TELKOM
NIKA Indonesia
n
Journal of
Electrical En
gineering
Vol.12, No.6, Jun
e
201
4, pp. 4283 ~ 4
2
8
9
DOI: 10.115
9
1
/telkomni
ka.
v
12i6.553
3
4283
Re
cei
v
ed O
c
t
ober 3
0
, 201
3; Revi
se
d Decem
b
e
r
31, 2013; Accept
ed Ja
nua
ry 2
5
, 2014
A Novel Clustering Routing Protocol in Wireless Sensor
Network
Wu Rui, Xia Ke
w
e
n*, Bai
Jianchu
a
n, Zhang Zhi
w
e
i
Schoo
l of Information En
gi
ne
erin
g, Heib
ei U
n
iversit
y
of T
e
chno
log
y
Beich
en Distrct
,
T
i
anjin, Ch
ina
*Corres
p
o
ndi
n
g
author, e-ma
i
l
: k
w
xi
a@h
e
b
u
t.edu.cn
A
b
st
r
a
ct
T
he p
h
e
n
o
m
e
non
of
netw
o
rk reso
urce-co
n
strain
ed ofte
n
a
p
p
ears in w
i
reless
s
enso
r
netw
o
r
k
(W
SN) beca
u
s
e
of n
o
d
e
s w
i
th
en
ergy-l
imited
, so it b
e
co
mes
a res
earc
h
top
i
c to study
o
n
h
i
gh-
perfor
m
a
n
c
e
routin
g pr
otoc
ol
in w
i
re
less
sensor
netw
o
r
k
. In a
ccor
d
a
n
c
e w
i
th th
e ch
aracteristic
of
qua
ntu
m
parti
cle
sw
arm opti
m
i
z
ation (QPSO),
a nove
l
cluster
i
ng routi
ng pr
o
t
ocol w
i
th QPSO algorith
m
is prop
osed b
a
se
d
on the low
energy a
daptiv
e clusterin
g
h
i
erarchy (L
EA
CH), w
h
ich in
clud
es severa
l
steps, such as
deter
mi
nin
g
th
e nu
mb
er of cl
uster-he
ad, pre
l
i
m
in
ary
cluster
i
ng, opti
m
i
z
a
t
i
o
n on te
mp
orary
cluster by QPSO
alg
o
rith
m, clust
e
rin
g
a
fter o
p
ti
mi
z
a
t
i
o
n
, an
d
so on.
Co
mp
ar
ative a
n
a
l
ysis
on si
mulati
on
exper
iments s
how
that the pro
pos
ed rout
i
ng
prot
ocol is s
uper
ior
to that of LEA
CH, and s
a
ves
energy c
onsu
m
pti
on of n
e
tw
ork
excell
ently a
n
d
bala
n
ces e
ner
gy of nod
es so
as to prolo
ng t
he life of the n
e
tw
ork.
Ke
y
w
ords
: w
i
reless se
nsor n
e
tw
ork, clustering routi
ng
pr
otocol, qu
antu
m
particl
e sw
arm
opti
m
i
z
at
ion
Copy
right
©
2014 In
stitu
t
e o
f
Ad
van
ced
En
g
i
n
eerin
g and
Scien
ce. All
rig
h
t
s reser
ve
d
.
1. Introduc
tion
The advan
ce
s in wirel
e
ss communi
cat
i
on techn
o
lo
gy, electroni
c techn
o
logy
, senso
r
techn
o
logy a
nd micro
-
ele
c
trom
echani
cal syst
em
s (MEMS) ha
s prom
oted the
developme
n
t
of
wirel
e
ss
sen
s
or net
work(WSN) ap
plications a
nd re
se
arch. Wi
rele
ss sen
s
or n
e
t
w
ork i
s
a net
work
comp
osed of
nume
r
ou
s se
nso
r
no
de
s which h
a
ve the
powe
r
s of pe
rce
p
tion, com
m
unication a
nd
comp
uting, it is conn
ecte
d by wi
rele
ss mode,
a
nd
i
t
has
a wid
e
appli
c
ation prosp
e
ct
in ma
ny
fields
[1,
2]. As
the important part
of the
net
work laye
r, the
network
ro
uting d
e
liver the
informatio
n from the sou
r
ce node to net
work, and it
’s
the efficient communi
catio
n
of the network.
