Indonesi
an
Journa
l
of El
ect
ri
cal
Engineer
ing
an
d
Comp
ut
er
Scie
nce
Vo
l.
13
,
No.
3
,
Ma
rch
201
9
, p
p.
884
~
891
IS
S
N: 25
02
-
4752, DO
I: 10
.11
591/ijeecs
.v1
3
.i
3
.pp
884
-
891
884
Journ
al h
om
e
page
:
http:
//
ia
es
core.c
om/j
ourn
als/i
ndex.
ph
p/ij
eecs
Traffic c
ongesti
on
detecti
on in a c
ity using
clusteri
ng techn
iqu
es
in VANE
Ts
An
it
a M
ohanty
1
,
S
udip
ta M
ahap
at
r
a
2
,
Ur
mi
la Bhanj
a
3
1
Depa
rtment of
El
e
ct
roni
cs
&
In
strum
ent
at
ion
E
ngine
er
ing, SIT
,
India
2
Depa
rtment of
El
e
ct
roni
cs
&
Elec
tr
ic
a
l
Com
m
unic
a
ti
on
Engi
ne
e
ring,
II
T, India
3
Depa
rtment of
El
e
ct
roni
cs
&
T
el
e
comm
unic
at
i
on
Engi
n
ee
ring
,
IGIT,
Ind
ia
Art
ic
le
In
f
o
ABSTR
A
CT
Art
ic
le
history:
Re
cei
ved
Ma
y
9
, 2
018
Re
vised
N
ov
27
,
2018
Accepte
d
Dec
7
,
20
18
Road
tra
ffi
c
con
gesti
on,
a
ser
iou
s
il
lne
ss
in
deve
l
oping
reg
ions,
is
one
of
the
bigge
st
proble
m
s
in
our
day
-
to
-
da
y
l
ife,
resultin
g
in
del
a
y
s,
wasta
ge
of
fue
l
and
m
one
y
.
In
t
his
pape
r,
a
ne
w
m
odel
is
dev
el
oped
using
Sim
ula
ti
on
o
f
Urban
Mobili
t
y
(SU
MO
)
si
m
ula
tor
for
sim
ula
t
in
g
a
re
al
ist
ic
tra
f
fic
sc
ena
rio
for
a
la
rg
e
c
ity
l
i
ke
Bhubane
sw
ar
where
,
tra
ffi
c
c
ongesti
on
is
a
cr
it
ical
issue.
In
a
c
ity
,
tr
aff
i
c conge
stion is c
h
ara
c
te
ris
ed
b
y
m
an
y
par
amete
rs s
uch
as
r
api
d
growth
of
popula
t
ion
,
num
ber
of
four
whee
le
rs,
ina
dequate
and
poor
roa
d
infra
struc
ture
s
a
nd
shortage
of
ph
y
sic
al
p
la
n
to
gover
n
the
dev
el
opm
ent
s
,
which
ar
e
foc
us
ed
on
enh
ancin
g
the
vo
lume
o
f
the
ro
ads
b
y
rai
sing
th
e
num
ber
of
la
n
es,
over
-
passes,
under
passes
a
nd
over
-
bridge
s
at
m
an
y
junc
ti
ons
.
How
eve
r,
for
th
e
succ
ess
of
the
se
m
aste
r
pl
ans
to
fully
over
come
the
cong
esti
on
i
ss
ues,
it
is
ne
ces
sar
y
to
tra
nsm
it
the
conge
stion
informati
on
to
vehicle
s
coming
towar
ds
a
co
ngesti
on
ar
ea
b
y
using
a
Vehi
cu
la
r
Ad
-
ho
c
Network.
Thi
s
pape
r
an
aly
z
es
cl
uster
ing
te
chn
ique
s
in
Vehic
u
la
r
Ad
-
hoc
Networks
to
det
ect
conge
stion
in
roa
ds
with
the
m
ini
m
al
infra
struct
ur
al
support.
The
ra
w
dat
a
from
vehi
cles
are
class
ified
using
cl
uster
ana
l
y
sis.
Out
of
a
num
ber
of
al
gorit
hm
s
tha
t
are
used
to
solve
the
conge
st
i
on
det
ection
proble
m
,
thr
ee
i
m
porta
nt
al
gor
ithm
s
such
as
Cent
roid
base
d
K
-
m
e
ans,
object
base
d
FC
M
and
FK
M
al
gorit
hm
s
are
compar
ed
in
thi
s
work
on
the
basis
o
f
dat
a
po
int
s
and
num
ber
of
cl
ust
e
rs.
The
result
s
o
f
the
al
gori
thms
are
cl
ose
to
ea
ch
oth
er
,
but
fuz
z
y
techniques
are
pre
fer
ab
le
as
the
tra
ffic
si
tua
ti
ons
ar
e
d
y
nami
c
in
na
tur
e.
Ke
yw
or
ds:
Fu
zzy
C
-
m
eans Clusterin
g
Fu
zzy
K
-
m
eans Clusterin
g
K
-
m
eans Clust
erin
g
Traffic
Co
nges
ti
on
Veh
ic
ular A
d Ho
c
N
et
wor
k
Copyright
©
201
9
Instit
ut
e
o
f Ad
vanc
ed
Engi
n
ee
r
ing
and
S
cienc
e
.
Al
l
rights re
serv
ed.
