TELKOM
NIKA
, Vol.14, No
.1, March 2
0
1
6
, pp. 319~3
2
5
ISSN: 1693-6
930,
accredited
A
by DIKTI, De
cree No: 58/DIK
T
I/Kep/2013
DOI
:
10.12928/TELKOMNIKA.v14i1.2743
319
Re
cei
v
ed Se
ptem
ber 26, 2015; Revi
se
d De
ce
m
ber
29, 2015; Accepted Janu
ary 18, 201
6
A Contention-Based Routing Protocol for VANET
Deling Hu
an
g*
1,2
, Yusong Yan
1
1
Colle
ge of Info
rmation Sci
enc
e and T
e
chno
l
o
g
y
, South
w
e
s
t Jiaotong U
n
iv
ersit
y
,
Che
ngd
u 61
00
31, Sichu
an, C
h
in
a
2
Colle
ge of Sof
t
w
ar
e Eng
i
n
eer
ing, Ch
on
qin
g
Univers
i
t
y
of Posts and T
e
lecommunic
a
tio
n
,
Cho
ngq
in
g 40
0
065, Ch
in
a
*Corres
p
o
ndi
n
g
author, em
ail
:
huang
dl@c
qu
pt.edu.cn
A
b
st
r
a
ct
In VANET
s, vehicl
es
as n
o
d
e
s are
se
lf-org
ani
z
e
d
an
d i
n
ter-co
m
mun
i
cat
ed w
i
tho
u
t ce
ntrali
z
e
d
author
ity. T
h
e
topol
ogy for
m
ed by
ve
h
i
cles
chan
ges
qu
ic
kly, w
h
ich
ma
kes routi
ng
be
come i
n
stabi
lit
y.
Po
si
ti
on
-b
a
s
e
d
ro
u
t
in
g
,
com
p
a
r
e
d
wi
th
tra
d
i
t
i
o
na
l
ro
u
t
in
g
,
is mo
re sca
la
ble
an
d fe
a
s
ib
le
. Th
u
s
i
t
ha
s
b
een
prove
n
stab
ler
for VANET
s t
han
conv
enti
o
nal r
outi
ng. H
o
w
e
ver, the fr
equ
ently
c
h
a
n
ged
topo
lo
gy
and
nod
es d
ens
ity
coul
d br
eak t
h
e p
a
th a
p
a
ck
et is fol
l
ow
in
g. T
hus
desi
gni
n
g
a r
o
b
u
st mul
t
i-hop
routi
ng i
n
VANET
is c
hal
len
g
in
g. T
h
is
p
aper
pro
pos
es
an
e
nha
nce
d
positi
on-b
a
se
d
rout
i
ng protoc
ol
c
a
ll
ed CBG
R
,
which takes into account th
e
velocity a
nd
directi
on of ve
hicles
in VAN
ET
. Simul
a
tion
results show
tha
t
CBGR achiev
e
s
a high lev
e
l
of routing p
e
rformanc
e
in terms of hop co
u
n
ts, netw
o
rk latency and p
a
c
k
et
deliv
ery ratio b
o
th in de
nse or
sparse veh
i
cul
a
r ad-h
o
c netw
o
rks.
Ke
y
w
ords
: VANET
, routing pr
otocol, gre
edy
forw
arding, loc
a
l opti
m
al
Copy
right
©
2016 Un
ive
r
sita
s Ah
mad
Dah
l
an
. All rig
h
t
s r
ese
rved
.
1. Introduc
tion
Vehicul
a
r Ad
hoc
Network (VANET
s)
[1] is a mea
n
s of inte
r-v
ehicl
e co
mm
unication,
usin
g for safe
driving, drive
r
assista
n
ce and
traffic ma
nagem
ent. In VANETs, vehicle
s
a
s
nod
e
s
are self-o
rga
n
ize
d
, they communi
catio
n
in a di
strib
u
ted fashio
n with no ce
ntralize
d
autho
ri
ty.
More
over, the vehicle
s
are move fast, the t
opology formed by them ch
ang
e
s
qui
ckly. Hi
ghly
dynamic topo
logy ma
ke
s
data tra
n
smi
ssi
on le
ss rel
i
able
,
sin
c
e
the highly dy
namic topolo
g
y
may re
sult in
netwo
rk di
sconne
ct fre
q
u
ently. So
the
key p
r
obl
em
of desi
gn
a routing p
r
oto
c
ol is
kee
p
ing it a
s
stable
as
po
ssi
ble, trying
to ke
e
p
the
link b
e
twe
e
n
two vehi
cle
s
while th
ey a
r
e
transmitting i
n
formatio
n.
Beside,
nod
e
s
in
VANET
s are n
o
t subj
ected
to
storage
and
po
wer
limitation as i
n
wirele
ss
se
nso
r
net
works. Vehi
cl
es
a
s
no
de in VA
NET, are
su
p
posed to hav
e
ample comp
uting po
wer and ene
rg
y. Howeve
r, efficient multi-hop
routi
ng in VANE
T
is
chall
engin
g
, due to the freque
ntly cha
nged to
polog
y. VANET routing protoco
l
s mu
st ope
rate
reliably in sce
nario
s em
bra
c
ing hi
gh spe
ed nod
es.
