TELKOM
NIKA
, Vol.13, No
.2, June 20
15
, pp. 711 ~ 7
2
1
ISSN: 1693-6
930,
accredited
A
by DIKTI, De
cree No: 58/DIK
T
I/Kep/2013
DOI
:
10.12928/TELKOMNIKA.v13i2.1438
711
Re
cei
v
ed
Jan
uary 28, 201
5
;
Revi
sed Ap
ril 7, 2015; Accepte
d
April 2
5
, 2015
Adaptive Energy-Aware Cluster Based Routing
Protocol for Mobile Ad Hoc Networks
Fateme
h Ha
kimifar
1
, Se
y
e
d-Amin Ho
sseini-Sen
o*
2
, Mohammad Hoss
ein Moattar
3
, Thair
Al-Dala’in
4
, Rahm
a
t
Bu
d
i
arto
4
1
Department o
f
Computer En
gin
eeri
ng, Pa
yamno
o
r Univ
er
sit
y
of Sari, Ira
n
2
Computer Net
w
o
r
k Res
earch
Labor
ator
y
,
F
e
rdo
w
si U
n
iv
ersi
t
y
of Mashh
ad,
Iran
3
Department o
f
Soft
w
a
re En
gi
neer
ing, Islami
c Azad Univ
ers
i
t
y
of Mash
had,
Iran
4
Colle
ge of Co
mputer Scie
nc
e and Inform
ati
on T
e
chnol
og
y, Al Baha Univ
ersit
y
, Sa
udi Ar
abi
a
*Corres
p
o
ndi
n
g
author, em
ail
:
hossein
i@um
.ac.ir
A
b
st
r
a
ct
Due to th
e do
w
n
side ch
arac
teristics of Mo
bile A
d
h
o
c N
e
tw
orks (MANET
s
) such as
dyna
mic
topol
ogy a
nd
ener
gy consu
m
pti
on a
nd c
ontrol ov
erhe
a
d
, netw
o
rk clus
tering is o
n
e
of the promi
s
in
g
soluti
ons. Cl
us
ter Based R
o
u
t
ing Protoc
ol (
C
BRP) is a ro
bust an
d scal
a
ble ro
utin
g pro
t
ocol for MAN
E
T
s
.
Clusteri
ng for
m
ation a
l
gor
ith
m
used in CBRP
is a variat
ion o
f
simpl
e
low
e
st-ID algorith
m
i
n
w
h
ich the nod
e
w
i
th a low
e
st ID a
m
o
ng its
n
e
ig
hbors
is e
l
e
c
ted as
th
e Cl
uster he
ad. N
e
glecti
ng
mo
bi
lit
y and
en
ergy f
o
r
selecti
ng cl
uster he
ad is
one
of the
w
eakne
ss points of th
e alg
o
rith
m. In
order to i
n
cre
a
se stab
ility of
th
e
netw
o
rk an
d to
preve
n
t re-cl
u
stering
an
ada
ptive e
ner
gy-a
w
a
re Cluster B
a
sed
Ro
uting
Protocol (AE
C
BRP)
is prop
osed. T
w
o algorith
m
s
have b
e
e
n
intr
oduc
ed i
n
AEC
BRP as enh
an
cement to the CBRP: improvi
n
g
the cluster for
m
ati
on a
l
g
o
rith
m by co
nsi
deri
ng rel
a
tive
mo
bility, resi
du
al
ener
gy an
d co
nnectiv
i
ty degr
e
e
metrics, a
nd a
dd in a
n
efficie
n
t cluster main
tenanc
e al
gorit
hm b
a
se
d on t
he ag
gre
gate
ener
gy metric of
cluster he
ad. Using NS-
2
w
e
eval
uate th
e rate
of clus
ter-hea
d cha
n
ges, the nor
mali
z
a
ti
on ro
utin
g
overh
ead
a
n
d
the
pack
e
t d
e
livery
rati
o.
Co
mp
aris
o
n
s
den
ote th
at th
e pr
opos
ed
A
E
CBRP
has
b
e
tter
perfor
m
a
n
ces
w
i
th respect to the orig
ina
l
CB
RP and Cr
oss-
CBRP.
Ke
y
w
ord:
Rou
t
ing in MANET
s
, CBRP, Cluster Formati
on A
l
gorit
hm, R
e
lati
ve Mobi
lity, Re
sidu
al Ener
gy
1. Introduc
tion
A Mobile Ad
hoc
Network (MA
N
ETs) includ
es
a set of wi
rele
ss
node
s
wh
ich can
comm
uni
cate
dynami
c
ally throu
gh
wireless m
u
lti-h
op net
wo
rks.
The
s
e n
e
tworks
can
be
config
ure
d
wi
thout an
infra
s
tru
c
ture o
r
centrali
zed
ad
ministration t
o
be
controlle
d. Each n
e
twork
node
can o
n
l
y comm
uni
cate directly
with n
ode
s t
hat
are in
its radio
ra
ng
e, therefo
r
e,
it is
requi
re
d that the node
s
perfo
rm routi
ng fun
c
tion
dedi
catedly. In MANET, due to net
work
dynamic stru
cture and
l
a
cking ce
ntrali
zed
man
age
m
ent, routing i
s
carried
out
by all availa
ble
node
s an
d via multi-ho
p way [1].