Corres
pond
in
g
Aut
h
or
:
An
it
a M
oh
a
nty
,
Dep
a
rtm
ent o
f El
ect
ro
nics
& I
ns
tr
um
entat
ion
Enginee
rin
g,
SI
T,
Bh
ubanes
war,
751024,
Ind
ia
.
Em
a
il
:
anita
@sil
ic
on
.ac.i
n
1.
INTROD
U
CTION
Transp
or
ta
ti
on
traff
ic
co
ntr
ol
is
a
crit
ic
al
prob
le
m
in
this
adv
a
nce
d
era.
A
lot
of
ti
m
e
and
fu
el
ar
e
wasted
eve
ryd
ay
by
veh
ic
le
s
facin
g
c
on
ges
ti
on
a
rou
nd
th
e
w
or
l
d
[1
]
.
T
he
reas
on
be
hi
nd
it
is
the
inc
rease
i
n
popula
ti
on
a
nd
the
num
ber
of
veh
ic
le
s
in
la
r
ge
ci
ti
es.
Be
caus
e
of
this,
a
n
autom
at
ed
traff
ic
con
t
ro
l
syst
e
m
is
require
d
t
o
m
anag
e
the
c
ongestio
n
pro
blem
s
m
oo
thly
and
on
a
c
onti
nuous
basis
[
2]
.
Tra
ff
ic
c
on
gestio
ns
occur
ei
ther
due
to
s
om
e
e
xter
nal
facto
rs
su
ch
a
s
r
oad
m
ai
ntenan
ce,
ru
s
h
hours,
heav
y
rain,
fog
a
nd
bo
tt
le
nec
k
c
on
diti
on
,
et
c.,
w
hich
a
re
pr
e
di
ct
able
or
unpredict
able
inci
de
nts
create
d
due
to
the
be
ha
viour
of
dr
i
ver
s
,
acci
de
nts,
et
c.
In
la
r
ge
ci
ti
es
traff
ic
congesti
on
is
beco
m
ing
wor
se
due
to
t
he
r
ise
in
popula
ti
on,
the
nu
m
ber
of
fou
r
w
heelers
,
a
r
apid
dev
el
op
m
ent
of
busine
s
s
centres
a
nd
an
inc
rease
in
so
ci
al
an
d
eco
no
m
ic
act
ivit
ie
s.
As
day
by
day
the
num
ber
of
veh
ic
le
s
is
incr
easi
ng,
traf
fic
congesti
on
bec
om
es
a
t
ypic
al
scenari
o
i
n
la
rg
e
ci
ti
es
w
hi
ch
waste
a
l
ot
of
ti
m
e
as
well
as
fu
el
.
In
la
rg
e
ci
ti
es,
w
e
o
bse
r
ve
cert
ai
n
hu
dd
le
s
li
ke
the
Evaluation Warning : The document was created with Spire.PDF for Python.
Ind
on
esi
a
n
J
E
le
cEng& C
omp Sci
IS
S
N:
25
02
-
4752
Tra
ff
ic
con
gestion
detect
io
n
i
n a city
usi
ng c
lusteri
ng tech
ni
qu
es
in
V
ANE
Ts
(
Anita
Moh
an
ty
)
885
beh
a
vior
of
ou
tsi
der
s,
co
ns
tr
uc
ti
on
/re
pair
w
ork
of
r
oa
ds
/p
avem
ents,
wea
ther
c
onditi
ons
an
d
the
be
ha
vi
or
of
street
hawker
s that
slow
dow
n
the
traff
ic
and
create
pr
oble
m
s
fo
r
pe
op
le
travell
in
g
from
on
e p
la
ce
to
an
oth
e
r.
Also
,
the
co
ns
t
ru
ct
io
n
of
r
oa
ds/
pav
em
ents
or
re
pair
of
pot
hole
s
ca
us
e
m
ore
am
ou
nt
of
de
la
y
and
in
t
urn
le
ad
to
sever
e
tra
ff
i
c
congesti
on.
Dr
ai
nag
e
syst
em
s
in
citie
s
o
n
rainy
days
work
bad
ly
an
d
introd
uce
de
la
ys
of
about
30
to
45
m
inu
te
s
in
tra
vel.
S
om
et
i
m
e
s,
co
ngest
io
ns
ta
ke
place
due
to
U
-
tu
r
n
of
ve
hicle
s
duri
ng
peak
hours.
T
he
ci
ty
adm
inist
rator
s
ar
e
try
in
g
t
o
work
out
th
e
pro
blem
by
wide
ning
the
r
oads,
by
co
ns
t
ru
ct
in
g
ov
e
r
-
bri
dges,
over
-
pa
sses,
un
derpasses
et
c.
at
the
ju
nctions.
S
uch
ste
ps
a
re
not
s
uffic
ie
nt
enou
gh
to
m
i
nim
iz
e
traff
ic
now
-
a
-
da
ys.
Hen
ce
, effici
en
t i
ntell
igent sys
tem
s ar
e req
uir
ed
to b
e im
ple
m
ented
to contro
l t
he
cu
rr
e
nt
traff
ic
j
am
s
in
bi
g
ci
ti
es,
w
her
e
ve
hicle
s
c
an
c
omm
un
ic
a
te
with
each
ot
her
us
i
ng
Ve
hicular
Ad
H
oc
Netw
orks
(VA
NETs
)
to
ove
rco
m
e
the
c
ongestio
n
pro
blem
[3
-
5]
.