Significant V
A
NET protocol
s have
bee
n pro
p
o
s
ed,
and they
’re classified a
s
topolo
g
y-
based
and
p
o
sition
-ba
s
e
d
[
2]. Topolo
g
y-ba
sed
rout
in
g
p
r
oto
c
ol
s perfo
rm pa
cket
forwarding
,
usin
g links’ i
n
formatio
n th
ey colle
ct fro
m
the
network. They find
out a path to
the de
stinati
on
before
rel
a
y the pa
cket. Position
-ba
s
e
d
(ge
ographi
c)
routin
gs
perf
o
rm p
a
cket f
o
rwarding, u
s
ing
the po
sition
i
n
formatio
n of
the
ultimate
destin
a
tion
a
nd of
the
one
-hop
n
e
ighb
o
r
s.
Usually th
ey
are
statele
s
s, with no n
eed to
keep
the lin
k inf
o
rmatio
n ab
out the wh
o
l
e network. An
interme
d
iate
node forwa
r
d
s
a pa
cket to its neighb
or
with the clo
s
est dista
n
ce to the positio
n
of
the destinatio
n. These day
s, it is comm
on that
vehicl
es get a GPS unit on board, from wh
ere
vehicle
s
can
get their l
o
catio
n
informati
on. It has m
ade
p
o
sition
-ba
s
e
d
routin
g ea
sy
impleme
n
ted
and
po
pula
r
.Literatu
r
e
s
have
con
d
u
c
ted evalu
a
tio
n
stu
d
y of t
opolo
g
y-ba
se
d
routing p
r
oto
c
ol
s and po
si
tion-ba
se
d o
nes[3, 4]. Un
able to cop
e
up with nod
es’ high mo
b
ility
and mi
nimu
m tran
smi
ssi
on del
ay, topology-ba
s
ed
routin
g p
r
ot
ocol
s p
e
rfo
r
m una
ccepta
b
ly in
VANETs
[5, 6].
Position
-ba
s
e
d
routing, a
s
is used by pro
t
ocol
s like G
r
eedy Perimet
e
r Stateless
Routin
g
(GPSR) [7] and Ge
og
ra
phic
Routin
g
Protocol
(G
RP) [8], is well suited fo
r highly dyna
mic
environ
ment
s such as VANET.
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 16
93-6
930
TELKOM
NIKA
Vol. 14, No. 1, March 2
016 : 319 – 3
2
5
320
Intermedi
ate
node i
n
GP
SR forwa
r
d
s
a pa
cket to a
radi
o nei
ghb
or
whi
c
h i
s
cl
ose
s
t to
the de
sinatio
n. This ap
pro
c
a
c
h i
s
calle
d
gre
edy forwardin
g
. It is a
s
sumee
d
that
ea
ch n
ode
can
determi
ne its own po
sitio
n
usin
g a GPS. Nodes e
x
chan
ge thei
r locatio
n
s
wi
th neihbo
rs,
and
obtain the
po
sition of d
e
sti
nation by a l
o
cation
se
rvice. In som
e
case
s, the
r
e
might be
no
next-
hop
whe
n
greedly forwa
r
ding, GPSR
introdu
ce
s a
strate
gy, ca
lled pe
rimete
r ro
uting. Ma
ny
resea
r
ch [9-1
2] have improved GPSR, most of
whi
c
h worke
d
on
plana
rized g
r
aph
s, whi
c
h
are
requi
re
d in
p
e
rimete
r m
o
d
e
. Ho
weve
r t
hese
schem
e
s
al
way
s
invo
lve sig
n
ifica
n
t overhead
s,
a
n
d
thus re
du
ce the efficien
cy
of position
-
ba
sed routing.
Nod
e
in GRP
maintain
s a neigh
bor tabl
e by perio
dically broad
ca
sting Hello m
e
ssage. It
is a
s
sume
d t
hat ea
ch
no
d
e
can
determ
i
ne its o
w
n
p
o
sition
u
s
ing
a GPS. Th
e
positio
ns of o
t
her
node
s
are
de
termine
d
thro
ugh flo
oding
[13]. The
wh
ole n
e
two
r
k i
s
divid
ed i
n
to
qua
dra
n
ts
with
different level
s
. They
are
d
eployed i
n
a
hiera
r
chy ma
nner.
Fou
r
lo
w level
s
com
pose a
high l
e
vel,
thus provide a distrib
u
ted locati
o
n
se
rvice. This sched
ule makes it overcomin
g
GPSR and m
any
other p
o
sitio
n
based p
r
oto
c
ol. G
R
P forwards
pa
ck
ets in a
n
g
r
eed
y mode until i
t
gets stu
c
k i
n
a
local optim
u
m
, where there are no ne
w node
s to
forward the packet to. It use
s
a backtra
cki
ng
mech
ani
sm,
whe
r
e the
pa
cket is
return
ed to the p
r
e
v
ious h
op a
n
d
a ne
w n
e
xt hop
sele
ction
can
be made.