MANETs rout
ing
protocols can be
cl
assi
fied into flat routing an
d hi
era
r
chical ro
u
t
ing. In
the flat routing schem
e, each nod
e o
n
a rout
e re
cords th
e ph
ysical n
e
xt hop towa
rd
s the
destin
a
tion a
s
its next hop for that rou
t
e. In
fa
ct, in these proto
c
ols, all node
s are enga
ged
in
routing fun
c
ti
on. So they incre
a
se co
ntro
l
packet overhead for
rout
e discovery p
r
ocess [2].
The hierarchi
c
al routin
g prot
ocol
s i
m
pro
v
e network
p
e
rform
a
n
c
e
s
esp
e
ci
ally wh
en the
netwo
rk size
i
n
crea
se
s. Clusteri
ng sche
mes
are
ty
pically
used by hiera
r
chi
c
al routing proto
c
ols.
The cl
uste
r b
a
se
d ro
uting
proto
c
ol
s de
crea
se the
nu
mber
of enga
ged no
de
s in
route a
nd al
so
the si
ze of n
e
ighb
or tabl
e. More
over
cl
usteri
ng i
s
on
e of the app
roache
s appli
ed in de
crea
sing
the traffic duri
ng the route
discov
ery
process [3],[4]. CBRP
is
a
ro
uting protocol
that is designed
for routin
g in
MANETs wi
th many nod
es. The
whol
e netwo
rk i
s
divided into
overlappi
ng
or
disjoi
nt cluste
rs. The no
de
which h
a
s bi
-directi
o
nal link and the lo
we
st ID amo
ng its neigh
b
o
rs
are ele
c
ted
a
s
clu
s
te
r-h
ea
d. The nod
e mobility
cau
s
es net
wo
rks
topology to chang
e fast [3].
Since e
a
ch cluster i
s
reco
gnized by its clu
s
ter h
ead
, which is ful
l
y depend
ent
on the cl
ust
e
r
head beh
aviour, cluste
rin
g
mechani
sm
dire
ctly
influen
ce
s the
overall n
e
twork pe
rform
a
nce.
Therefore,
a
wi
se
clu
s
te
r formatio
n a
s
a
ma
i
n
stre
am p
a
rt of
CBRP
may i
m
prove
net
work
perfo
rman
ce.
Cluste
rin
g
a
l
gorithm of
CBRP
due to
not co
nsid
eri
ng the mo
bility and node
'
s
energy which
are
con
s
ide
r
ed a
s
t
w
o M
A
NETs li
mi
ta
tions,
cau
s
e
s
the
wea
k
ne
ss of th
e routi
n
g
proto
c
ol. T
o
improve
clu
s
ter hea
d el
ection,
a
n
e
w
clu
s
terin
g
algo
rithm i
s
introd
uced
that
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 : 711 – 72
1
712
con
s
id
ers rel
a
tive mobility, resid
ual e
n
e
rgy an
d con
nectivity degree of no
de
s. In addition, t
he
clu
s
ter
stabili
ty is maintai
ned by
an al
gorithm th
at
con
s
id
ers the
agg
reg
a
te e
nergy m
e
tri
c
of
clu
s
ter he
ad
s.
The re
st of the pape
r is
orga
nized as follo
ws. Section 2 explains the CBRP briefly.
Section 3 giv
e
s a b
r
ief su
mmary of rel
a
ted wo
rks.
S
e
ction 4
pro
p
ose
s
an
effici
ent clu
s
ter
ba
sed
routing p
r
oto
c
ol (AECBRP
)
. Section 5 discu
s
ses
si
mul
a
tion setup, a
nd re
sults. Fi
nally, Section
6
pre
s
ent
s the pape
r’s
con
c
l
u
sio
n
s.
2.
Ov
er
v
i
e
w
of CBRP
The CB
RP is a distribute
d
, efficient and scalabl
e prot
ocol that use
s
clu
s
teri
ng a
ppro
a
ch
to decrea
s
e the traffic of route discove
r
y me
ssa
ge
s in the netwo
rk. CBRP ha
s less ove
r
he
ad
and
highe
r th
roug
hput
co
mpared to
Ad ho
c
On-De
m
and
Di
stan
ce Ve
cto
r
(A
ODV) protocol. I
n
this p
r
oto
c
ol
the wh
ole
ne
twork i
s
divid
ed into
overl
appin
g
o
r
di
sjoint clu
s
te
rs.
Each
clu
s
te
r
contai
ns a cl
uster-h
ead, g
a
teway
s
and
membe
r
s. A gateway is a node throug
h
which m
e
mb
er
node
s comm
unicate with
the adja
c
ent
cluste
r-he
a
d
.
The clu
s
tering algo
rithm
of CBRP is a
variation of
si
mple
lo
we
st-I
D clu
s
terin
g
algorith
m
. Th
e nod
e
with t
he lo
we
st-ID
in its n
e
igh
b
o
u
rs
is el
ecte
d a
s
clu
s
te
r-h
ead
. Each
cl
ust
e
r-hea
d
con
s
iders
all n
e
ig
hbou
rs havin
g bi
-directio
n
al
links, as me
mbers. Each
node m
a
intai
n
s a n
e
igh
b
o
u
r table
(NT
)
and a
clu
s
ter adjacen
cy table
(CAT
). The
n
e
ighb
our tabl
e is
used fo
r receivin
g t
he li
nk
statu
s
for
sen
s
in
g an
d f
o
rmin
g cl
uste
rs.
The cl
uste
r a
d
jacen
c
y table kee
p
s th
e informatio
n
of adjacent clu
s
ters an
d is
use
d
by CBRP'
s
Adjace
nt Clu
s
ter
Discove
r
y Proce
dure. These t
able
s
are up
date
d
by peri
odi
c hello me
ssa
ge.