Due to th
e li
mitations
of the ha
rd
wa
re
whi
c
h h
a
s th
e wid
e
di
strib
u
tion of no
de
s, limited e
n
e
r
gy
[3], the small
memo
ry, the ba
d
comp
uting p
o
we
r and narro
w band
width
i
n
wirele
ss
se
nso
r
netwo
rks, th
e routin
g al
gorithm
s u
s
e
d
in many
tradition
al fixed-n
e
two
r
k a
nd mobil
e
self-
orga
nizi
ng n
e
t
work
can
not
be effectively
use
d
in
wi
re
le
ss
se
ns
or
ne
tw
o
r
k
[4
]. In
this
case, it ha
s
extremely im
portant
signif
i
can
c
e to th
e pro
p
o
s
ed routing proto
c
ol
of
low e
nergy
a
dapti
v
e
clu
s
terin
g
hi
e
r
archy (LEACH)
[5], whi
c
h define
s
the "
r
ound"
co
ncep
t and all
o
ws
different n
o
d
e
s
become the
cluste
r-hea
d with the
rando
m probability P in different time, the relay
comm
uni
cati
on busi
n
e
ss
is avera
ge share
d
in WS
N. Comp
are
d
to the same plane ro
uting
proto
c
ol
s, LEACH p
r
oto
c
o
l
improve
s
the over
all p
e
rform
a
n
c
e
of the netwo
rk an
d bala
n
c
e
s
energy con
s
umption
of n
ode
s. Based
on LEA
CH,
this p
ape
r p
r
opo
se
s th
e i
m
provem
ents to
solve ina
deq
uaci
e
s of LE
ACH.
Sun Jun
et al
[6]
p
r
op
oses a kind
of
qua
ntum pa
rticle
swa
r
m optimizatio
n (QPSO)
algorith
m
ba
sed on
qua
ntu
m
com
puting,
due to th
e
pa
rticle
ca
n ap
p
ear
any po
siti
on in the
entire
feasibl
e
sea
r
ch
sp
ace
with a
certain
p
r
oba
bility, an
d ha
s a
better fitne
s
s val
ue, so th
e gl
obal
sea
r
ch perfo
rmance for Q
PSO algorith
m
is sup
e
ri
o
r
to that of
PSO algorith
m
. The state of
particl
e is o
n
l
y describ
ed
with the po
si
tion vector,
and only o
n
e
paramete
r
need
s to
b
e
determi
ned in
QPSO algori
t
hm, so the Q
PSO algorith
m
has mo
re a
d
vantage
s:
(1) T
he few p
a
ram
e
ters, si
mple progra
mming an
d e
a
sy to imple
m
ent;
(2) T
he ra
pid
conve
r
ge
nce, and qui
ckly find t
he optima
l
solution in th
e global
scop
e.
Therefore,
a
c
cordi
ng to
t
he a
d
vantag
es
or
characteristics of t
he QPSO
al
gorithm,
literature
is replete with appli
c
ation
s
of
Q
PSO in
WSNs [7]. Low
ene
rgy
awa
r
e
clu
s
te
ring
hiera
r
chy (LE
A
CH) is a
si
mple an
d efficient al
g
o
rith
m. Cluste
ring
is an NP
-ha
r
d optimi
z
atio
n
probl
em [8],
whi
c
h
QPSO
ca
n h
andl
e
efficiently. Cl
usteri
ng
or cl
uster-h
ead
selectio
n i
s
n
o
t
a
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 23
02-4
046
TELKOM
NI
KA
Vol. 12, No. 6, June 20
14: 4283 – 4
289
4284
one-tim
e
a
c
tivity; therefore, the si
mpl
e
r the
optimi
z
ation
algo
rit
h
m, the b
e
tter the
net
wo
rk
efficien
cy is. This i
s
an re
aso
n
why QP
SO is a effici
ent choi
ce fo
r WSN
clu
s
te
ring. Thi
s
pa
per
prop
oses
a novel WSN clu
s
terin
g
ro
uting p
r
oto
c
o
l
com
b
inin
g
with LEA
CH,
whi
c
h i
s
used in
actual a
ppli
c
a
t
ion. The mai
n
work o
r
inn
o
vation is a
s
follows.
(1) Th
e n
u
m
ber of fixed
cluster-h
ead
is at
5%
of the
total nu
mbe
r
of
sen
s
o
r
n
o
des in
LEACH [6], which
is only e
x
perime
n
tal d
a
ta with
out a
n
y ba
sis. T
h
i
s
p
ape
r o
p
tim
i
ze
s the
num
ber
of clu
s
ter-h
e
ad b
a
sed o
n
the math
em
atical m
odel
of the p
h
ysi
c
al imag
e
sen
s
or net
wo
rk,
and
dynamically adjust
s
the pro
portion
so a
s
to
decrea
s
e the network e
nergy con
s
u
m
ption.