The afo
r
em
e
n
tioned
wo
rki
ng
sch
edul
e
of GRP m
a
kes it a
suitab
le ro
uting p
r
o
t
ocol fo
r
VANET, but t
here are
still
some short
c
omings.
T
he
packet forwarding algorithm
is too si
mpl
e
,
that it only m
a
kes
use of
the location information to f
o
rward
the packets. T
he
m
obility of packet’s
sou
r
ce an
d
destin
a
tion a
r
e
con
s
ide
r
e
d
, while th
e
interme
d
iate
node’
s mov
e
ment isn’t
well
con
s
id
ere
d
when
sel
e
ctin
g
next h
op. A
s
the i
n
te
rm
ed
iate no
de
mo
ving, the li
nk
betwe
en th
em
will probably vanish,
whi
c
h result
in the
routing instabil
i
ty. This
generates a routing rebuilt, whi
c
h
bring
mo
re
n
e
twork
overh
ead a
nd
co
st
. Thus G
R
P
may not p
e
rf
orm
as
effici
ent a
s
expe
cted
whe
n
deploy
ed in VANET, where vehicl
es are always in a relatively high mobility.
In the next section, a
ge
ogra
phi
c rout
ing
p
r
oto
c
ol
calle
d CB
RG
is p
r
op
osed
, which
improve the forwarding de
cisi
on
s throu
gh cont
e
n
tion
among nod
e
s
. CBGR i
s
propo
se
d ba
sed
on G
R
P to
o
p
timize th
e e
fficiency, it rede
sign
s th
e
forwarding
a
nd b
a
cktra
c
ki
ng al
gorith
m
of
GRP, Th
e pu
rpo
s
e
of CB
GR i
s
to im
prove the
p
a
cket delivery
ra
tios
while
red
u
ce
the n
e
twork
delay. CBG
R
keep
s the a
d
vantage
s
of
GRP.Furth
e
r more, it inco
rpo
r
ate
s
geo
grap
hic l
o
cati
on
with no
de’
s
mobile velo
ci
ty and dire
cti
on, to
be th
e pre
c
o
nditio
n
for
cho
o
sin
g
the next h
op.
Simulation
re
sults sh
ow th
at CBG
R
a
c
hi
eves a
hig
h
l
e
vel of ro
utin
g pe
rform
a
n
c
e in terms of
hop
cou
n
ts, net
work l
a
ten
c
y a
nd pa
cket de
livery ratio b
o
th in den
se
or sparse v
ehicular
ad-h
o
c
netwo
rks.
2.
Contentio
n
-Based Ge
ographic Ro
uting
(CBGR)
2.1
.
Con
t
en
tion-bas
e
d F
o
r
w
a
r
ding
We
notice th
at, Locations
Service
s
a
n
d
Forw
a
r
din
g
algorith
m
a
r
e
the two
key
probl
em
s
in routing
pro
t
ocol
s ba
sed
on geo
gra
phi
c location,
wh
ich would aff
e
ct the ro
utin
g perfo
rma
n
ces.
The hi
erarch
y locatio
n
se
rvice
empl
oyed in
G
R
P o
v
erco
me
s m
any othe
r
ro
uting p
r
oto
c
o
l
s in
usin
g the d
e
s
tination
zo
n
e
inste
ad of
destin
a
tion l
o
catio
n
. Thi
s
makes
GRP
be on
e of t
h
e
highe
st effici
ent po
sition-b
a
se
d ro
uting
proto
c
ol
s, as
it can d
e
crea
se the l
o
cation indete
r
mina
cy
cau
s
e
by th
e mobility of
the de
stinat
ion. CB
G
R
keep
s this ad
vantage, an
d
red
e
si
gns the
forwa
r
di
ng al
gorithm, imp
o
sin
g
an ad
ditional co
nte
n
tion in spe
ed and direction. Meanwhile,
CBGR
p
r
op
o
s
e a rep
a
ir strategy
when
a
pa
cket get
s
stuck i
n
a
local optimum,
while GRP
si
mply
gives the pa
cket ba
ck to th
e previou
s
forwarde
r.
2.1.1. Conte
n
tion Ba
sed
on Speed
Priority i
s
int
r
odu
ce
d in
CBGR. A
nod
e calcul
ate
s
the p
r
iority o
f
its n
e
ighb
or befo
r
e
forwa
r
d it. We use th
e pri
o
rity to sele
ct the mo
st suita
b
le nod
e to relay, rather t
han ma
ke a
p
u
re
gree
dy forwarding sele
ction
.