The h
e
llo
me
ssage
in
clud
es th
e n
ode
ID, the
nod
e role (clu
ste
r-h
ead, me
mbe
r
, unde
cid
ed).
If
the hell
o
me
ssag
e i
s
n
o
t re
ceived
from
a
sp
ecifi
c
n
ode
, that entry
wi
ll be
rem
o
ved
from th
e tabl
e
[5].
CBRP is ba
sed on source
r
outing that usin
g clu
s
ter
stru
cture to minimize the floodin
g
traffic du
ring t
he ro
ute di
scovery
process. Furth
e
rm
ore,
the use of
uni-di
r
e
c
tiona
l links in
cre
a
se
s
the net
wo
rk
con
n
e
c
tivity.
In ro
ute di
scovery p
r
o
c
ed
ure
cl
uste
r-h
ead
s
sea
r
chi
ng fo
r a
sou
r
ce
route a
r
e flo
oded
with Route Re
que
st (RREQ
)
Packets. Th
e clu
s
ter-he
ad
forwa
r
d
s
RREQ
packet only o
n
ce a
nd neve
r
sen
d
s it to a
node t
hat ha
s alre
ady re
corde
d
in the route [6].
The advanta
ge of CBRP is that only cluster-
h
ead
s
excha
nge rou
t
ing informati
on. Thus,
comp
ared to
the traditio
n
a
l floodin
g
method
s,
the
cont
rol ove
r
head t
r
an
smi
tted is far le
ss.
Ho
wever, CBRP
is
li
ke
other hierarchical ro
uting
proto
c
ol
s th
at ha
s cl
ust
e
r fo
rmation
and
maintena
nce overhe
ad.
For pe
rform
ance optimi
z
ati
on, CBRP
re
comme
nd
s a
sho
r
t
enin
g
route. Sin
c
e CBRP
use
s
a sou
r
ce
routing sch
e
me,
a
no
de gets all
in
fo
rmation a
bout
route
wh
en
receivin
g a
pa
cket.
Nod
e
s expl
oit route sho
r
te
ning a
s
next hop to minimi
ze the h
op n
u
mbe
r
and
a
dapts to n
e
twork
topology cha
nge
s to cho
o
se the m
o
st distant
nei
ghbo
ring n
o
d
e
in a route.
Local
rep
a
ir is
anothe
r opti
m
ization met
hod that is employed by
CBRP. It checks the ro
uting informa
t
ion
contai
ned in
the pa
cket wheneve
r
a n
ode ha
s a
p
a
cket to forward a
nd the
next hop is
not
rea
c
ha
ble. In a route, if the
next hop or the hop
after the next hop is rea
c
ha
ble through on
e of its
neigh
bors, the packet is fo
rwa
r
d
ed thro
ugh the ne
w route [7].
3. R
e
lated
Works
The clu
s
te
rin
g
algorithm
s divide MANETs into clu
s
ters. Clu
s
te
r head
s man
age the
clu
s
ter a
nd
communi
cate
with othe
r cl
u
s
ters. Clu
s
te
ring algo
rithm
con
s
tru
c
t a l
ogical topolo
g
y
for routing al
gorithm an
d allows feedb
ack from ro
ut
ing algo
rithm
in order to a
d
just that logi
cal
topology
an
d ma
ke
cl
u
s
terin
g
de
cisions. S
o
th
e cl
uste
r-he
ad
stability
is im
porta
nt for
perfo
rman
ce
of networks [6].
The lo
we
st I
D
alg
o
rithm
[8] is the
mo
st comm
on te
chni
que to
ra
ndomly
sele
ct clu
s
ter
head
s. Ea
ch
nod
e i
s
i
d
e
n
tified by
a
uniqu
e ID,
a
nd the
n
ode
with
the l
o
we
st ID in i
t
s
neigh
borhoo
d
is co
nsi
dere
d
as
clu
s
ter h
ead. Since
this heu
ri
stic is
biased to cho
o
se n
ode
s
with
smalle
r IDs a
s
clu
s
ter he
a
d
s, those n
o
des with
sma
ller ID's
suffe
r from the ba
ttery drainag
e,
resulting sho
r
t lifetime span of the networks.
In the high
e
s
t co
nne
ctivity cluste
ring
(HCC
) al
gorit
hm [9] the d
egre
e
of a
n
ode i
s
comp
uted
ba
sed
on its di
stance f
r
om
others. Each
n
ode b
r
oa
dcasts its ID to
th
e nod
es th
at
are
within its tran
smissio
n
ran
ge. The
nod
e with m
a
xim
u
m num
be
r
of neigh
bou
rs (i.e., maxi
mum
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
1693-6
930
Adaptive En
e
r
gy-Aware Cl
uster Ba
se
d Routin
g Pr
oto
c
ol for Mo
bile
.... (Fatem
eh Hakim
i
far)
713
degree
) is ch
ose
n
as a
cl
uster h
ead. S
i
nce the n
o
d
e
is forced to
leave its clu
s
ter after find
ing
anothe
r clu
s
t
e
r hea
d with t
he high
er con
nectivity,
the
clu
s
ter he
ad
s do not play their role well for
very long. S
o
this
algo
rithm con
s
tru
c
ts un
st
abl
e cl
usters. Whe
n
e
ver
the nu
mber of
ordi
nary
node
s in a cl
uster in
crea
ses, efficien
cy and net
work
perfo
rman
ce
degrade
s.