These
c
onge
sti
on
detect
io
n
m
echan
ism
s
i
m
ple
m
ented
in
a
n
intel
li
gen
t
syst
e
m
can
be
cat
egorized
int
o
two
ty
pes:
tr
aff
ic
m
anag
em
ent
con
tr
ol
un
it
base
d
c
onge
sti
on
detect
ion
a
nd
VANET
base
d
co
ng
est
i
on
de
te
ct
ion
.
In
t
his
first
a
ppr
oach
a
ple
nty
of
se
ns
ors
a
re
i
ns
ta
ll
ed
to
gathe
r
traf
fic
inf
or
m
at
ion
an
d
the
co
ntr
ol
unit
is
us
ed
to
gove
r
n
the
eve
nt
of
r
oad
c
onge
sti
on
by
sc
ru
t
inizi
ng
the
data
c
ollec
te
d
f
r
om
the
sens
or
s
[
6
-
9].
I
n
th
e
sec
ond
a
ppr
oach
ve
hicle
s
in
m
ov
i
ng
conditi
on
are
us
e
d
to
colle
ct
the
in
f
or
m
at
ion
of
ve
hicle
s
in
cl
os
e
pro
xim
i
ty
and
ta
ke
decisi
on
about
c
ongesti
on
an
d
e
xc
ha
nge
the
ro
a
d
c
onditi
on
due
to
c
onge
sti
on
with
the
oth
e
r
ve
hi
cl
es.
Acc
ordin
g
to
a
num
ber
of
congesti
on
det
ect
ion
m
et
ho
ds
s
uc
h
as
autom
at
ic
tr
aff
ic
co
ngest
io
n
ide
ntific
at
ion
bas
ed
on
gai
n
am
plifie
r
theor
y
[
10]
,
co
ng
est
ion
recog
niti
on
us
i
ng
wav
el
et
te
chn
i
qu
e
[11]
an
d
co
ng
est
i
on
de
te
ct
ion
by
patte
rn
rec
ogniti
on
[12],
t
he
am
ount
of
m
essages
received
f
r
om
ind
iv
idu
al
ve
hicle
s
are
m
or
e.
As
t
he
ba
ndwidt
h
avail
abili
ty
in
a
VANET
is
finite
,
a
m
essage
ag
gre
gation
sc
hem
e
is
re
qu
i
red
to
be
c
onside
re
d.
A
str
uctu
re
-
f
ree
m
essage
a
ggre
gation
sc
hem
e
i
s
descr
i
bed
in
[13]
wh
e
re
ind
i
vi
du
al
ve
hicle
with
the
othe
r
veh
ic
le
s
m
ay
beh
ave
as
a
m
e
ssage
ag
gregat
or,
but
the
disa
dvanta
ge
is
t
hat
it
shou
l
d
be
within
a
pre
def
ine
d
area
of
t
he
e
ve
nt.
C
ooper
at
i
ve
T
raffic
Co
ngest
io
n
Detect
ion
(C
oT
EC)
is
a
m
et
h
od
[
14]
,
w
hich
h
an
dles
a
m
ess
age
a
ggre
gatio
n
t
ech
ni
qu
e
ba
sed
on
f
uzzy
l
ogic
to
detect
the c
onge
sti
on
on the
road
. B
ut, none
of these m
et
hods
a
re a
ble to r
edu
ce
the
ba
nd
width re
quirem
ent.
As
the
m
ov
e
m
ent
of
veh
ic
le
s
is
dynam
ic
in
nat
ur
e,
a
f
uzzy
log
ic
bas
ed
cl
us
te
rin
g
te
chn
i
qu
e
is
pr
e
fer
a
ble.
At
a
jun
c
ti
on,
w
he
n
veh
ic
le
s
a
r
e
facin
g
c
onge
sti
on
,
nea
rer
ve
hicle
s
ha
ve
t
heir
par
am
et
ers
ve
ry
cl
os
e
to
eac
h
oth
e
r.
So,
by
us
in
g
a
ppr
opri
at
e
cl
us
te
rin
g
al
gorithm
a
cl
us
te
r
can
be
f
or
m
ed
with
pa
ram
et
ers
m
or
e
si
m
il
ar
to
each
oth
e
r.
When
t
heir
cl
us
te
r
centres
a
re
cl
os
e
to
eac
h
othe
r,
it
m
eans
that
the
veh
ic
le
s
a
re
in
a close
pro
xim
it
y l
eading
t
o
c
ongestio
n.
On
e o
f
t
he
popula
r
te
ch
niques
-
m
eans
cl
us
te
ring
(a h
ar
d
cl
ust
ering
te
c
hn
i
que)
is
fast, r
ob
us
t,
easi
e
r
to
i
m
ple
m
ent
and
bette
r
co
m
pu
ta
ti
on
al
tim
e.
But
it
is
u
ns
uc
cessf
ul
in
getti
ng
ov
e
rla
pp
i
ng
cl
us
te
rs
[15].
I
n
fu
zzy
cl
us
te
ri
ng
te
chn
i
qu
e
s,
an
obj
ect
is
no
t
on
ly
the
m
e
m
b
er
of
a
cl
us
te
r
bu
t
m
e
m
ber
of
m
any
cl
us
te
rs.
But
this
syst
e
m
is
relat
ively
co
stl
ie
r
than
hard
com
pu
ti
ng
t
echn
i
qu
e
s.