(2) T
he e
n
e
r
gy con
s
um
ption bet
ween
the node
s i
s
only bal
an
ced
by clu
s
t
e
ring i
n
LEACH. O
n
the ba
si
s of L
EACH,
the p
r
opo
sed
routin
g proto
c
ol i
n
this pa
per
ca
n reb
a
lan
c
e t
h
e
energy of each nod
e in cl
uster, an
d effectively av
oid the emerge
n
c
e of a lot of
blind nod
es a
nd
further p
r
olo
n
g
the netwo
rk lifetime.
2. Principle
of QPSO Alg
o
rithm
2.1. Algorith
m
O
v
er
v
i
e
w
Quantum
pa
rticle swa
r
m
optimizatio
n
(QPSO)
algo
rithm is th
e improvem
ent
of the
cla
ssi
cal PS
O algorith
m
[5], the velo
city and pos
i
t
ion of particle are cl
osel
y related wit
h
a
para
m
eter
in the QPSO algorith
m
. To ensu
r
e the
conve
r
ge
nce of the algorit
hm, a media
n
optimal po
sition is ad
opted
to calcul
ate the next it
erative variable fo
r parti
cle, whi
c
h is d
e
fined
as
the average v
a
lue of
perso
nal be
st
p
o
sit
i
ons fo
r all
pa
rticle
s (
mbes
t
). Th
e
formula
for
mbes
t
is
as
follows
:
1
11
1
11
1
,,
MM
M
ii
i
d
ii
i
mb
es
t
P
P
P
MM
M
（）
(1)
Her
e
,
M
is the numbe
r of pa
rticle
s,
i
P
is the
personal b
e
st position of the i-th pa
rticl
e
in
d
-dime
n
si
on
spa
c
e.
And the parti
cle evolutio
n equatio
n ca
n be obtain
ed a
s
follows:
12
1
2
()
/
(
)
id
i
d
g
d
PP
P
(2)
(1
(
)
l
n
(
1
/
)
id
id
X
tP
m
b
e
s
t
X
t
u
）
(3)
Her
e
,
id
P
is d
-
th
dimen
s
ion
coordi
nate of i
-
th pa
rticle,
g
d
P
is the gl
obal
b
e
st po
sition i
n
all particles,
id
X
is d
-
th di
men
s
ion
coordina
te of i-th
pa
rticle i
n
pa
rticl
e
qua
ntum p
a
rticle
swarm
spa
c
e;
12
,,
u
are
the rand
om
numbe
r bet
wee
n
0 and
1;
is cont
ractio
n-expan
sion
coeffici
ent, whose rol
e
is t
o
co
ntrol the
spe
ed of
con
v
ergen
ce,
wh
en
0.5
u
, the formula (3
)
t
a
ke
s “
”, otherwi
se takes
“
”.
The algo
rithm
steps fo
r QPSO are a
s
follows:
Step 1: Initialize each qu
antum particle position
ii
1
i
2
i
d
,,
,
X
xx
x
in
d-dim
e
n
s
ion
spa
c
e.
Step 2: Evaluate the fitness val
ue of pa
rticle. If the fitness
i
f
x
in current positio
n is
bigge
r th
an th
e fitness
i
f
pbes
t
of the pa
st
be
st p
o
sition, th
en
ii
p
be
st
x
; then
compa
r
e th
e
fitness valu
es of all current
particle
s
, if
i
f
x
is bigg
er than
the fitness of
global be
st p
o
sition
(
f
gbe
s
t
), the global o
p
timal solutio
n
is
i
gbe
s
t
x
.
Step 3: Updat
e the quantu
m
particl
e swarm in
a
c
cord
ance with formula (2
) and
(3).
Step 4: Che
ck
wheth
e
r t
he end
co
nd
ition is
satisf
ied or n
o
t. If it is met, Stop the
iteration; othe
rwi
s
e, retu
rn t
o
Step 2 to continue the it
eration.
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
2302-4
046
A Novel
Clu
s
tering
Routin
g
Protocol in
Wirel
e
ss Sen
s
or
Network (Wu Rui)
4285
2.2. Application and Anal
y
s
is on QPSO Algorithm
In order to verify the validity of QPSO
algo
rithm, we compa
r
e th
e QPSO, Ba
sic PSO
(BPSO) and Genetic Algorithm (GA
)
by comput
er simulation. The followi
ng common test
function i
s
sel
e
cted fo
r com
pari
s
on.
Sphere functi
on:
2
1
()
n
i
i
f
xx
, find a minimum value
in the rang
e [-5,5].