Stability is an impo
rtant requireme
nt of rout
ing in
a
wirel
e
ss n
e
twork. A hig
h
mobility
may cause the route fail
ure if we use pure greedy
forwardi
ng. To i
n
crease
reliability, GRP reads
the velocity of each nod
e, and sel
e
ct reliable
nod
es
with the lowe
r velocity, abando
ning oth
e
rs
even if they
are cl
oser to the
destinatio
n. On the other han
d, nod
es with lo
wer velocity wou
l
d
make
a
sma
ller lo
cation
cha
nge
s
whil
e exch
angi
n
g
Hell
o me
ssag
e, this
will prolo
ng th
e
neigh
bor tabl
e’ lifetime.
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
1693-6
930
A Contention
-
Based
Routin
g Protocol for VANET (Deli
ng Hu
ang
)
321
There a
r
e two app
roa
c
h
e
s to get the n
e
i
ghbo
r’s vel
o
ci
ty. One can
write it
s o
w
n
spe
ed in
hello me
ssag
e, thereby th
e spe
ed is
known by
others
along
with the bro
a
d
c
asting of hell
o
messag
e. Another way is that, each node
can
cal
c
ulate the ve
locity of its neigh
bor by
the
following formula:
(
node
_lat
_
lat
node
_long
_long)
(1)
Whe
r
e no
de_
lat, node_lon
g and timest
amp
j
are the
latitude, longi
tude and tim
e
of the
neigh
bor at time j respe
c
tively, and nbr_lat, nbr_lo
n
g
and timesta
m
p
I
are the l
a
titude, longit
ude
and time of the neig
hbo
r
at time i read
from
neig
h
b
o
r table.
We
add an
additi
onal field n
a
m
ed
velocity to the neigh
bor ta
ble. To minim
i
ze
the
comp
uting co
st wh
en rel
a
y a pa
cket, each
no
de
doe
s this calculation when
receiving the
hello
me
ssag
e, and upd
ate
s
the neig
hbo
r table.
Whe
n
forwa
r
ding p
a
cket
s, we
use P
v
to de
cide
the
prio
rity of the
neigh
bo
rs. F
o
rmul
a 2
defi
ne
the veloc
i
ty priority:
1
(2)
Thus CB
GR
sele
cts the n
e
xt hop b
a
se
d on
a
conte
n
tion of
spe
e
d
. The
one
of the
lowe
st sp
eed
will get a high
est velocity priority.
2.1.2. Conte
n
tion Ba
se o
n
Direc
t
ion
Becau
s
e
of the ob
sta
c
les,
node
s in diff
erent
road
s cannot commu
nicate with
e
a
ch other
unle
ss
one
is an
exten
s
i
on of the
other
.
Con
s
id
e
r
sce
nari
o
in
Figure 1,
Node
N1
i
s
at
the
cro
s
sing. At t
hat mom
ent,
the current fo
rwa
r
d
e
r
C
wil
l
pic
k
N
1
a
s
t
he n
e
xt hop
with a
differe
nt
moving direct
ion. Ho
weve
r, since th
e n
ode is
alway
s
moving fa
st in VANET,
N
1
may be o
u
t of
C’
s radio
ra
n
ge. In fa
ct it
woul
d b
e
b
e
tter if
C
ch
oo
s
e
s
N
2
at
that
time. So it’s
necessa
ry for the
neigh
bou
r to
content b
a
se on the mo
ving dire
ctio
n
.
For a long
-lasting lin
k, the one
with
the
same di
re
ctio
n as the curre
n
t forwa
r
de
r will be a go
od
candi
date.
Figure 1. Jun
c
tion no
de
s moving towa
rds differe
nt directio
n
As discusse
d before, we can u
s
e the l
o
catio
n
at two different m
o
ments to cal
c
ulate the
dire
ction of a node.
node_lat
node_lat
node_long
node_long
(3)
Whe
r
e
no
d
e_la
t
and
n
ode_
long
a
r
e the
coordi
nate
s
, a
n
d
t1
an
d
t2
are two different time
points. In ord
e
r to red
u
ce the impact o
n
relay pa
cket, node cal
c
ulates
Dir
wh
en re
ceiving
the
Hello me
ssag
e, and then u
pdate
s
its nei
ghbo
r table. Let
P
d
stand for the prio
rity base
d
on no
de
moving direct
ion, we defin
e P
d
as the following fo
rmul
a:
1
|
|
(4)
N
o
de
r
e
ad
P
d
from neigh
bo
r table to deci
de a better n
e
xt-hop befo
r
e forwa
r
di
ng.
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 16
93-6
930
TELKOM
NIKA
Vol. 14, No. 1, March 2
016 : 319 – 3
2
5
322
2.2
.
De
al
w
i
t
h
Dead
-en
d
Problem
Sometimes
when g
r
ee
dy forwarding p
a
c
kets, it
ca
n lead to blo
c
ke
d route
s
whe
r
e the
r
e
are n
o
next h
ops to forwa
r
d the pa
cket to. GRP
ado
pts ba
ckt
ra
cki
n
g
, which retu
rn the pa
cket to
the p
r
eviou
s
nod
e
whe
r
e
a n
e
w next
hop
sele
ctio
n can
be
m
ade [8]. T
h
is me
chani
sm
will
prob
ably increase the hop
s to the desti
nation,
theref
ore in
cre
a
se the transmission delay. Th
e
long-kn
own d
ead-end p
r
o
b
l
em is
depi
cte
d
in Figure 2.