O
ur
pap
e
r
com
par
es
three
di
ffere
nt
cl
us
te
rin
g
al
gorithm
s
to
de
te
ct
congesti
on
by
ta
king
s
om
e
of
the
ve
hicle
pa
ram
e
te
rs
li
ke
s
pee
d,
fuel
consum
ption
a
nd
C
O
2
e
m
issio
n
i
nto
c
on
si
de
rati
on
i
n
a
V
A
NET
e
nv
ir
onm
ent.
I
n
our
rese
arch
w
or
k,
-
m
e
ans,
Fu
zzy
C
-
m
eans and
F
uzzy
K
-
m
eans clustering alg
o
rithm
s ar
e a
naly
sed b
ased
on their
e
xecu
ti
on ti
m
e.
These
a
bove
cl
us
te
rin
g
te
ch
niques
are
te
s
te
d
to
co
ntr
ol
the
traf
fic
prob
le
m
in
a
big
ci
ty
li
ke
Bhuba
nesw
a
r,
an
adm
inist
rativ
e,
inf
orm
ation
te
chnolo
gy,
edu
cat
io
n
an
d
tourism
dr
ive
n
ci
ty
.
Altho
ugh
these
te
ch
niques
exi
st,
these
we
re
ap
plied
for
t
r
aff
ic
c
ongestio
n
detect
ion
ea
rlie
r.
In
this
pa
per
the
real
world
veh
ic
le
data
ar
e
extracte
d
a
nd
sim
ulate
d
in
m
at
la
b
us
in
g
t
hese
te
c
hn
i
qu
e
s.
Sect
io
n
2
giv
es
a
n
ov
e
r
view
of
a
VANET
.
T
he
co
ngest
ion
a
t
Kali
nga
Hospita
l
J
unct
io
n
of
B
huba
nes
war
Ci
ty
is
create
d
us
i
ng
SU
MO
si
m
ulator
an
d
var
i
ou
s
pa
ram
et
ers
are
e
xtra
ct
ed
f
or
cl
ust
erin
g
as
ex
plain
ed
in
Sect
io
n
3.
T
he
m
et
ho
dolo
gies
are
ex
plaine
d
i
n
Sect
io
n
4.
T
hese
are
il
lustr
at
ed
ta
king
one
exam
ple
scenari
o
with
t
he
resu
lt
s
are
re
porte
d
in
Sect
ion
5. Fina
ll
y, Sect
ion
6
c
on
cl
ud
e
s the
paper
.
2.
VANET:
A
N OVER
VIEW
VANETs
are
c
os
t
ef
fecti
ve,
di
stribu
te
d
traf
fic
congesti
on
de
te
ct
ion
syst
em
s.
These
generall
y
requir
e
a
set
of
inex
pe
ns
ive
de
vices,
wh
ic
h
can
be
i
ncor
porated
int
o
ve
hicle
s
and
wh
ic
h
com
m
u
nicat
e
with
a
sat
el
li
t
e
to
accum
ulate
the
data
and
tr
ansf
e
r
it
to
the
syst
e
m
.
In
an
In
te
ll
igent
Tr
ans
port
Syst
em
(I
TS),
eac
h
dev
ic
e
works
as
a
se
nsor
,
recei
ver
a
nd
r
ou
te
r
to
broad
ca
st
the
in
f
or
m
at
ion
thr
ou
ghout
the
network,
f
or
a
sa
fe
an
d
co
m
fo
rtable
dri
vin
g
a
nd
tra
ve
ll
ing
e
xp
e
rien
ce.
T
his
c
on
ti
nuous
e
xc
hang
e
of
in
f
or
m
at
i
on
bet
wee
n
ve
hicle
s
include
s
data
a
bout
the
s
peed
of
the
ve
hicle
s
and
t
heir
locat
ion
s
.
V
ANETs
are
us
e
d
in
a
n
ITS
to
im
pr
ov
e
the
dr
i
ving effic
ie
ncy, tra
ff
ic
sa
f
et
y and
c
om
fo
r
t
and also
d
et
e
ct
r
oa
d
c
ongest
ion
[
16,
17
]
.
The
m
ai
n
com
pone
nts
of
a
n
I
TS,
as
sho
wn
in
Fi
gure
1
are:
A
pp
li
cat
ion
U
nits
(
AU
s
),
ve
hicle
Boa
r
d
Un
it
s
(B
Us
)
an
d
Roa
d
Si
de
U
nits
(RS
Us),
in
sta
ll
ed
separ
at
el
y
or
integ
rate
d
with
B
Us.
A
Us
are
sop
histi
cat
ed
dev
ic
es
w
h
ic
h
pro
vid
e
a
ppli
cat
ion
s
relat
ed
to
ve
hicle
saf
et
y.
BUs
are
instal
le
d
on
bo
ard
of
a
veh
ic
le
an
d
com
m
un
ic
at
e
with
BUs
i
ns
t
al
le
d
in
ot
her
veh
ic
le
s
or
wi
th
Roa
d
Side
Un
it
s.
T
hey
al
so
c
omm
un
ic
a
te
with
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
S
N
:
2502
-
4752
Ind
on
esi
a
n
J
E
le
cEng& C
omp Sci,
V
ol.
13
, No
.
3
,
Ma
rch 2
019
:
884
–
891
886
AU
s
.