Adaptive mul
t
i-hop
clu
s
te
ring [10]
set
s
upp
er
and
l
o
we
r b
oun
ds (U a
nd
L)
on the
numbe
r of cl
u
s
ter-mem
bers within
a gro
up that
a cl
uster hea
d can
handl
e. Whe
n
the num
ber of
clu
s
ter-mem
b
ers in a
cl
ust
e
r i
s
le
ss tha
n
the lo
we
r b
ound, the
cl
u
s
ter
nee
ds to
merg
e
with
one
of the neig
h
b
ourin
g cl
uste
rs. On the
co
ntrary, if
the
numbe
r of
cl
uster-me
mbe
r
s i
n
a g
r
ou
p
is
greate
r
than t
he upp
er bo
u
nd, the clu
s
te
r is divided int
o
two clu
s
ters.
For mobility based
cluster format
ion, Lowest Relative Mobility clustering [11] applies a
new met
r
ic
. A relative mobility with res
p
ec
t to a
neighb
or is a
c
hieve
d
usin
g the ratio of receiv
ed
power between
two succ
essive packets. In [4] this rel
a
tive
mobility
techni
que is
used and
Cross-
CBRP
routin
g proto
c
ol
is i
n
trodu
ce
d. It is a n
e
w
cr
oss
-
la
yer
ap
p
r
oa
c
h
to
for
m
a c
l
us
te
r
in
whic
h
each node achieves its m
obility by the received po
wer level
s
of two hello message from
each
neighbor. If
each node has M
neighb
ours, so it will have M
values
relativ
e
mobility that
aggregate
ap
proa
ch i
s
introdu
ced in thi
s
work.
Every node sets t
he agg
re
gate
mobility in hello
messag
e an
d broa
dcast
to other nod
es. To achie
v
e the maximum stability
,
a node wit
h
the
lowest aggregate mob
ility is sel
e
cted as
the cluster-head.
The limitation
s
of the
aforement
ion
ed
algorith
m
are
that to form
the clu
s
te
rs t
hey only
con
s
id
er a si
ngle feature o
f
a node.
The
weig
hted
clu
s
te
ring
al
gorithm
(WCA) [12],[13], i
s
b
a
sed
on t
he u
s
e
of a
combine
d
weig
ht metri
c
that take
s into accou
n
t se
veral system para
m
eters like
the node
-de
g
ree,
distan
ce
s wit
h
all its neigh
bours, nod
e spe
ed and
th
e time spent
as a clu
s
te
r-h
ead. Each n
o
de
obtain
s
the
weight valu
es
of all oth
e
r
no
des an
d
info
rmation of
oth
e
r
clu
s
ter he
a
d
s i
n
the
sy
stem
throug
h
reb
r
oad
ca
sting.
As a
result, the ove
r
he
ad
indu
ce
d
by
WCA
is very
high.
If a
n
ode
moves i
n
to a
regio
n
that i
s
not covere
d b
y
any
cl
uste
r-head, th
en th
e cl
uste
r
set-up p
r
o
c
ed
ure
i
s
invoke
d throu
ghout the
wh
ole sy
stem. T
h
is le
ad
s to o
v
erhea
ds. In
addition, in
th
is alg
o
rithm
the
node
spe
ed i
s
u
s
ed a
s
a
mobility prop
erty whe
r
ea
s the relative
mobility between nei
ghb
o
u
ring
node
s si
gnificantly affects cluster
stability.
Tao et. al. [6] select a
cluster head based on
the rel
a
tive mobility with the
connectivity
degree is u
s
ed. The relati
ve mobility m
e
tric is o
b
tain
ed usin
g the locatio
n
information provid
ed
by Glob
al Po
sitionin
g
Syst
em (GPS)
an
d velo
city. Ho
wever,
ene
rg
y metric is no
t con
s
id
ere
d
.
If a
node
with lo
we
st mobility and hig
h
e
s
t con
n
e
c
tivity i
s
selecte
d
a
s
a clu
s
ter-he
ad whil
e ha
s
little
resi
dual e
nergy, the cluste
r must be
re
c
onstructe
d. It
prod
uces a hi
gh overh
ead.
4.
The Propos
e
d
AECBRP
As me
ntione
d
previo
usly i
n
Section
2,
cl
uste
r formatio
n alg
o
rithm
in
CBRP
is a v
a
riation
of sim
p
le l
o
west-I
D
clu
s
teri
ng al
gorith
m
s in
whi
c
h th
e
node
with
a l
o
we
st ID am
o
ng its nei
ghb
ors
is ele
c
ted as the Cluste
r-head. We propo
se
a protocol n
a
med
as AECBRP
whi
c
h enh
an
ce
s
CBRP in te
rms of: (i
) ele
c
ting the
clu
s
ter he
ad by
t
a
kin
g
into a
c
cou
n
t its mo
bility, conne
ctivity
degree a
nd resid
ual en
erg
y
. (ii) maintai
n
ing the fo
rm
ed clu
s
te
rs
b
y
con
s
ide
r
ing
the agg
regat
e
energy metric of cluster he
ads. The det
ails of
the enhan
ceme
nts
are de
scrib
e
d
in the following
se
ct
ion
s
.
