
s
ed Algo
rithm
for Paramete
r Life-Tim
e
Life-Time
w
i
th 40
0 nodes
Life-Time
w
i
th 20
0 nodes
Algorithm
3253 s
1632 s
Proposed Algorit
hm
2688 s
1416 s
WCA
21.01 %
15.25 %
Increment %
Con
s
id
erin
g
Table
2, it is kno
w
n tha
t
the pro
p
o
s
ed alg
o
rithm
life-time ha
s be
en
improve
d
in compa
r
ison to two other offered
al
gorith
m
s and this i
m
provem
ent is about 21%
in
some p
o
ints
and finally, it
results in hig
h
st
ability of the network in
the propo
se
d
algorithm.
4.2. Consum
ed Energ
y
Amount (E
N)
The con
s
um
ed ene
rgy a
m
ount in ne
twork nod
es in the given time to transmit
informatio
n i
s
call
ed the
u
p
load
or con
s
ume
d
e
n
e
r
g
y
amount
of
the net
work.
The le
sse
r
th
e
amount, the
more
stable t
he clusters are.
This amount is expressed in mill joul
e.
In this expe
ri
ment, the tra
n
smi
ssi
on
ra
nge
wa
s 10
0-200 m, a
nd t
hen
we p
e
rfo
r
med thi
s
experim
ent u
n
til Figure 5 i
s
obtain
ed.
Figure 5. Con
s
ume
d
ene
rg
y amount wi
th increa
sing t
r
an
smi
ssi
on range
In Figure 5, it was in
dicate
d that less e
nergy was
co
nsum
ed in th
e prop
osed a
l
gorithm
than al
gorith
m
s
WCA
an
d LEACH
an
d in thi
s
ca
se by in
crea
si
ng the
tran
smissi
on
ra
ng
e of
node
s ene
rgy
consumptio
n
decrea
s
e. At first, t
he algorithm WCA perfo
rms efficiently, however
whe
n
the
tra
n
smi
ssi
on
ra
nge
of net
wo
rk nod
es
in
creased, the
e
nergy
co
nsu
m
ption in
crea
sed
signifi
cantly in this alg
o
rit
h
m and
sho
w
ed lo
we
r
efficien
cy than
our al
gorith
m
. This shows that
the p
r
op
osed
algo
rithm
co
nsum
es very
little ene
rg
y b
e
ca
use of
the
prope
r
stru
ct
ure
of its
ste
p
s
and sel
e
ctin
g
the
b
e
st clu
s
ter-h
ead
s whi
c
h are
s
upe
ri
or th
an th
eir
neigh
bors'
inf
o
rmatio
n le
ad
s
to the incre
a
se of stability and life-time of
cluste
rs.
4.3. Number
of Clus
terin
g
and Re
-clu
stering O
p
er
ations
If a cluster i
s
destroyed (removed
)
, the re-c
lu
sterin
g
phase will b
e
initiated an
d one is
sele
cted
as t
he cl
uste
r-he
ad amo
ng av
ailable n
ode
s and a
ne
w
clu
s
ter
will b
e
gen
erate
d
. It
results i
n
con
s
umin
g hig
h
energy and
finally de
st
royi
ng the
whol
e
netwo
rk sta
b
i
lity; the less t
h
is
amount, the
more
stable t
he network is.
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
1693-6
930
A New Clus
tering Algorithm Us
ing Links
'
We
ight to Dec
r
ease ....
(Abbas
Afs
h
arfarnia)
417
In this exp
e
riment, the n
u
mbe
r
of
re-clu
sterin
g o
p
e
ration
s
of
each alg
o
rith
m wa
s
investigate
d
in different tra
n
smi
ssi
on ra
nge
s.
We re
g
u
lated the tra
n
smi
ssi
on ra
nge from 10
0
to
200 m until Fi
gure 6 i
s
dete
r
mine
d.
Figure 6. Nu
mber of cl
ust
e
ring a
nd re-clu
sterin
g op
eration
s
with i
n
crea
sing tra
n
smi
ssi
on ra
nge.