RS
Us
a
r
e
fixe
d
unit
s
in
sta
ll
ed
al
ong
t
he
si
de
of
the
r
oad
to
pro
vid
e
the
co
ve
rag
e
a
nd
co
nnect
ivit
y
to
al
l
veh
ic
le
s
[19].
Figure
1. A
rch
i
te
ct
ur
e
of
an
ITS [
18]
Figure
2.
Proce
ss Flow
for
e
xt
racti
ng Real
Tim
e d
at
a
from
v
ehicl
es
3.
E
X
T
R
AC
TI
O
N
O
F P
ARA
METE
RS US
I
NG SU
MO
SI
MU
L
ATO
R
Si
m
ulati
on
of
Urba
n
Mob
il
it
y
(S
UMO
)
sim
ulator
is
us
e
d
to
create
a
traffi
c
scenario
f
r
om
wh
ic
h
the
par
am
et
ers
of
veh
ic
le
s
are
e
xtracted
to
det
ect
the
con
ges
ti
on
in
a
par
ti
cular
junc
ti
on.
The
proces
s
flow
f
or
extracti
ng
real
tim
e
data
fr
om
veh
ic
le
s
is
sh
own
in
Fig
ur
e
2.
The
real
m
a
p
of
Kali
ng
a
H
os
pital
Junct
io
n,
the
m
os
t
cro
wd
e
d
j
unct
io
n
of
Bh
ub
a
nes
war,
is
ta
ken
from
Op
en
Street
Ma
p
a
nd
giv
e
n
to
S
UMO
f
or
si
m
ulati
on
of
a
real
tim
e
traf
fic
sce
nari
o.
S
UMO
is
an
op
e
n
sourc
e,
hi
gh
ly
port
able,
m
ic
ro
sco
pic
an
d
co
nti
nuou
s
ro
a
d
traf
fic
s
i
m
ulati
on
pac
kag
e
desig
ne
d
to
handle
la
rg
e
ro
a
d
ne
twork
s.
T
he
s
cenari
os
in
S
UMO
si
m
ulator
has
t
wo
pa
rts:
ro
a
d
netw
ork
(m
ap
)
inclu
ding
ro
a
ds
,
st
reets,
tra
f
fic
li
gh
ts
j
unct
ion
s
et
c.
a
nd
t
raffic
dem
and
expres
sing
the
detai
ls
of
ve
hicle
s
li
ke
sp
ee
d
of
ve
hicle
s,
directi
on,
de
pa
rtur
e
ti
m
e
and
arr
ival
tim
e,
po
sit
io
n
et
c.
Figure
3(
a
)
sh
ows
t
he
i
m
po
rt
net
work
of
Kal
ing
a
Hospita
l
Ju
nc
ti
on
ta
ken
from
http:/
/ openst
re
etm
ap.
org [20]
.
(a)
(b)
Figure
3. (a
) O
rigin
al
O
pen St
reet M
ap o
f Ka
li
ng
a
Hospita
l
Ju
nc
ti
on, Bhu
ba
nes
war
[
20
]
a
nd
(
b) I
m
po
rte
d
m
ap
from
O
S
M i
n
S
UMO
This
dow
nlo
a
de
d
m
ap
save
d
i
n
.
os
m
file
for
m
at
is
i
m
po
rted
to
S
UMO
to
create
traf
fic
e
nv
i
ronm
ent
wh
ic
h
is
save
d
in
.c
f
g
file
as
show
n
in
Figure
3(b)
with
the
help
of
Netco
nve
rt,
P
olyc
onvert
an
d
rand
om
Trips.
py
too
ls.
T
he
c
ongestio
n
on
a
ro
a
d
in
S
UMO
sim
ulato
r
is
create
d
by
de
la
yi
ng
a
ve
hicl
e
on
a
la
ne,
w
hich
c
an
be
ass
um
e
d
as
an
acci
de
nt
on
a
r
oa
d
in
real
world
.
At
Kali
nga
Ho
s
pital
Ju
nct
ion
th
e
congesti
on
is
create
d
in
S
UM
O
as
show
n
in
the
Figure
4.
T
hen
the
raw
outpu
ts,
wh
ic
h
con
ta
in
s
la
ne
id
,
CO,
CO
2
,NO
x,,
PM
x,
no
ise
,
f
uel
con
s
um
ption
,
m
axim
u
m
sp
eed,
m
ean
sp
eed
et
c.
are
extracte
d
from
the
si
m
ulator
for
sim
ulati
on
.
O
pen S
t
reet Map
(O
SM
)
(
www.ope
ns
treetmap.o
rg)
Si
mul
ati
on of Urb
an Mob
i
l
i
ty
(SUM
O
)
P
y
thon
Extr
ac
ti
on of pa
r
ameters
Evaluation Warning : The document was created with Spire.PDF for Python.
Ind
on
esi
a
n
J
E
le
cEng& C
omp Sci
IS
S
N:
25
02
-
4752
Tra
ff
ic
con
gestion
detect
io
n
i
n a city
usi
ng c
lusteri
ng tech
ni
qu
es
in
V
ANE
Ts
(
Anita
Moh
an
ty
)
887
Figure
4. Co
ngest
ion
c
reated
at
the K
al
i
ng
a
Ho
s
pital
Juncti
on
4.