4.1. Net
w
o
r
k Model
Let us con
s
id
er a net
work
rep
r
e
s
ente
d
by a graph
G (
V
, E)
, where
V
is the nu
mber of
node
s an
d
E
is the numb
e
r
of bi-directi
onal lin
ks. Interme
d
iate no
des h
e
lp ea
ch sou
r
ce nod
e to
sen
d
data to
a de
stinatio
n nod
e. If
N
x
is the n
u
mb
er of nei
ghb
our n
ode
s
x
,
C
degree
(x)
is th
e
con
n
e
c
tivity d
egre
e
of nod
e
x
that is de
fined by the numbe
r of ne
ighbo
rs in the
neighb
or tab
l
e.
C
degree
(x, y)
i
ndicates that
the node
x
ge
ts the con
n
e
c
tivity degree
of node
y
.
(1)
We a
s
sume that each nod
e aggregate
s
the
conne
cti
v
ity degree o
f
its neighbo
rs. The
aggregate
conne
ctivity degre
e
of n
o
de
x
is an
averag
e of t
he conn
ectiv
i
ty degre
e
of
its
neigh
bou
rs, i
s
define
d
in Equation (2).
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 : 711 – 72
1
714
x
N
y
ree
ree
ree
y
x
C
x
C
x
AC
)
,
(
)
(
1
)
(
deg
deg
deg
(2)
In mobile
ad
hoc net
works du
e to
ran
dom move
of
node, i
n
ste
a
d
of con
s
ide
r
ing the
speed of
nodes movement, the rel
a
tive mobilit
y is used. By
compar
i
ng the receive si
gnal
strength of neighbors with
the pervious value in cache. The rel
a
tive mobility
)
(
x
M
rel
y
b
e
twee
n
n
y
and
n
x
, is defi
ned in Equati
on (3
):
(3)
Whe
r
e
y
x
P
R
new
r
x
is the powe
r
cu
rre
n
t node
n
y
tha
t
has re
ceive
d
from
n
x
,
y
x
P
R
old
r
x
is the
power n
ode
n
y
that has
previously recei
v
ed from
n
x
. If
0
)
(
x
M
rel
y
, it indicates that two nod
es
are
gradu
ally
moving
away,
othe
rwi
s
e
th
e two
n
ode
s
are
moving
cl
ose
to
ea
ch
o
t
her. Sup
p
o
s
e a
node with
M
neigh
bors, it has
M
number relative values that t
he aggregate local
mobility values
[4] is calculat
ed usi
ng Equ
a
tion (4
).
(4)
All node
s in
MANET a
r
e
mobile
with li
mited ene
rgy
sou
r
ces,
whi
l
e any comm
unication
in a netwo
rk involves en
ergy con
s
um
ption. T
he en
ergy con
s
um
ption of each n
ode de
pen
ds on
its sen
d
ing a
nd re
ceiving t
r
an
smi
ssi
on
[14] as expressed in Equ
a
tion (5
)
(5)
M
and
D
a
r
e con
s
tants,
rep
r
e
s
entin
g the
protocol used, se
nding
and
receivin
g
informatio
n a
nd, are dete
r
mined
by the
hardware.
T
able 1
sho
w
s the en
ergy
con
s
um
ption
in
v
a
riou
s st
at
e
s
.
Table 1. Power co
nsumpti
on mea
s
u
r
em
ents [14]
Par
a
meter
M(
µ W. sec)
D(
µ W. sec)
Broadcast Send
1.9
266
Point to point Send
1.9
454
Broadcast Receive
0.50
56
Point to point Re
ceive
0.50
356
Idle 843
(m W
)
Each n
ode
cal
c
ulate
s
its resi
dual e
n
e
rgy de
pen
ding on its
se
nding a
nd receivin
g
informatio
n. This value in e
v
ery moment
is cal
c
ul
ated
usin
g Equatio
n (6)
(6)
Having
don
e
the calcul
atio
n of the
re
sid
ual en
er
gy of
node
s, this value i
s
set in
the hello
messag
e
and
broad
ca
sted
amo
ng
ea
ch
othe
r.
E
residual
(x,y)
in
dicat
e
s th
at n
ode
x
r
e
c
e
ives
th
e
resi
dual e
nergy of node
y
.
(7)
4.2. The neighbor ta
ble and hello message
forma
t
in AECBRP
In this pape
r we, extend the structu
r
e
of nei
ghb
orin
g table by ad
ding 4 field
s
, inclu
d
ing
relative mo
bi
lity, aggreg
ate mobility, resid
ual
e
nergy and
co
nn
ectivity degree a
s
sho
w
n in
Figure 1.
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
1693-6
930
Adaptive En
e
r
gy-Aware Cl
uster Ba
se
d Routin
g Pr
oto
c
ol for Mo
bile
.... (Fatem
eh Hakim
i
far)
715
Neig
hbo
rID
Neig
hbo
r
Status
Link
status
Relativ
e
mobility
Aggre
gate
mobility
Re
sidu
al
Energy
C
o
nn
ec
tivity
degree
Figure 1. Nei
ghbo
r Tabl
e in AECBRP
This informat
ion i
s
o
n
ly u
s
ed
to fo
rm
a cl
uste
r. Ea
ch
nod
e le
arns i
n
form
atio
n from
received h
e
ll
o messa
ge.
The hell
o
m
e
ssag
es
co
ntain not only
a neigh
bo
r table an
d cl
u
s
ter
adja
c
en
cy table, but also
other info
rma
t
ion of
node
x, including
a
ggre
gate mo
bility, conne
ct
ivity
degree an
d resid
ual ene
rg
y (see Fig
u
re
2).