Figure 2
.
No
de
x
’s
voi
d
with res
p
ec
t to
D.
The di
stan
ce
betwe
en D a
nd x is the ra
dius
of the a
r
c aroun
d nod
e D. The a
r
c
arou
nd x
stand
s fo
r its
radio
ra
nge.
Although th
ere are two exi
s
ting
path
s
to
D,
x-y-z-D
a
nd x-w-v-D, x
will
cho
o
se n
e
ith
e
r
of them
u
s
ing g
r
e
edy fo
rwa
r
di
ng
[7]. There need
s
to be some o
t
her mechani
sm
for x to forward pack
e
t
s
in this
s
i
tuation.
Cla
ssi
cal pro
t
ocol
GPS
R
use
s
pe
r
i
me
te
r
fo
rw
ar
d
i
ng
to solve the local
optimal. When
there i
s
no
n
e
w
nod
e to
relay in
gre
e
d
y
mode, th
e
l
ong-kn
own
right ha
nd
rule
is i
n
trod
uced
to
find a path around the
void
. Perimeter forwa
r
di
ng will
come b
a
ck to
greedy forwardin
g
as
soo
n
as a
ne
w nod
e
ap
pea
rs, which
is geo
graphi
cally
cl
oser to th
e p
a
cket’s d
e
stin
ation than th
e n
o
d
e
swit
ch to
pe
rimeter mo
d
e
. The
short
c
omin
g of
p
e
rimete
r fo
rwarding
is p
l
anar g
r
ap
hs are
requi
re
d. An
algorith
m
fo
r rem
o
ving
e
dge
s from th
e g
r
ap
h that
are n
o
t p
a
rt
of the
Relativ
e
Neig
hbo
rho
o
d
G
r
ap
h
or
Gabri
e
l Gra
p
h
a
r
e
also
re
quire
d to
yie
l
d a
network
with n
o
crossing
links [14, 15]. This will defi
n
it
ely increase the routing
cost.
GRP, as men
t
ioned befo
r
e
,
use a backt
rackin
g mech
anism to retu
rn the pa
cke
d
to the
previou
s
no
d
e
, where a n
e
w next hop
sele
ction
u
s
in
g gree
dy mo
de ca
n be m
ade. Thu
s
no
new
mech
ani
sm i
n
trodu
ce
d to solve the p
r
o
b
lem. A node
will eliminate
local o
p
timal
neighb
ors o
n
e
by one until
a valid neig
h
bor i
s
foun
d
or the p
a
ck
et
-life co
me
s to the en
d. This trial
and
error
method will p
r
oba
bly incre
a
se the h
op counts a
nd the
network late
ncy.
Figure 3
.
C
i
s
the local opti
m
al node of s with destin
a
tion
D
The intuitio
n
behin
d
CBG
R
i
s
eve
n
if
we can’t
g
e
t an
y clo
s
er to th
e de
stinatio
n, we
won’t
go far
away from it. So we go b
a
ck, b
u
t neithe
r
ret
u
rn
s that mu
ch to th
e pre
v
ious h
op a
s
in
GRP, no
r fol
l
ow a
fixed
cou
n
ter
wi
se
seq
uen
ce
a
r
oun
d the
vo
id
as in GP
SR. We let the
neigh
bors of the local opti
m
al node co
ntent again.
The one cl
osest to the destination got
the
highe
st p
r
io
r. If
C
get m
o
re
than
on
e nei
ghb
ors with th
e
sa
me
clo
s
est
distan
ce
to t
he
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
1693-6
930
A Contention
-
Based
Routin
g Protocol for VANET (Deli
ng Hu
ang
)
323
destin
a
tion, t
he o
ne
whi
c
h
is furthe
st to
C
will
win the contention.
T
h
is i
s
because there i
s
a
void
ahea
d; the further the n
o
d
e
is away fro
m
C
, the easi
e
r it can byp
a
ss the
voi
d
. See Figure 3
for
an exam
ple.
N
1
and
N
2
lies on th
e a
r
c
wi
th the
sam
e
distan
ce
to n
ode
D.
Note t
hat in th
e figu
re,
N
2
is mo
re sui
t
able than
N
1
to be the next hop.
2.3
.
Combini
ng All Conte
n
tions
We
no
w p
r
e
s
ent th
e full
CBG
R
routi
ng al
gorith
m
, whi
c
h
com
b
ine
s
all
the
above
mentione
d co
ntention
s
tog
e
ther.
Step 1.