The maximu
m numb
e
r of
iteration
ma
x
ite
r
is
set to 100 time
s for th
ree
al
gorithm
s, the
siz
e
M
of the particle
swa
r
m
is 10, the
nu
mber
of pa
rticle dim
e
n
s
io
n (
d
) is 2,
is
0.7
5
.
T
h
e
cro
s
sove
r pro
bability is 0.9
5
and the mut
a
tion pro
babil
i
ty is 0.05 in GA.
For th
e
sake
of co
mpa
r
ison, the p
e
rfo
r
man
c
e
indi
cator fo
r evalu
a
tion i
s
set a
s
me
an
s
q
uare error,
whic
h is
as
follows
.
2
1
1
ˆ
()
n
ii
i
ex
x
n
(4)
Her
e
,
i
x
is the actual o
u
tput value,
ˆ
i
x
is the desi
r
ed o
u
tpu
t
value.
The e
r
ror e
repre
s
e
n
ts th
e mea
n
squa
re e
r
ror b
e
tween the
iterat
ion result in functio
n
optimizatio
n and the true
value, which display
s
the optimizatio
n pre
c
isi
on of the algorithm.
Figure 1 sh
o
w
s the ite
r
ation er
ror
cu
rve of Sphere f
unctio
n
.
I
t
er
at
i
on er
r
o
r
I
t
e
r
a
tion time
s
Figure 1. The
Iteration Erro
r
Cu
rve for Sphere Fun
c
tio
n
From
Figure
1, the
GA and BPSO al
gorith
m
still maintain a big
error value after 70
iteration
time
s,
but
the error cu
rve of QPSO
al
gorit
hm decrea
s
e
s
rapi
dly with the number
of
iteration time
s increa
sin
g
. After 42th iteration
times,
the erro
r valu
e beco
m
e
s
small and stab
le
.
Among the
m
, error
e=2.73
85×10
-2
when
iteration tim
e
s at 7
3
for
G
A
, erro
r e
=
1.
6582
×1
0
-3
wh
en
iteration times
at
53 for BPSO al
gorithm,
and the error e=1.8578×10
-7
wh
en ite
r
a
t
ion time
s at
45
for QPSO algorithm. Obvious
ly, QPSO algorithm is
far superior to GA
and BPSO algorithm in
the conve
r
ge
nce
spe
ed an
d optimizatio
n pre
c
isi
on.
3. Cluste
ring
Routing Pro
t
ocol Based
on QPSO
3.1. Net
w
o
r
k
Modeling
We ado
pt the wirel
e
ss communi
catio
n
system
mo
del in refere
nce [10] to calcul
ate,
whi
c
h i
s
con
s
tituted by the
transmissio
n
circuit,
po
we
r amplifier,
an
d the
receivin
g ci
rcuit. Wh
en
the b bit data
in tran
smissi
on po
rt is tra
n
smitted
to th
e re
ceiving p
o
rt thro
ugh th
e distan
ce
d, the
energy con
s
u
m
ption by tra
n
smitting is a
s
follows:
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 23
02-4
046
TELKOM
NI
KA
Vol. 12, No. 6, June 20
14: 4283 – 4
289
4286
2
10
4
20
Tx
e
l
e
c
Tx
e
l
e
c
E
bE
bu
d
d
d
E
bE
bu
d
d
d
(5)
The ene
rgy consumption b
y
receiving i
s
as follo
ws:
Rx
elec
Eb
E
(6)
Her
e
,
2
01
2
/
du
u
is the critical
value
of distan
ce. T
he ene
rgy co
nsum
ption fro
m
the
transmitting circuit to re
ceiving ci
rcuit is
5
0
n
J
/b
it
ele
c
E
, the en
ergy con
s
um
ption of the
amplifier ci
rcuit is
2
1
10pJ/b/m
u
a
nd
4
2
0.0013pJ/b/m
u
, the e
nergy
co
nsum
ption
of data
fusion i
s
5nJ/
b
i
t
F
E
.
Assu
me that
there i
s
N n
o
des
whi
c
h a
r
e di
stri
buted i
n
the sq
uare
area
of side l
ength A
with Poi
s
sion
dist
ributio
n o
f
distri
bution
den
sity
2
/
NA
, agg
regation
no
de
(Sin
k) is lo
cated
outsid
e
the di
stributio
n are
a
, radiu
s
of p
r
opa
gation
ra
nge of nod
es is
r
, clustering probability is
P
, final cl
uste
ri
ng n
u
mbe
r
i
s
KN
P
. Di
stributio
n
den
sity of
cl
uster-h
ead
is
1
P
, and
distrib
u
tion
d
ensity of
ordi
nary n
ode
is
0
1-
P
（）
. If
0
and
1
are
each ind
epe
n
dent, we
can g
e
t the numbe
r
[]
1
/
c
E
nP
P
of
membe
r
nod
es in ea
ch
cl
uster, a
nd th
e averag
e
distan
ce fro
m
cluste
r mem
bers to clu
s
te
r-h
ead i
s
1
1/
to
C
H
d
.