Nod
e
ID
Nod
e
Status
Aggre
gate
mobility
Re
sidu
al
Energy
C
o
nn
ec
tivity
degree
Neig
hbo
rID
Neig
hbo
r
Status
Link status
… …
…
Clu
s
ter adj
acent ID
…
Figure 2. Hell
o messag
e in
AECBRP
4.3. Cluster
Formation
Algorithm
The
basi
c
idea of the
cluster form
ation algori
thm
i
s
to
consi
der
mobility, connectivity
degree
and
the re
sid
ual
energy of n
ode
s to
sele
ct a
clu
s
ter.
The
clu
s
ter head
form
a
t
ion
algorith
m
is d
e
scrib
ed a
s
follows.
1.
All node
s
sta
r
t wo
rki
ng i
n
unde
cid
ed
state and
se
t the timer with
the spe
c
ific
time interval
and broad
ca
st a hello messag
e. Every node b
r
oa
dc
a
s
ts its o
w
n m
obility, conne
ctivity degree
and re
sidu
al energy
(
M
and
C
degree
are i
n
itialize
d
to
0
and
en
ergy i
s
initiali
ze
d t
o
40
0 at
the
begin
n
ing of
operation
s
) in
a hello message to its 1-h
op neig
hbo
rs.
2.
By receivin
g
the hell
o
m
e
ssag
e, a n
ode
com
pare
s
its agg
reg
a
te
mobility values with
i
t
s
neigh
bors an
d the node
wi
th the lo
we
st aggregate m
obility value
M(
x)
<
M(
y)
is considered.
3.
In additio
n
th
e no
de
com
p
are
s
its conn
ectivity
degre
e
with
the
ag
greg
ate
con
n
ectivity degre
e
of its neig
hbo
rs a
nd the
no
de with
th
e hi
ghe
st co
nne
ctivity degree
C
degree
(x) >
A
C
degree
(x)
is
c
o
ns
ide
r
ed
.
4.
At the end the node
with the high
est re
sidu
al ene
rgy
E
resid
ual
(x) >
E
residua
l
(y)
is
selected.
A node
can
be a
clu
s
ter-head if it ha
s less mobility
,
more resi
d
ual en
ergy a
nd mo
re
con
n
e
c
tivity
degree to its neighb
ors. T
h
is n
ode
w
ill
cha
nge it
s st
ate to clu
s
ter-hea
d state.
By
broa
dcastin
g
hello m
e
ssa
ge, all n
ode
s having
bi-d
i
r
ectio
nal lin
ks with
this
cl
uster-h
ead, a
r
e
recogni
ze
d a
s
memb
ers.
4.4. Cluste
r Mainten
a
nc
e
Algorithm
Whe
n
cluste
rs a
r
e fo
rme
d
,
to preve
n
t sud
den
de
crement of
clu
s
ter-he
ad e
n
e
rgy, the
clu
s
ter-he
ad
aggregate
s
t
he resi
dual
energy of it
s membe
r
s a
nd continu
o
u
s
ly co
mpa
r
e
s
its
resi
dual
en
ergy with thi
s
aggregate
va
lue. When
t
he
clu
s
ter-he
ad en
ergy is less tha
n
t
he
aggregate
en
ergy of it
s
cl
uster mem
b
e
r
s, th
e
cl
uste
r-h
ead
chang
es to
mem
b
e
r
st
ate an
d t
h
e
clu
s
ter format
ion algo
rithm
is perfo
rme
d
again in the
same
clu
s
ter.
It is worth to
note that after
cha
ngin
g
the
clu
s
ter-h
ead
nod
e
state t
o
a m
e
mb
er,
the
clu
s
ter
doe
s n
o
t re
st
ructu
r
e,
and
the
node
with the
highe
st resi
d
ual ene
rgy in that cluste
r wi
ll be the clu
s
ter-hea
d.
Gene
rally, the purpo
se of
the pro
p
o
s
e
d
algo
rithm i
s
to prevent
the reforma
t
ion of
clu
s
ters. This appro
a
ch cre
a
tes sta
b
le cl
usters.
5.
Simulation setup an
d res
u
lts
To evalu
a
te t
he p
r
op
osed
proto
c
ol, the
simulato
r
NS
-2(ve
r
si
on
2.34) i
n
Ubu
n
tu 10.0
4
environment was
per
form
ed. The mobi
lity scenarios that
use the Random Way Point mobility
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 : 711 – 72
1
716
model with 30-1
30 nod
e
s
a
nd rand
o
m
ly
distri
b
u
ted in
a 67
0
m
×67
0
m a
r
e
a
are rand
o
m
ly
gene
rated. T
able 2 dem
on
strate
s the si
mulation pa
ra
meters.
Table 2. Simulation setting up paramet
ers
Parameter Values
Simulation Duration
600s
Pause time
0s
Maximum Speed
of the node
5-30 m/s
Transmission range
150- 250m
Packet Rate
4 pkt /sec
Number of
nodes
30-130
Traffic Model
CBR
Max connection
40
Initial Energ
y
400j
Area
670m×670mm
5.1. In
v
estigation of clu
s
ter hea
d
cha
nges in net
w
ork conditio
n
In the first scenari
o
, the numbe
r of clu
s
te
r head
cha
nge
s is illust
rated again
s
t the sp
eed
cha
nge
s. Th
e numbe
r of
cluste
r he
a
d
cha
nge i
s
the total number of cl
uster head
ch
a
nge
s
durin
g the
wh
ole si
mulatio
n
ru
n time. A
small va
l
ue
of clu
s
ter
hea
d ch
ang
e reflects the
stabi
lity
of the
clu
s
ter structu
r
e. Fi
gure
1
dem
o
n
strate
s
the rate
of clu
s
ter-hea
d cha
n
g
e
s
i
n
cre
a
se
s by
increa
sing
th
e spee
d of
n
ode
s. Due
to
mobilit
y in
crement, the
n
e
twork topol
o
g
y is serio
u
sly
cha
nge
d and
the clu
s
ter f
o
rmatio
n ope
ration
s are
re
peated. F
r
om
Figure
3, it is found
out that
the prop
osed
protocol, co
nsid
er mobilit
y, ener
gy and conn
ectivity degree du
ri
ng the sele
ct
ing
clu
s
ter, ha
s b
e
tter perfo
rm
ance com
p
a
r
ed to
the origi
nal CBRP a
n
d The Cro
s
s-CBRP.