Whe
n
a so
urce n
ode o
r
an int
e
rme
d
iate
no
de send
s pa
ckets, it ch
ecks i
n
it
s
neigh
bor tabl
e, to find o
u
t
the entry
of the pa
cket’s u
l
timate de
stin
ation. If there
is a
n
e
n
try, the
packet
will be delivered
directly to the
destinati
on
successfully, and CB
GR
ends. If there i
s
n’t
any entry, th
e sende
r
will
sel
e
ct a
no
d
e
in it
s si
ngl
e-ho
p radio
neigh
bor,
whi
c
h i
s
g
eog
ra
phic
clo
s
e
s
t to the destinatio
n. This
step loo
p
s an
d quits
whe
n
the followin
g
two ca
se sho
w
up:
Ca
se 1: if there a
r
e mo
re
than one no
des, wh
ich form a set
call
ed set_
1, sat
i
sfy the
rule, goe
s ste
p
2;
Ca
se 2: if there are no no
d
e
s
satisfy the rule, goe
s ste
p
4;
Step 2.
Cont
end
b
a
sed
o
n
velocitie
s
. The send
er cal
c
ulate
P
v
, usin
g form
ul
a 2, for
each
can
d
ida
t
es in
set_1,
cho
o
ses the
one
with hi
gh
est P
v
as the
next ho
p. Th
e alg
o
rithm
g
oes
to step 1. If
there a
r
e two n
ode
s and ab
o
v
e, which form a set calle
d set_2, sati
sfy the rule, goes
step 3;
Step 3.
Co
ntend ba
se
d on
dire
ction. Th
e sen
d
e
r
cal
c
ulates P
d,
u
s
i
ng formul
a 4,
for eac
h
can
d
idate in
set_2, choo
ses the on
e with highe
st
P
d
as the next hop, and the
algorithm g
o
e
s
back to
step
1; If there
a
r
e
still mo
re
than on
e
n
o
des sati
sfy the rule, the
sen
der pi
cks one
rand
omly, an
d the algorith
m
goes b
a
ck
to step 1.
Step 4.
B
r
ea
k throug
h d
e
a
d
-en
d
. Th
e sende
r relays the pa
ck, u
s
in
g sch
edule
prese
n
te
d
in se
ction
3.3
.
If there is
n
o
neig
hbo
r e
x
cept
the
on
e wh
ere the
packet i
s
fro
m
, the se
nde
r will
carry the pa
cked u
n
til there is a fre
s
h
hello me
ssag
e or expi
re T
T
L, then the
algorith
m
go
es
bac
k t
o
st
ep
1
.
By competitions, no
de can pick up
a be
tter next-hop tha
n
G
R
P in vehicl
e ad hoc
netwo
rk.
We
must
cla
r
ify there
i
s
rarely po
ssi
b
ility for ou
r al
gorith
m
de
gene
rat
ed to
pu
re
greedy
without co
ntention. We
define
th
e t
w
o
nod
es wi
th the
sam
e
dista
n
ce fro
m
send
er, if
the
differen
c
e i
s
within
10 m
e
ters. So
there will
always be
co
ntentio
ns, e
s
p
e
ci
all
y
in cro
s
sro
a
d
,
whe
r
e the di
rection
conte
n
t
ion is highly required.
Re
call that
al
l nod
es
main
tain a n
e
igh
b
o
r tabl
e, whi
c
h
stores the
location
s, velocitie
s
and directio
n
s
of their si
ng
le-ho
p
ra
dio n
e
ighb
ors. Thi
s
table p
r
ovid
es all informa
t
ion requi
re
d for
CBGR’
s
fo
rwardin
g
d
e
ci
si
ons, b
e
yond
the pa
cket
s t
hemselves. F
r
om thi
s
p
o
in
t of view, CB
GR
make a n
egli
g
ible spa
c
e cost to improv
e the routing
perfo
rman
ce
of GRP.
3. Simulation Resul
t
In this section we will com
pare the performance
of CBGR
with GRP. The ex
perim
ents
were
conducted on OP
Net model
er. M
edium
ac
cess cont
rol (M
AC)
i
s
IEEE 802.11g,
with a
radio rang
e of 250m. The mobility trac
e
s
we
re gen
erated on an area of 3000m
1000m u
s
ing
VANetMobiSi
m [16], a
vehicul
a
r t
r
affic gener
ator. The mi
cro-mobility is controll
ed
by
IDM_IM[17].
Table 1. Simulation Para
meters
Parameter Value
Net
w
ork simulator
OPNet 14
.5
Mobility
simulator
VANetMobisim
Dimension 3000m×1000m
Simulation time
600s
Routing Protocol
GRP, CB
GR
Average vehicle speed
50 km/hr
Transmission pow
e
r
0.005
w
802.11g r
a
te
11Mbps
Source-destination
Pairs
Random
Packet interval
0.2s
Packet size
500B
y
t
e
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 16
93-6
930
TELKOM
NIKA
Vol. 14, No. 1, March 2
016 : 319 – 3
2
5
324
For ea
ch
sim
u
lation run, 2
0
% sen
der-recei
ve
r pai
rs from the net
work
were ra
ndomly
sele
cted. Ea
ch p
a
ir ex
ch
ange 2
0
pa
ckets
over
5
seco
nd
s. We
measured th
e pa
cket d
e
li
very
rate (PDR) versus the n
u
m
bers of veh
i
cle
s
parti
ci
p
a
te in (Figu
r
e
4). Each poi
nt in the gra
phs
come
s
out from 10
inde
p
ende
nt ru
ns.