Above mathe
m
atical m
ode
l is e
s
tabli
s
h
ed, we
can g
e
t the total e
nergy
co
nsu
m
ption of
WSN [9]:
2
4
1
2s
i
n
2(
1
)
(
)
total
e
le
c
F
to
k
e
l
e
c
ub
A
N
Eb
N
E
b
E
N
K
b
u
d
E
K
(7)
Whe
r
e
sin
to
k
d
is the distan
ce bet
wee
n
clu
s
ter-head a
nd ag
g
r
egatio
n nod
e
(
Sink).
Formul
a (7
) i
s
the fun
c
tion
with an inde
pend
ent vari
able
K
, then using the m
e
thod of
Lapla
c
e extre
m
um, let
0
total
E
K
, and we can get
as follo
ws:
1
4
2s
i
n
()
to
k
e
le
c
u
KA
N
ud
E
(8)
From
form
ula
(8
), it i
s
o
b
vious that th
e i
deal
clu
s
te
rin
g
nu
mbe
r
K
is
related to
si
ze of
a
given are
a
an
d the numbe
r
of the sensor node
s, not
N
*5% in LEACH protocol.
3.2. The Specific Steps o
f
Clus
tering
Rou
t
ing Protocol
(1)
Determina
t
ion of the number of cl
ust
e
r-hea
d
To dete
r
min
e
the cl
uste
ri
ng nu
mbe
r
K
based o
n
th
e mathe
m
ati
c
al m
odel
a
nd the
formula (8), then get cl
ust
e
ring p
r
o
babil
i
ty
/
PK
N
.
(2) Pr
elimina
r
y
cluste
ring
Firstly, the netwo
rk
sh
o
u
ld be initial
i
zed a
nd p
r
elimina
r
y clu
s
tere
d with
LEACH,
determi
ning t
he tempo
r
ary
cluste
rs a
nd
clu
s
ter- hea
d. At the same time, the tempora
r
y clu
s
ter-
head
collect
s the
state inf
o
rmatio
n of
each me
mbe
r
in
clu
s
te
r, i
n
clu
d
ing
loca
tion informati
o
n
(sp
a
tial co
ord
i
nates) and e
nergy info
rma
t
ion.
(3) Optimiz
e
t
he temporary
c
l
us
ter with t
he QPSO algorithm
The p
opul
atio
n si
ze
of qu
a
n
tum pa
rticle
s a
r
e i
n
itialize
d
with
clu
s
te
r numb
e
r
K
, an
d the
spatial
co
ord
i
nates
of the
sen
s
o
r
n
o
d
e
s
co
rre
sp
on
d to the
po
sition of
qua
ntum pa
rticle
s.
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
2302-4
046
A Novel
Clu
s
tering
Routin
g
Protocol in
Wirel
e
ss Sen
s
or
Network (Wu Rui)
4287
Becau
s
e th
e
spatial
co
ordinate
s
of th
e se
ns
or n
o
des
are
ra
n
dom an
d di
screte, q
uant
u
m
particl
e'
s po
si
tion sho
u
ld b
e
adju
s
ted af
ter parti
cle u
pdating
with Step 3, which is adju
s
ted
by
the minim
u
m
Euclid
ean
distan
ce
met
hod i
n
thi
s
pape
r,
that mean solvin
g
the
Eu
clid
ean
distan
ce
s bet
wee
n
updat
e
d
quantum p
a
rticle
s an
d all sen
s
o
r
no
des, an
d po
sition of quant
um
particl
e is ad
justed to sp
atial coo
r
din
a
te values
of the neare
s
t sen
s
or n
o
d
e
.Then the total
energy con
s
u
m
ption of WS
N ca
n
be calculated on Eq
uation (7
).