In the secon
d
scen
ari
o
, the rate of
cl
us
ter-h
ead
chang
es i
s
computed
versu
s
the
transmissio
n rang
e ch
ang
e
s
. Figure 4 sh
ows that
by increa
sing the
transmi
ssion
range, the ra
te
of clu
s
ter-he
ad chang
es
decrea
s
e
s
.
Having d
one
increa
sing t
he tran
smi
ssion ra
nge, m
o
re
node
s
are
wit
h
in the
range
of othe
r n
ode
for l
ong
er
pe
riod
s of
time.
Hen
c
e, l
e
ss
o
f
larg
e
clu
s
ters
formed a
nd their mobility doe
s not allo
w them to
move freque
ntly in and out of range of e
a
ch
other. T
herefore, the
nu
m
ber
of cl
uste
r-hea
d
ch
a
n
g
e
s
de
cre
a
ses. Whe
n
the
transmi
ssion
range
is de
cre
a
sed
the rate of cluster-h
ead chang
es
in th
e AECBRP will get better perfo
rma
n
ce in
comp
ari
s
o
n
with the origi
nal CBRP a
n
d
The Cro
s
s-CBRP.
Figure 3.
Nu
mber of Cl
ust
e
r-hea
d Ch
an
ges vs. Node
Speed
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
1693-6
930
Adaptive En
e
r
gy-Aware Cl
uster Ba
se
d Routin
g Pr
oto
c
ol for Mo
bile
.... (Fatem
eh Hakim
i
far)
717
Figure 4.
Nu
mber of Cl
ust
e
r-hea
d Ch
an
ges vs. Tran
smissi
on Rang
e
In the thi
r
d
scen
ario,
the
rate
of cl
uste
r h
ead
chan
ges versu
s
t
he n
u
mb
er
o
f
node
'
s
c
h
an
g
e
is calc
u
l
a
t
ed
. As
s
h
ow
n in
F
i
gu
r
e
5
,
by i
n
crea
sing
the
numbe
r
of n
ode
s the
rat
e
of
clu
s
ter hea
d
cha
nge
s i
n
creases. A
s
th
e no
de
den
sit
y
increa
se
s,
AECBRP p
r
o
duces con
s
ta
ntly
less num
ber
of clu
s
ter h
e
ad chang
es i
n
com
p
a
r
iso
n
with the
CBRP and
Cross-CBRP. A
s
a
result AECB
R
P give
s
better
perfo
rma
n
c
e i
n
te
rm
s of
the
numb
e
r
of clu
s
te
r h
e
a
d
chan
ge
s
when
the node d
e
n
s
ity in the network is hi
gh.
Figure 5.
Nu
mber of Cl
ust
e
r-hea
d Ch
an
ges vs. Numb
er of node
s
In the fourth
scena
rio, the
numbe
r of clus
ter
hea
d chang
es i
s
ca
lculate
d
agai
nst the
cha
nge of p
ause time. Whe
n
pau
se
time increa
se
s the re
qu
ired n
u
mbe
r
of cluste
r h
ead
cha
nge
s are very low. Figure 6 indi
cat
e
s that wh
e
n
the pause ti
me is 0 s, the most mobili
ty is
within the
ne
twork an
d it is the result of incr
ea
sing
clu
s
ter h
ead
cha
nge
s. In
the pau
se ti
me
600s, no m
o
bility is in the network, the
rate of
cl
uster head
changes i
s
zero. From Figure
7 i
t
is
clea
r that AECBRP pe
rforms better tha
n
both,
the ori
g
inal CB
RP a
nd the Cross-CBRP.
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 : 711 – 72
1
718
Figure 6.
Nu
mber of Cl
ust
e
r-hea
d Ch
an
ges vs. Pau
s
e Time
In the fifth scen
a
rio, the
numbe
r of cl
us
ter
hea
d chang
es i
s
ca
lculate
d
agai
nst the
cha
nge of pa
cket rate. Wh
en traffic inje
ction to
the n
e
twork in
crea
se
s so
me rea
s
on
s can
cau
s
e
packet
s
do
not received by a downs
t
r
eam node, for example, lack
of route or
impossibility to
acce
ss to the
medi
a –
so p
a
ckets will
ho
ld in th
e inte
rf
ace
qu
eue. If
this b
u
ffer
overflows the l
a
st
incomi
ng p
a
cket will di
scard. Ther
efore,
if there are some hell
o
pa
cket in this q
ueue the
s
e
h
e
llo
packet
s
rea
c
h to the
nei
ghbo
urs n
o
d
e
s by
del
ay.