PDR
sh
ow
s
good
re
sult
s
for ou
r a
p
p
r
o
a
ch
co
mpa
r
e
d
to
GRP. This is becau
se
CBGR select
s the best suit
a
b
le node, u
s
i
ng the mobili
ty information, to
relay the
pa
cket to.
We
ca
n see i
n
the
very le
ft
pa
rt,
wh
en network
i
s
relatively
sp
arse, CBRG
has ove
r
whel
ming adva
n
ta
ge over G
R
P
in PDR, t
han
ks fo
r allo
win
g
ca
rrying
pa
cket whe
n
th
ere
is no n
ode to
be delivered
. The figure il
lustrate
s
CBGR i
s
stabl
e with a hig
h
PDR a
bove 9
0
%
,
neverthel
ess
the netwo
rk i
s
den
se o
r
sp
arse.
Figure 4. PDR vs. Nod
e
d
ensity
Fi
gure 5. Hop
count vs. No
de den
sity
Figure 5
sho
w
s th
e ave
r
a
ge nu
mbe
r
of
hop
s for
a p
a
cket du
ring
routing
to its
ultimate
destin
a
tion
compa
r
e
with
GRP. The
re
ductio
n
in h
o
p
co
unt for
CBGR i
s
du
e to not ba
cktra
cki
ng
whe
n
come
t
o
the
dea
d-e
nd. CB
GR ke
eps the
pr
ogress the
pa
cket ha
s m
ade
by last
hop,
while
CBGR obliterate
it.
Figure 6 de
picts the
en
d-to-end d
e
l
a
y. The sho
r
ter laten
c
y
for GRP e
a
rly in the
simulation is the result of small PDR. Packets
that
do
not get d
e
livered to th
e de
stination
do n
o
t
contri
bute to
the latency.
Thus
CBG
R
gets l
a
rg
e
r
PDR.
Ho
w
e
v
e
r it is b
e
tter to re
ce
ive
somethi
ng th
an nothi
ng.
Whe
n
no
de
cou
n
t rea
c
h
e
s
mo
re tha
n
100, CB
GR
gets le
ss lat
ency
becau
se of th
e more
re
aso
nable
choi
ce
of the interm
ediate no
de
s
unde
r compa
r
able P
DR
wi
th
GRP. Recall
that althoug
h the choi
ce
is ba
se
d on
som
e
extra
com
puting,
we a
r
rang
e the
cal
c
ulatin
g at the time receiving a Hello
mess
age rather than the time relaying t
he packet. Th
us
the comp
utation co
st ra
rely
contrib
u
tes t
o
the latency.
Figure 6. Latency vs. No
d
e
den
sity
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
1693-6
930
A Contention
-
Based
Routin
g Protocol for VANET (Deli
ng Hu
ang
)
325
4. Conclusio
n
This
work ai
ms at i
m
proving the
rout
e st
ability in vehicular ad-hoc
networks.
The
simulatio
n
re
sults
have
proved the
pro
posed al
gor
it
hm ad
apts it
self to VANET
in varying
n
ode
den
sities. In f
u
ture,
we pl
a
n
to ma
ke th
e forwarding
deci
s
io
n with
an ove
r
lap
o
f
city map. T
hat
will make our algorithm perf
o
rm
better in variou
s scenarios.
Ackn
o
w
l
e
dg
ements
This
wo
rk wa
s
sup
porte
d
by Natio
nal
Natural S
c
ien
c
e F
oun
datio
n of
China
(Grant
No.
6130
9032
)Progra
m
fo
r I
nnovation
T
eam Buil
din
g
at In
stitutions of
High
er Ed
ucation
in
Cho
ngqin
g
(Grant
No. KJT
D
201
310
).
Scientifi
c
a
nd T
e
chnolo
g
ical
Research P
r
og
ram
of
Cho
ngqin
g
M
unici
pal Edu
c
ation Com
m
ission (K
J1
400
431)
Referen
ces
[1]
F
an Li, Yu W
ang. Ro
uting i
n
Vehic
u
lar A
d
Hoc Net
w
or
ks: A Surve
y
.
IEEE Vehicul
a
r Technol
ogy
Maga
z
i
ne
. 2
0
0
7
; 2(2): 12-22.
[2]
K Katsaros, M
Dian
a
ti, Z
S
un.
Surve
y
of
Rou
t
ing Pr
otoco
l
s i
n
Ve
hicu
lar
Ad
Hoc
Net
w
o
r
ks
. Advanc
es
in Veh
i
cul
a
r Ad
-Hoc Netw
orks: Develo
p
m
ent
s and Ch
all
e
n
g
e
s
. 2010: 1
49-
170
.