To dete
r
mine
the fitness fu
nction i
s
the
most
imp
o
rta
n
t thing in the
stage. Ba
se
d on the
mathemati
c
al
model a
nd t
he eq
uali
z
ati
on re
qui
re
me
nts, the fitne
ss val
ue of
cluster-h
ead
n
o
t
only rel
a
tes
to own
state
inform
ation,
but al
so
sh
ould
be
affected
by stat
e information
of
membe
r
n
o
d
e
s in
cl
uste
r. Beca
use th
e ro
uting p
r
o
t
ocol
with a
singl
e-h
op transmi
ssion
mode
(i.e., transmit
information
dire
ctly between mem
b
e
r
node in
clu
s
t
e
r an
d clu
s
te
r-h
ead n
ode
),
the
farther di
stan
ce th
e
clu
s
ter-hea
d n
ode, t
he mo
re
rel
a
tive ene
rgy th
e mem
ber no
de in
cl
uste
r,
on
the contrary,
the le
ss e
nergy t
he
nea
r
membe
r
nod
e in
clu
s
te
r.
So the fitne
s
s fun
c
tion
ca
n be
establi
s
h
ed a
s
follows:
1
1
()
l
n
1
n
ji
i
i
ij
fj
e
e
nr
(9)
Her
e
,
,
0,
1
and
1
, rep
r
e
s
ent th
e ene
rgy imp
a
ct facto
r
of t
he no
de
j
and the rem
a
ining no
de i
n
clu
s
ter re
spectively,
j
e
rep
r
esents the
energy of the node
j
,
i
r
rep
r
e
s
ent
s th
e di
stan
ce b
e
twee
n no
de
i
and
no
de
j
,
n
re
pre
s
e
n
ts
the num
ber o
f
node
s in
clu
s
ter. The
more
suitabl
e
the cluste
r-h
ead no
de
s, the bigge
r the fitness functio
n
value.
The po
sition
s of quantum
particl
es a
r
e
update
d
with
the steps of
QPSO. The pbesti
and gb
esti of
particl
e is co
nfirmed a
c
co
rding to the va
lue of fitness. The tempo
r
a
r
y clu
s
ter-he
ad
can
be optimi
z
ed i
n
acco
rd
ance, and th
e formal
cl
u
s
ter-hea
d in thi
s
ro
und i
s
d
e
t
ermine
d so as
to achieve to
rebal
an
ce the
energy of no
des in
clu
s
ter.
(4)
Clu
s
terin
g
after optimiz
ation
The tempo
r
a
r
y clu
s
ter-he
ad broad
ca
st
s the inform
ation of form
al clu
s
ter-he
ad to the
node
s in
clu
s
ter, then th
e clu
s
teri
ng i
n
this
roun
d and the d
e
te
rminatio
n of clu
s
ter-he
ad
are
finis
h
ed.
4. Simulation and Analy
s
is
4.1. Simulation En
v
i
ronment
In the simulat
i
on experim
e
n
t, the wirele
ss
sen
s
o
r
net
work con
s
ist
s
of 100 sen
s
o
r
node
s
with GPS
po
sitioning fu
ncti
on in
the
are
a
of 1
50m
×1
50m. Th
e di
st
ance b
e
twe
e
n the Si
nk (b
ase
station)
nod
e
and the
cen
t
er of the area is 1
50m.
, the contraction-expa
nsi
o
n co
efficient,
values 0.75.
and
in the fitn
ess fun
c
tion
value 0.6
an
d 0.4,
re
spe
c
tively .The
numbe
r
o
f
iteration
s
is 1
00. The sim
u
l
a
tion paramet
ers fo
r
ra
dio model an
d ne
twork are given in Table 1.
Table 1. The
Simulation Param
e
ters
Parameter Value
Netw
or
k size
150m×150m
Base station loca
tion
X=75m,
y
=225m
Simulation round
450
Number of
node,
n
100
Cluster head p
r
o
babilit
y
of LEA
C
H,
P
0.05
Initial energ
y
,
E
o
0.1J
Packet size
5000bit
Transceiver ener
g
y
,
E
el
e
d
50nJ/bit
A
gg
re
g
ation en
er
gy
pe
r bit,
E
F
5nJ/bit
Multipath amplifier ener
g
y
,
u
2
0.0013pJ/bit/m
4
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 23
02-4
046
TELKOM
NI
KA
Vol. 12, No. 6, June 20
14: 4283 – 4
289
4288
4.2. Simulation Res
u
lts a
nd Analy
s
is
First of
all, the sim
u
la
tion expe
rim
ents a
nd
compa
r
ative
analysi
s
o
n
ene
rgy
con
s
um
ption
of netwo
rk a
r
e ca
rrie
d
out
in acco
r
dan
ce with the formula (7
), Fig
u
re 2
sho
w
s the
relation
shi
p
b
e
twee
n network e
n
e
r
gy co
nsum
ption an
d the numbe
r of cluste
r-h
e
ad.