CLUS
TE
RI
N
G A
L
GO
RIT
HMS FO
R D
ET
ECTION
OF
CONGES
TION
Mostl
y,
ve
hicle
s
instal
le
d
with
a
var
ie
ty
of
on
-
bo
a
rd
sen
sors
gen
e
rate
ple
nty
of
m
essages
that
yi
el
d
the
issue
of
c
ha
nn
el
com
petition
an
d
ex
ha
ust
the
lim
i
te
d
avail
able
band
width.
In
a
V
AN
E
T,
on
boa
rd
unit
s
are
instal
le
d
i
n
ve
hicle
s
to
ac
cum
ulate
the
outp
uts
of
the
se
ns
ors
s
uc
h
as
t
he
veh
ic
le
s
s
pe
ed,
f
uel
c
onsum
pt
ion
and
CO
2
em
iss
ion
i
nto
a
si
ng
le
m
essage
an
d
tra
ns
m
it
to
a
ll
veh
ic
le
s
out
of
w
hich
one
beh
a
ves
a
s
a
node
to
process
them
us
ing
cl
ust
erin
g
te
chn
iq
ue.
Usi
ng
cl
us
te
rin
g
t
echn
i
qu
e
s,
the
dataset
is
pr
eci
sel
y
par
ti
ti
on
e
d
into
c
lusters
s
uc
h
t
hat
the
data
i
n
each
cl
ust
er
ha
s
the
sam
e
disti
nguish
e
d
at
tri
bu
te
.
T
he
veh
i
cl
es
with
t
he
s
a
m
e
or
near
ly
sam
e
sp
eed
are
gro
up
e
d
to
gethe
r
into
a
sing
le
cl
us
te
r.
T
he
m
ini
m
u
m
distance
bet
ween
t
he
cente
rs
of
the cluster
s
de
ci
des
the
close
ness betw
ee
n
t
he
cl
ust
ers
a
nd
u
lt
i
m
at
el
y t
he
congesti
on i
n
a
lane.
4.1.
K
-
me
an
s
C
lus
tering
-
m
eans
cl
us
te
ri
ng
is
a
par
ti
ti
onin
g
al
go
rithm
w
her
e
obj
ect
s
of
a
data
set
are
orga
nized
into
par
ti
ti
on
s
(
≤
)
w
her
e
the
pa
rtit
i
on
s
a
re
re
prese
nted
as
a
cl
us
t
e
r.
He
re,
t
he
obj
ect
s
belo
ng
to
a
cl
us
te
r
a
re
sai
d
to
be
“si
m
il
ar”
to
each
ot
her
a
nd
“
dissim
il
ar”
to
ob
j
ec
ts
in
oth
er
cl
ust
ers
in
te
rm
s
of
the
at
tribu
te
s
of
th
e
data
set
[
21
-
22]
.
I
n
t
he
Ce
ntr
oid
-
base
d
-
m
e
ans
cl
us
te
ri
ng
te
chn
iq
ue
the
Ce
ntro
i
d
of
a
cl
us
te
r
i
s
the
c
enter
po
i
nt
a
nd
is
diff
e
ren
ti
at
ed
fro
m
data
po
ints
by
E
uclidean
di
sta
nce
bet
wee
n
the
tw
o
obj
e
ct
s
(or
points)
and
.
4.2.
F
uz
z
y
C
-
mea
ns clus
ter
ing
The
F
uzzy
C
-
m
eans
(F
CM
)
cl
us
te
rin
g
is
an
un
s
uper
vis
ed
cl
us
te
rin
g
al
gorithm
wh
ic
h
cre
at
es
cl
us
te
rs
by
ta
kin
g
the
data
points
ha
vi
ng
a
high
de
gr
ee
of
belo
ngin
gnes
s
to
that
cl
us
te
r.
T
he
distan
ce
from
any
giv
e
n
data
poi
nt
to
t
he
c
luster
center
is
ex
presse
d
as
m
ini
m
u
m
obj
e
ct
ive
f
unct
io
n
[23
-
24]
.
(
2
)
is
us
e
d
to
cal
c
ulate
the
tim
e
com
plexity
of
the
FCM
al
go
r
it
h
m
,
her
e
t
he
total
num
ber
of
obj
ect
s
is
,
the
nu
m
ber
of clus
te
rs
is
an
d
the
nu
m
ber
of it
er
at
ion
is
.
4.3.
F
uz
z
y
K
-
me
an
s
cl
ust
e
ri
ng
In
fu
zzy
-
m
ea
ns
cl
us
te
rin
g,
a
giv
e
n
group
of
featu
re
ve
ct
or
s
conve
rted
into
an
im
pro
ved
on
e
thr
ough
par
ti
ti
on
i
ng
data
poi
nts.
T
his
proce
ss
sta
rts
with
a
gro
up
of
i
ntrodu
ct
or
y
cl
us
te
r
centers
an
d
re
runs
this
process
ti
ll
it
sat
isfie
s
a
stoppin
g
crit
er
ion
.
It
is
ex
pe
ct
ed
that
tw
o
cl
us
te
rs
don’
t
hav
e
t
he
sam
e
cl
us
t
er
centers.
If
t
hey
are
sam
e,
then
a
cl
us
te
r
ce
nt
er
com
es
ou
t
of
t
he
proce
ss
to
av
oid
c
oin
ci
den
ce
[
25
]
.