From Fi
gure
6 it reveal
s that le
ss
re-clu
steri
n
g
operation
s
are p
e
rfo
r
me
d in the
prop
osed al
g
o
rithm than
algorith
m
s
WCA and
LEACH
and by i
n
crea
sing th
e Tra
n
smi
s
si
on
rang
e, this n
u
mbe
r
will de
cre
a
se. Re
ga
rding to t
he f
a
ct that this
para
m
eter i
s
one of the m
o
st
signifi
cant pa
ramete
rs fo
r whi
c
h cl
uste
ri
ng algo
ri
thm
s
sea
r
ch, a
nd
then it can b
e
prop
osed that
the propo
se
d
algo
rithm
ca
n obtain
a
si
gnifica
nt
diag
ram. It wa
s i
m
prove
d
be
cause the
clu
s
ter-
head
s we
re determi
ned
p
r
ope
rly
a
nd
t
he clu
s
ters maintainin
g pha
se wa
s p
l
anne
d
effici
e
n
tly
su
ch that after the failure of
one clu
s
ter,
t
he next clu
s
ter-hea
ds a
r
e
sele
cted
correctly.
In Table
3 t
here
is a
co
mpari
s
o
n
in
two p
a
rts of
the di
agram
and
the
differen
c
e
s
betwe
en num
bers of gene
rated clu
s
ters
are stu
d
ied.
Table 3. Co
m
parin
g the Propo
sed Algo
ri
thm for Clu
s
tering a
nd Re-Clu
sterin
g Pa
ramete
r
Number of cluste
ring
w
i
th 175
Transmission range
Number of cluste
ring
w
i
th 140
Transmission range
Algorithm
19.1
26.5
Proposed Algorit
hm
21.9
27.2
LEACH
12.78 %
2.57 %
Decrement
%
Con
s
id
erin
g
Table
3 it i
s
cle
a
r th
at t
he n
u
mbe
r
of re
-cl
u
ste
r
i
ng o
peration
s
in
the
prop
osed
alg
o
rithm
ha
s b
een i
m
proved
in
com
pari
s
o
n
to t
w
o
othe
r me
ntione
d
algorith
m
s. E
v
en
at some
poin
t
s, we see 1
2
% decrem
e
nt which fi
nal
ly leads to hi
gh stability of
netwo
rk in t
he
prop
osed alg
o
rithm.
5. Conclusio
n
In this pap
er,
an algo
rithm
to cluste
r no
des
in m
obile
ad-h
o
c n
e
tworks (MENET
s) h
a
s
been
sugg
ested in
whi
c
h
so
me
node
s a
r
e
sele
cte
d
a
s
clu
s
ter-head
s wh
ose
links have
t
he
highe
st value
.
To dete
r
min
e
the lin
ks' va
lue bet
we
e
n
node
s, fou
r
fa
ctors a
r
e a
ppl
ied i.e.: ene
rg
y,
mobility, neig
hborhoo
d
sh
are
and
lin
k'
s le
ngth. To
estimate th
e
final value
of ea
ch n
ode,
the
obtaine
d links value and lo
ad-b
a
lan
c
in
g factor a
r
e u
s
e
d
.
Usi
ng the lin
ks' weight e
s
ti
mation, the final
wei
ght of
each nod
e ca
n be e
s
timate
d ba
sed
on obtain
ed i
n
formatio
n from its neig
h
b
o
r nod
es
wh
e
r
ea
s in previo
us meth
od
s, the final wei
g
h
t
of each n
ode
is estimate
d by using o
n
ly its own limite
d
informatio
n.
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 16
93-6
930
TELKOM
NIKA
Vol. 12, No. 2, June 20
14: 411 – 41
8
418
In pe
rform
e
d
sim
u
lation
s, it wa
s
reve
aled
th
at ou
r propo
sed
al
gorithm
sta
b
i
lity was
improve
d
sig
n
ificantly in compa
r
ison to
previo
u
s
alg
o
rithm
s
and i
t
resulted in
increa
sing th
e
clu
s
ters' life-ti
me and de
cre
a
sin
g
ene
rgy con
s
um
ption.
Referen
ces
[1]
JY Yu, PHJ
C
hon
g. A S
u
rve
y
of Cl
usteri
ng
Schem
es for
Mobil
e
A
d
H
o
c
Net
w
orks.
T
h
e El
ectroni
c
Maga
z
i
ne of Origin
al Pe
er-Re
v
iew
ed Survey
Articles.
200
5; 7(1): 32-4
8
.
[2]
IG Sha
y
eb, A
H
Huss
ei
n, AB
Naso
ura. A
S
u
rve
y
of Cl
usterin
g
Sch
e
mes
fo
r Mobile Ad-
H
oc Net
w
or
k
(MANET
)
.