En
e
r
g
y
c
o
nsum
ption
(J)
Numbe
r
of
cluste
r
-
head
E
n
er
gy
c
ons
umpt
i
o
n
(J
)
S
i
de
l
e
ng
th of
a
r
ea
(m)
Figure 2. The
Relation
ship
betwe
en the
Energy Con
s
umption an
d the Num
b
e
r
of
Nod
e
s
Figure 3. The
Relation
ship
betwe
en Network
Energy Con
s
umption an
d Side Length
of Area
From Fig
u
re
2, there is a minimum i
n
the netwo
rk ene
rgy co
nsum
ption when the
numbe
r of
clu
s
ter-he
ad
K
=3, 4, and the t
heoretical
re
sult
K
=3.431
8
cal
c
ulate
d
by
the formul
a (8)
is ba
sically consi
s
tent wit
h
the actual
value in
Figu
re 2
.
It is clear that the minimization on
netwo
rk e
n
e
r
gy con
s
umpti
on ca
n be a
c
hieved by
opt
imization of the numb
e
r of
cluste
r-hea
d.
Figure 3
sho
w
s th
e relatio
n
shi
p
bet
wee
n
network en
ergy con
s
um
ption an
d si
d
e
length
of are
a
.Fro
m
Figure 3,
with
the in
crea
sin
g
of
the
si
de
length A
of th
e net
work
area, the
ene
rg
y
con
s
um
ption
of two prot
ocol in
crea
ses. The
r
e i
s
a little diffe
ren
c
e in e
n
e
r
gy co
nsump
t
ion
betwe
en the
two protocols when th
e si
de length
of
area i
s
small
,
but when th
e A>30
0m, the
energy con
s
u
m
ption value
of LEACH is
highe
r than t
hat of prop
osed prot
ocol i
n
this pape
r, with
faster i
n
crea
sing rate
almo
st in expo
nen
tial form. Thi
s
is
be
cau
s
e
the clu
s
te
ring
numb
e
r of th
e
prop
osed p
r
o
t
ocol is a dy
namic va
riati
on, whi
c
h
is i
n
fluen
ced by
the factors such a
s
the si
de
length A and so on, but the cluste
ring p
r
oportio
n
of
LEACH is fixed at 5%. Clearly, the propo
sed
proto
c
ol is
su
perio
r to LEACH p
r
oto
c
ol.
In orde
r to a
nalyze th
e p
e
rform
a
n
c
e o
f
t
he entire
WSN, we ad
opt the net
work lifetim
e
para
m
eter to
evaluate
performan
ce
of
WSN. Fi
gure
4 sho
w
s the
simul
a
tion
curves of n
e
twork
lifetime from
two ro
uting p
r
otocols.
He
re, the netwo
rk lifetime is the time
of all
su
rviving no
des
from the begi
nning of the simulation to the death.
N
u
m
b
e
r
o
f
s
u
r
v
iv
in
g
n
o
d
e
s
Ro
und
Figure 4. Co
mpari
s
o
n
on
Networ
k Lifetime of Two Protocol
s
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
2302-4
046
A Novel
Clu
s
tering
Routin
g
Protocol in
Wirel
e
ss Sen
s
or
Network (Wu Rui)
4289
5. Conclusio
n
In this p
ape
r,
the cl
uste
rin
g
ro
uting
prot
oc
ol
of wi
rele
ss sen
s
or net
works b
a
sed
on the
QPSO can
effectively overcome
the
sh
ortco
m
ing
s
,
such
a
s
the
lo
w e
nergy of
wirel
e
ss
se
nsor,
the poo
r utili
zation, a
nd
so on. By u
s
e
of clu
s
te
ring
optimization
and e
n
e
r
gy e
quali
z
ation, t
he
netwo
rk lifeti
m
e is prolon
g
ed and n
e
two
r
k resource i
s
better utilize
d
.
In addition, f
o
r the
sin
g
le
-hop
commu
nicatio
n
may
re
sult in
en
ergy lo
ss in
the long
-
distan
ce
tran
smissio
n
[11],
we
can
ado
p
t
the multi-ho
p commu
nica
tions or
multi-layer clu
s
teri
ng
s
t
rategy to res
o
lve.
Ackn
o
w
l
e
dg
ements
This work was supp
orte
d by the National Natu
ral Scien
c
e
Found
ation
of China
(No.5
120
816
8), Tianji
n
Natu
ral
Scien
c
e
Foun
dati
on (No.1
1
JCYBJC00
9
00,
No.13
J
CYBJC37
700
), Hebei Provin
ce Na
tural
Science Found
ation (No.F
201
320
2254,
No.F20
132
02
102) a
nd Heb
e
i Province Found
ation for
Returned S
c
h
o
lars (No.C2
0120
0303
8).