The
r
efore
Neig
hbo
r tab
l
e upd
ates a
n
d
informatio
n a
bout the
status of n
e
igh
b
o
ring
nod
es
can
be d
e
lay
ed an
d in
cre
a
se
s the
rate
of
clu
s
ter he
ad
cha
nge.
Figure 7.
Nu
mber of Cl
ust
e
r-hea
d Ch
an
ges vs. pa
cke
t
rate
5.2. In
v
estigation on nor
malization r
outing ov
erhead in net
w
o
r
k condition
In the sixth scen
a
rio, the r
outing overhe
ad metri
c
is compa
r
ed to speed
cha
nge
s. Thi
s
metric
dete
r
mines the ov
erhe
ad
cau
s
ed by tra
n
sm
itting routing
packet
within
the network
and
the metri
c
eq
uals the fracti
on of th
e n
u
m
ber of
s
ent
routing
pa
cke
t
on the
num
ber
of all
re
ce
ived
data packet. Figure 8 dem
onstrates that
incre
a
si
ng the spee
d of node
s will increase the routi
ng
overhe
ad.
In
cre
a
si
ng sp
e
ed cau
s
e
s
a
fast
chan
ge
of the net
wo
rk top
o
logy b
e
ca
use with
this
cha
nge, no
de
s will exchan
ge more routi
ng messa
g
e
s
.
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
1693-6
930
Adaptive En
e
r
gy-Aware Cl
uster Ba
se
d Routin
g Pr
oto
c
ol for Mo
bile
.... (Fatem
eh Hakim
i
far)
719
Figure 8. Normalizatio
n Ro
uting Overh
e
ad vs. Speed
of node
s
In the seven
t
h sce
nari
o
, norm
a
lization
rout
ing ove
r
hea
d is calculate
d
agai
nst the
cha
nge of n
u
m
ber of n
ode
s. From Fi
gu
re 9 we fi
nd
o
u
t that the more n
ode
s, the more
routin
g
overhe
ad. As there a
r
e m
o
re n
ode
s, b
o
th proto
c
ol
s must mai
n
ta
in more ro
uting informatio
n in
ca
che
and
great amo
unt o
f
control me
ssag
e shoul
d
be forwa
r
de
d
.
However, th
e thre
e line
s
are
very close, sh
owin
g the AECBRP only in
cre
a
ses a little norm
a
lization routin
g overhe
ad.
Figure 9. Normalizatio
n Ro
uting Overh
e
ad vs. Numb
er of node
s
5.3. In
v
estigation of p
a
c
ket deliv
er
y
ratio in net
w
ork conditio
n
In the eig
h
th
scena
rio, th
e pa
cket deli
v
ery
ratio i
s
comp
ared to
the ch
ang
e o
f
spee
d.
Packet delive
r
y ratio is defi
ned a
s
the total numbe
r
of data pa
ckets
sent by traffic sou
r
ces to th
e
total numbe
r
of data pa
ckets received
at destin
a
tion
s. Figu
re 1
0
indicates th
at increa
sing t
he
spe
ed in all three p
r
oto
c
ol
s, the packet
delivery ratio
decrea
s
e
s
.
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 : 711 – 72
1
720
Figure 10. PDR vs. Spee
d of Node
s in
the Netwo
r
ks
In the ni
nth
scen
ario
,
th
e p
a
cket d
e
livery ratio i
s
e
s
ti
mated ve
rsus the n
u
mbe
r
of nod
es
cha
nge. Fi
gu
re 1
1
d
e
mon
s
trate
s
th
at i
n
crea
sing
the
numb
e
r of n
ode
s
will de
crea
se th
e p
a
c
ket
delivery. The reason is because some paths w
ill
becom
e
longer if ther
e are more nodes in
netwo
rk,
more an
d mo
re
p
a
ckets are d
r
oppe
d in th
e
pro
c
e
s
s of tra
n
smi
ssi
on, th
en the t
w
o
da
ta
delivery rate
s decrea
s
e.
Figure 11. PDR vs. Numb
er of
Nodes
in the Network
s
6. Conclu
sion
The cl
uster-based routi
ng
protocol
s impact the net
work
sc
alability. In CBRP the
cluster
formation
alg
o
rithm, the lo
we
st ID alg
o
r
ithm
do
es
n
o
t con
s
id
er
mobility and
node
s e
nerg
y
in
MANETs
. In t
h
is
paper, the c
l
us
ter formation algori
thm that us
es
the relati
ve mobility metric
, the
resi
dual e
nergy and con
n
e
ctivity degre
e
is introdu
ced. After forming the clu
s
ter, to prev
ent
sud
den
de
cre
m
ent of cl
ust
e
r h
ead
ene
rgy, an efficie
n
t clu
s
ter
mai
n
tenan
ce
alg
o
rithm b
a
sed
on
the ag
gre
gat
e en
ergy
met
r
ic of
clu
s
ter
head
is
p
r
opo
sed. Thi
s
al
g
o
rithm create
s
stable
cl
ust
e
rs.
Comp
ared to
the origin
al
CBRP an
d
Cro
s
s-CB
RP, the rate of clu
s
ter-he
ad ch
ang
es
has
signifi
cant im
provem
ent that cau
s
e
s
bet
te
r throu
ghp
u
t
and lifetime of the network.
In the future, we
plan
to
impleme
n
t a
nd eval
uate
the AECBRP
perfo
rma
n
ce
usi
n
g
differen
c
e mo
bility models
and furthe
rm
ore to implem
ent the AECBRP in a test-b
ed MANET
s
.
Evaluation Warning : The document was created with Spire.PDF for Python.