Ame
r
ican Jo
urna
l o
f
Scientific Res
earch.
20
11
;
2
0
: 135-1
51.
[3]
S Chin
ara, SK
Rath. A Surve
y
on One-
Ho
p Cluster
ing A
l
gorit
hms in M
obil
e
Ad H
o
c Net
w
orks.
Journ
a
l Netw
Syst Manage.
2
009
;
17: 1
83–
2
07.
[4]
X Li
u, J Shi. Clusterin
g R
outin
g Algor
ith
m
s In W
i
reless Sensor Net
w
o
r
ks: An Overvie
w
.
KSII
T
r
ansactio
n
s o
n
Internet And
Information Sy
stems.
20
12
;
6
:
1735-1
7
5
5
.
[5]
E Sakh
ae
e, A
Jamali
po
ur. Stabl
e C
l
usteri
ng
an
d
Commu
ni
cations
in
Pse
udo
lin
ear
Hi
ghl
y M
o
b
ile
A
d
Hoc Net
w
orks.
IEEE Transactions on Vehic
u
lar Technology.
2008; 5
7
(6): 3
769-
377
7.
[6]
Y Kim, KY Jung, T
H
Kim, J Kim.
A distribute
d
en
erg
y
-effici
ent cluster
i
n
g
s
c
heme for
de
pl
o
y
in
g IDS i
n
MANET
s
.
Telecommunic
a
tion System
s.
201
1.
[7]
M Chatterjee,
SK Das,
D
T
u
rgut. WCA:
A Weig
hted
Clusteri
n
g
Alg
o
rithm for M
o
bile
Ad
Hoc
Net
w
orks.
Klu
w
er Acade
mic
Publ
ishers. Ma
nufacture
d in T
he Neth
erla
nds
.
2002; 5: 193-
204.
[8]
VS Anitha, M
P
Sebasti
an.
Scenar
io-b
ase
d
Di
am
eter-b
o
und
ed A
l
gor
ith
m
for Cluster
Creati
on a
n
d
Mana
geme
n
t in Mobil
e
Ad h
o
c Net
w
orks.
13th IEEE/ACM International Sym
pos
ium
on Distribute
d
Simulati
on a
n
d
Real T
i
me Ap
plicati
ons.
2
0
0
9
: 97-10
4.
[9]
C Li, Y W
ang
, F
Huang, D
Yang.
A Nov
e
l Enh
anc
ed
W
e
ighte
d
Clus
t
ering Al
gorit
h
m
for Mob
ile
Netw
orks
. IEEE conferenc
e 9
78-1-
424
4-3
6
9
3
-4/09. 20
09.
[10]
S Leng, L Z
h
a
ng, H F
u
, J Ya
ng. A Novel L
o
cati
o
n
-Servic
e
Protocol Bas
ed on k-Ho
p C
l
usteri
ng for
Mobile Ad Hoc
Net
w
orks.
IEEE Transactions
on Vehic
u
lar Technology.
200
7;
56(2): 81
0-8
17.
[11]
S Muthurama
l
i
ngam, R R
a
ja
Ram, K Petha
perum
al, VK Devi. A D
y
n
a
m
ic Clusteri
n
g
Algorithm fo
r
MANET
s
b
y
m
odif
y
i
ng W
e
i
g
h
t
ed Cl
usteri
ng
Algorit
hm
w
i
th
Mobil
i
t
y
Pred
ic
tion.
Intern
atio
nal J
our
nal
of Computer a
nd Electric
al E
ngi
neer
in
g.
20
10
;
2(4): 709-
714.
[12]
C Konstanto
p
oul
os, D Gavala
s, G Pantziou. Clust
erin
g in mobi
le
ad hoc n
e
t
w
o
r
ks through
nei
ghb
orh
ood stabilit
y-
base
d
mobil
i
t
y
pred
iction.
Co
mputer Netw
orks.
200
8
;
52: 179
7–1
8
24.
[13]
J Akbari T
o
rke
s
tani, MR Me
ybod
i. A Mobi
lit
y-Bas
ed
Cluste
r
F
o
rmation A
l
gorithm for W
i
r
e
less M
obi
le
Ad-hoc Net
w
orks.
Cluster Computi
ng-T
he
Journ
a
l of Net
w
orks Softw
a
re T
ools And
Appl
icatio
ns.