Referen
ces
[1]
L H
u
a
w
e
i
. AC
O-Based
Routi
ng A
l
gor
ithm f
o
r W
i
rel
e
ss S
e
nsor N
e
t
w
ork.
Chin
ese
Jo
urn
a
l
of Sens
or
s
and Actuat
ors
. 200
7; 20(1
1
): 2450-
245
1.
[2]
Z
e
y
u
S
un, Z
h
enp
ing
LI.
W
i
reless Se
nsor
Net
w
ork Path
Optimizatio
n
Based o
n
H
y
br
i
d
Algor
ithm.
T
E
LKOMNIKA Indon
esi
a
Jour
nal of
Electric
al
Engin
eeri
ng.
2
013; 11(
9): 535
2-53
58.
[3]
L Guoh
ua, T
Hui. A Nove
l
Routi
ng Al
gorit
hm Ba
sed
on
Energ
y
Po
lic
y
for W
i
reless S
ensor N
e
t
w
ork
.
Journ
a
l of Elec
tronics & Informati
on T
e
ch
no
logy
. 20
06; 28(
1): 167-1
71.
[4]
T
Yong. T
he Researc
h
o
n
R
o
uting
an
d Br
oa
dcas
tin
g
Al
gori
t
hms for W
i
rel
e
ss Se
nsor
Ne
r
w
o
r
ks. P
h
D
T
hesis. Che
n
g
D
u: D
o
ctoral
D
i
ssertatio
n, Un
i
v
ersit
y
of E
l
ect
r
onic
Scie
nce
and
T
e
chnol
og
y
of C
h
i
na.
200
7.
[5]
Heinz
e
lm
an
W
,
Cha
ndr
akasa
n
A. E
nerg
y
-
efficient
Com
m
unic
a
tion
Pr
otocol
for W
i
r
e
less
Se
nso
r
Net
w
orks
. IEEE Computer S
o
ciety
. 200
0; 1
75-1
87.
[6]
J Sun, B
F
e
n
g
. Particl
e
s
w
arm optim
izati
on
w
i
t
h
particl
es h
a
vin
g
q
u
a
n
tum b
ehav
ior.
Con
g
ress
o
n
Evoluti
onar
y C
o
mputati
on.
IEEE
. 2004: 32
5-
331.
[7]
H Your
ui, T
Yiming
. Si
mulati
on R
e
searc
h
of the QoS M
u
lt
icast R
outin
g in W
S
N B
a
s
ed o
n
QPSO
Algorit
h
m
. T
h
ird Internati
o
n
a
l
S
y
mp
osi
u
m on
Intelli
gent Info
rmation T
e
chn
o
lo
g
y
Ap
plic
ati
on. 20
09; (3)
:
39-4
3
.
[8]
H Dong
xue, Z Ruihu
a
. PSO-base
d
Dou
b
l
e
Cluste
r-h
ea
ds Clusteri
ng Al
g
o
rithm for W
i
reless Se
nsor
Net
w
ork.
Co
mputer Eng
i
n
eeri
ng.
201
0; 36(1
0
): 100-1
03.
[9]
Handy
M, Haase M.
Low
Energy
Ad
apti
v
e
Clusteri
n
g
Hierarc
hy w
i
th
Deter
m
inistic
Cl
uster-He
a
d
Selecti
on.
Fourth IEEE Conf
erenc
e on Mobile and Wi
reless Comm
unic
a
tions
Net
w
ork
s
. 2002; 368-
372.
[10]
Heinz
e
l
an W
,
Cha
ndrak
asa
n
A. An applicat
ion sp
ec
ific pro
t
ocol arch
itectu
re for
w
i
r
e
less
microsens
or
net
w
o
rks.
On W
i
reless Co
mmu
n
ic
ations, C
o
mmunic
a
tio
n
society
. 200
2; 1(4): 660-
67
0.
[11]
He N
i
ng
hu
i, Li
Hon
g
sh
eng,
Gao Jin
g
. En
e
r
g
y
-s
avin
g R
o
uting A
l
g
o
rithm
Based
on
Cl
u
s
ter in W
S
N.
T
E
LKOMNIKA Indon
esi
a
Jour
nal of Electric
al
Engin
eeri
ng.
2
013; 11(
2): 835
-847.
Evaluation Warning : The document was created with Spire.PDF for Python.