He
re
the
fu
zzy
relat
ion
s
hip
betwe
en
a
data
poi
nt
an
d
cl
us
te
r
ce
nters
is
represe
nted
by
a
m
e
m
ber
sh
ip
∈
[
0
,
1
]
val
ue
wh
ic
h rep
rese
nt
s the
degree
of b
el
o
ngin
gn
e
ss
of
data point
and cluste
r
ce
nt
er
.
5.
COMP
ARAT
IVE
ANALY
S
IS O
F
THE
C
LUSTE
RI
NG
APP
ROAC
H
ES
5.1.
The
Data Se
t
We
hav
e
us
e
d
the
data
set
as
show
n
i
n
Ta
bl
e
1
to
c
ha
racteri
ze
the
m
essages
with
at
tri
bu
te
s
of
sp
ee
d
(
km
/hr)
, fuel c
onsu
m
ption (
m
l/
s)
and
CO
2
(m
g/s)
em
i
ssion w
hich
a
r
e taken
fro
m
SU
MO
Sim
ulato
r
. T
he
m
essages
colle
ct
ed
f
or
detect
ion
of
c
onge
sti
on
a
re
ge
ne
rated
by
the
on
boar
d
unit
s
inst
al
le
d
in
the
ve
hicle
s.
Fo
r
t
he
detect
ion
of
tra
ff
ic
co
ng
e
sti
on,
a
t
otal
of
27
ve
hicle
s
are
ta
ken
to
f
or
m
the
data
set
.
T
hese
sa
m
ples
are
groupe
d
to
form
d
iffer
e
nt
cl
us
te
rs
a
nd th
en
a
re c
om
par
ed.
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
S
N
:
2502
-
4752
Ind
on
esi
a
n
J
E
le
cEng& C
omp Sci,
V
ol.
13
, No
.
3
,
Ma
rch 2
019
:
884
–
891
888
Table
1
. T
he
Dat
a Set
Attribu
tes
Sa
m
p
le
Nu
m
b
er
Sp
eed
(K
m
/Hr
)
Fu
el Co
n
su
m
p
tio
n
(
m
l/sec
)
CO
2
E
m
iss
io
n
(
m
g
/sec)
1
0
1
.13
2
6
2
4
.72
2
1
.70
1
.37
3
1
8
0
.87
3
2
.54
1
.18
2
7
4
3
.50
4
3
.12
1
.57
3
6
5
5
.93
5
2
.86
1
.49
3
4
6
0
.67
6
9
.30
3
.60
8
3
8
2
.62
7
1
0
.43
3
.00
6
9
8
3
.71
8
1
1
.56
3
.86
8
9
7
1
.36
9
1
5
.64
3
.64
8
4
7
5
.07
10
1
6
.48
3
.87
9
0
0
1
.59
11
1
7
.19
5
.39
1
2
5
3
6
.11
12
1
7
.33
4
.46
1
0
3
7
1
.44
13
2
2
.34
7
.74
1
8
0
0
6
.88
14
2
1
.98
7
.60
1
7
6
8
0
.25
15
2
1
.73
5
.73
1
3
3
2
2
.30
16
2
7
.17
2
.59
6
0
2
0
.8
3
17
2
6
.11
7
.75
1
8
0
3
5
.66
18
2
6
.44
6
.47
1
5
0
4
4
.55
19
2
7
.30
2
.28
5
2
9
7
.24
20
5
.84
2
.40
5
5
8
4
.19
21
6
.76
1
.89
4
4
0
7
.81
22
2
7
.64
3
.67
8
5
4
9
.24
23
1
9
.88
7
.02
1
6
3
3
0
.06
24
1
8
.22
4
.79
1
1
1
4
7
.94
25
1
2
.9
4
.54
1
0
5
5
3
.88
26
6
.59
2
.74
6
3
8
5
.8
1
27
8
.26
2
.83
6
5
7
8
.9
8
5.2.
E
xp
eri
m
ent
al R
es
ults
an
d
Obs
e
rv
at
i
on
s
The
-
m
eans
Clu
ste
rin
g,
Fu
zz
y
C
-
m
eans
Cl
us
te
rin
g
a
nd
F
uzzy
-
m
eans
cl
us
te
rin
g
te
ch
ni
qu
es
a
re
i
m
ple
m
ented
in Mat
la
b 2
015.
5.2.1. Im
plem
ent
at
io
n
of
-
M
eans
Clus
terin
g
The
m
essages
from
veh
ic
le
s
in
t
he
×
data
m
atr
ix
w
her
e
is
the
num
ber
of
data
m
essages
a
nd
is
t
he
num
ber
of
at
trib
utes
of
th
os
e
m
ess
ages
are
gr
ouped
int
o
cl
us
te
rs.
T
he
-
m
ea
ns
grap
h
f
or
t
he
veh
ic
le
data
set
(sp
ee
d,
fu
e
l
con
s
um
ption
and
C
O
2
em
issi
on)
re
pr
ese
nts
three
cl
us
t
ers.
T
he
gr
a
phic
al
represe
ntati
on
of
the
scat
te
re
d
ve
hicle
s
ha
vin
g
t
hr
ee
at
trib
utes:
sp
ee
d,
f
ue
l
con
s
um
ption
an
d
CO
2
em
i
ssion
a
s
m
entioned in
the
dataset
is s
how
n
i
n
the
Fig
ur
e
5
(a).