[3]
Loch
e
rt C, H
a
r
t
enstein
H, T
i
an J,
F
u
ssl
er H,
Herma
nn,
D,
Mauve, M.
A r
outin
g strate
gy
for ve
hicu
lar
ad hoc n
e
tw
orks in city envir
on
me
nts. 2003
.
Proceedi
ngs
of IEEE Intelligent Veh
i
cl
es S
y
mp
osi
u
m
.
200
3:
156-
161.
[4]
Naum
ov V, B
a
uman
n R, Gro
ss T
.
An eval
u
a
tion
of Inter-V
ehicl
e A
d
Hoc
Netw
orks Base
d o
n
R
e
a
listi
c
Vehic
u
lar T
r
ac
es.
Proc. ACM MobiH
o
c’0
6
C
onf. 200
6: 108-
119.
[5]
Sanj
a
y
K, Dhur
and
her, Moh
a
mmad S Obai
d
a
t.
GROOV: A
Geogra
phic
Ro
uting Over VA
NET
s
and Its
Performanc
e E
v
alu
a
tion.
Gl
ob
ecom 2
0
1
2
-Co
mmunicati
ons
QoS, Reli
ab
ilit
y
and
Mod
e
l
i
ng
S
y
m
posi
u
m.
201
2: 167
0-16
75
[6]
R Hussey
,
E Huff,
Z
Shinw
a
ri, V Hnat
yshin.
A comp
arative study
of proactive a
nd reactiv
e
geo
grap
hic
a
l r
outin
g protoc
ol
s for MANET
.
Procee
din
g
s of
12th Int. Conf.
on W
i
rel
e
ss N
e
t
w
.
20
13
: 1
-
6.
[7]
B Karp, HT
Kung.
GPSR: gr
e
edy p
e
ri
meter
stateless r
outin
g for w
i
rel
e
ss
netw
o
rks.
Proc
. ACM/IEEE
MOBICOM. 20
00: 243-
25
4.
[8]
Z
Li.
Geogr
ap
hic ro
utin
g pr
o
t
ocol
and
si
mu
latio
n
.
Proce
e
d
in
gs of
2nd
Int. W
o
rkshop
on C
o
mp
ute
r
Scienc
e an
d Engi
neer
in
g. 20
09: 404-
40
7.
[9]
Kim YJ, Govindan R, KARP B, Shenker S.
On the Pitfalls of
Geograph
ic Routin
g.
Proc. of the 3rd
Internatio
na
l W
o
rksho
p
on DI
ALM-POMC
.
2005: 34-
43.
[10]
H F
r
e
y
, I Sto
j
menov
ic.
On d
e
livery
gu
ara
n
tees of fac
e
a
n
d
co
mb
in
ed
greedy-fac
e ro
uti
ng i
n
a
d
h
o
c
and se
nsor net
w
o
rks.
Proceedin
g
s of the 12th ann
ual i
n
te
rnatio
nal co
nfe
r
ence o
n
Mobi
l
e
computi
n
g
and n
e
t
w
orki
ng
. 2006: 23-
29.
[11]
Y Kim, R Gov
i
ndan, B Kar
p
,
S Shenker.
L
a
z
y
cr
oss-li
nk re
mov
a
l f
o
r g
eog
raph
ic ro
uting.
Procee
din
g
s
of the 4th inter
natio
nal c
onfer
ence o
n
Embe
dde
d net
w
o
rke
d
sensor s
y
ste
m
s. 2006,
[12]
W
ang Yin
g
, Xu Hui-
bin, C
h
a Dai-fe
ng. A
Novel
Routi
n
g Protoco
l
for VANET
S
. TE
LKOMNIKA
Indon
esi
an Jou
r
nal of Electric
al Eng
i
ne
eri
ng.
2013; 1
1
(4): 2
195-
219
9.
[13]
OPNET
Model
er. Riverb
ed T
e
chn
o
lo
g
y
, Inc.
[14]
Agabr
iel
K, So
kal R.
A n
e
w
s
t
atisti
cal
appr
o
a
ch to
g
eogr
ap
hic v
a
riati
on
an
al
ysis
. Syste
m
atic Z
o
o
l
ogy
.
196
9; 18: 259
–
278.
[15]
T
oussaint G.
The rel
a
tive ne
i
ghb
orho
od gr
a
ph of a finitep
l
anar set.
Pattern Recog
n
iti
on.
1980; 12(
4):
261-
268.
[16]
VANetMobisim home.
http://V
ANet.eureeom.
f
r/.
[17]
Vaib
hav Go
db
ol. Intell
ige
n
t D
r
iver Mob
ilit
y M
ode
l an
d T
r
affic Pattern Gen
e
r
ation
base
d
O
p
timizati
on of
Reactiv
e
Proto
c
ols for V
e
h
i
c
u
lar A
d
h
o
c N
e
t
w
orks
.
Inter
n
a
t
iona
l Jo
urna
l
of Informatio
n
and
Netw
ork
Security.
201
3;
2(3): 207-2
15.
Evaluation Warning : The document was created with Spire.PDF for Python.