201
1; 14: 311
–
324.
[14]
SC W
ang,
HH
Pan, KQ Ya
n, YL
Lo. A
un
ifie
d frame
w
o
r
k fo
r cluster ma
na
ger e
l
ecti
on
an
d cluster
i
n
g
me
ch
an
i
s
m in
mo
b
i
le
ad
ho
c n
e
t
w
o
rks.
C
o
mp
uter Stan
da
rds & Interface
s
-Elsevier.
200
8
;
30: 32
9–
338.
[15]
H Ali, W
Sha
h
zad, F
A
Kha
n
. Energ
y
-
e
fficient
cl
usterin
g
in mob
ile
ad-
hoc n
e
t
w
orks
usin
g multi-
obj
ective p
a
rticle
s
w
arm opti
mization.
Appli
e
d
Soft Comp
utin
g.
2012
;
1
2
: 19
13–
19
28.
[16]
A Karimi, SM
Abed
ini, F
Z
a
r
a
fshan, SAR A
l
-Had
da
d. Clus
t
er Hea
d
Sel
e
c
t
ion Usi
ng F
u
z
z
y
L
o
g
i
c an
d
Cha
o
tic Base
d Genetic Al
g
o
rithm in W
i
r
e
less Se
nsor
Net
w
ork.
Jour
nal of Bas
i
c
and A
ppl
ie
d
Scientific R
e
se
arch.
201
3
;
3(4
)
: 694-70
3.
[17]
L T
homas, B Anna
pp
a.
Applic
ation of Para
ll
e
l
K-Means Cl
us
te
ring Al
gorith
m
for Predicti
o
n of Optima
l
Path in Se
lf Aw
are Mobil
e
Ad-Hoc Netw
orks w
i
th Link
Stability
. Adv
ances i
n
Com
putin
g an
d
Commun
i
cati
o
n
s (ACC 20
11)
, India. 201
1.
[18]
L T
homas, K M
anj
app
a, B A
n
nap
pa, GRM
R
edd
y. P
a
ral
l
el
iz
ed K-Me
ans
Cl
usterin
g
Al
gorit
hm for Se
lf
A
w
a
r
e Mo
bil
e
Ad-Hoc N
e
t
w
or
ks.
ACM 978-1
-
450
3-04
64-
1/11/02 - ICCCS
11.
201
1
:
152-
155.
[19]
S Chin
nasam
y, C Mu
thusamy, G Ramani.
An Anal
ys
is o
n
the Performance of F
u
zz
y C -Means
Clusteri
n
g
Al
g
o
rithm for
Car
d
iotoc
ogram
D
a
ta C
l
usteri
ng.
Cas
p
ia
n J
our
nal
of A
ppl
ie
d
Scie
nces
Research (CJA
SR).
2012
;
1(1
3
): 35-42.
[20]
A Karimi, A Af
sharfarnia, F Za
rafsha
n, SAR
Al-Ha
dda
d. A
Novel
Cl
usteri
ng Al
gorit
hm for Mob
ile
Ad-
Hoc Net
w
orks
Based o
n
Det
e
rminati
on of
Virtual
Li
nks'
W
e
ight to Increase N
e
t
w
ork Stabil
i
t
y
.
Th
e
Scientific W
o
rl
d Journ
a
l.
2
014
.
(In Press).
[21]
S Kang, S Le
e, S Ahn, S An. Energ
y
Effic
i
ent T
opolo
g
y
Contro
l bas
ed
on Soci
ol
ogic
a
l Cluster i
n
Wireless S
ens
or Net
w
orks.
KSII T
r
ansactions o
n
Inter
n
e
t
And Infor
m
ati
on Syste
m
s
. 2
012;
6: 34
1-
360.
[22]
T
Issariy
a
k
u
l,
E Hossain.
In
troductio
n
to N
e
tw
ork Simulat
o
r NS2
. USA:
Scienc
e Bus
i
ness Media
-
Sprin
ger. 20
09
.
[23]
W
B
Heinze
lma
n
, AP Chan
dra
k
asan, H Ba
lak
r
ishn
an.
An Ap
plicati
on-S
pecif
ic rotocol Arch
i
t
ecture for
Wireless Micr
osensor
Net
w
or
ks.
IEEE Transactions
on Wire
less Comm
unications.
20
02
;
1(4): 6
60-
670.
Evaluation Warning : The document was created with Spire.PDF for Python.