(a)
(b)
(c)
Figure
5. (a
)
-
m
eans cl
us
te
rin
g o
f data
m
essa
ges, (
b) Fuzz
y
C
-
Me
ans
cl
us
t
erin
g of
data m
essages
and
(c)
Fu
zzy
K
-
Me
ans
cl
us
te
rin
g o
f data
m
essa
ges
5.2.2. Im
plem
ent
at
io
n
of Fu
zz
y
C
-
Me
an
s
Clust
eri
n
g
The
Fu
zzy
C
-
m
eans
Cl
us
te
ring
(
FCM
)
is
us
ed
to
perf
or
m
cl
us
te
ring
of
diff
e
re
nt
m
ess
ages
recei
ve
d
from
var
iou
s
veh
ic
le
s
at
a
jun
ct
io
n.
The
f
un
ct
io
n
FCM
ta
kes
the
data
set
fr
om
the
ve
hicle
s
an
d
a
desire
d
Evaluation Warning : The document was created with Spire.PDF for Python.
Ind
on
esi
a
n
J
E
le
cEng& C
omp Sci
IS
S
N:
25
02
-
4752
Tra
ff
ic
con
gestion
detect
io
n
i
n a city
usi
ng c
lusteri
ng tech
ni
qu
es
in
V
ANE
Ts
(
Anita
Moh
an
ty
)
889
nu
m
ber
of
cl
ust
ers
are
ge
nerat
ed.
Fi
gure
5(
b)
is
a
t
hr
e
e
di
m
ension
al
plo
t
of
t
he
th
ree
at
tribu
te
s,
sp
ee
d,
fu
el
consum
ption
a
nd CO
2
em
issio
n f
or each
of t
he vehicl
es a
nd
red X
m
ark
s
sh
ows
the ce
nt
ers of
clusters
.
5.3.3
Im
plem
e
nt
at
io
n
of Fuz
z
y
K
-
me
an
s
Cl
ust
eri
ng
So
m
et
i
m
es
m
os
t
of
the
ve
hicle
s
do
no
t
ha
ve
cl
ear
at
trib
utes.
He
nce
a
n
i
nterm
ediary
nature
in
qu
al
it
y
and
ty
pe
exist
s
betwee
n
th
e
veh
ic
le
s
for
wh
ic
h
a
s
of
t
div
isi
on
is
re
quire
d.
T
he
fu
z
zy
K
-
m
eans
(F
KM)
cl
us
te
rin
g
te
c
hniq
ue
is
a
best
m
et
ho
d
to
w
ork
on
the
a
bove
sai
d
pro
blem
.
The
Fu
zzy
-
m
e
ans
grap
h
with
t
he
veh
ic
le
data
set
(sp
ee
d,
fu
el
consum
ption
a
nd
C
O
2
em
issio
n)
re
presents
three
cl
ust
ers.
Figure
5(
c
)
shows
a
scat
te
red
F
uzz
y
-
m
eans
gr
ap
h
of
veh
ic
le
da
ta
set
with
thr
ee
at
tribu
te
s:
s
peed,
f
uel
co
nsum
ption
an
d
CO
2
e
m
issi
on
.
5.3.4
E
xp
eri
m
ent
al R
es
ults
The
ef
fici
ency
of
FCM
,
-
m
e
ans
an
d
F
uzzy
-
m
eans
te
chni
qu
es
a
re
te
ste
d
in
Ma
tl
ab
[
26
]
.
Al
l
com
pu
ta
ti
on
s
are
pe
r
form
ed
on
H
P
I
ntel(R)
Core
(TM)
i3
-
4000M
CP
U
@
2.4
0GHz
w
it
h
4GB
RAM.
In
ou
r
exp
e
rim
ent,
the
data
are
the
m
essages
com
i
ng
from
veh
ic
le
s
m
ov
in
g
to
w
ar
ds
a
c
ongest
ed
area
.
27
m
e
ssages
are
re
cei
ve
d
with
at
trib
utes
of
s
pee
d,
fu
e
l
consum
ption
an
d
C
O
2
em
i
ssion.
T
hat
m
eans
t
he
data
set
is
consi
sti
ng
of
27
data
po
i
nts.
T
he
a
ve
rag
e
com
pu
ti
ng
ti
m
e
(in
s
eco
nd
s)
for
-
m
eans,
FCM
a
nd
F
uz
zy
-
m
eans
are
li
ste
d
i
n
t
he
Ta
ble
2
with
50
num
ber
s
of
it
erati
ons.
It
is
obser
ve
d
from
the
Ta
ble
2
that
-
m
ea
ns
cl
us
te
rin
g
te
c
hniq
ue
c
ons
ume
s
le
ss
a
ve
rage
com
pu
ti
ng
t
i
m
e
than
FC
M
an
d
Fu
zzy
-
m
eans
cl
us
t
erin
g
te
chn
iq
ue.
The
distances
bet
ween
t
he
cl
us
t
er
centers
for
diff
e
re
nt
te
chni
qu
es
a
re
li
ste
d
in
Ta
ble
3
,
Table
4
and
Table
5
.
Fr
om
our
re
sul
ts,
it
is
sh
own
that
al
l
the
distances
m
easur
e
d
betwee
n
the
cl
us
te
rs
i
n
Fu
zzy
-
m
eans ar
e
very
less. T
hat m
e
ans
F
uzzy
-
m
e
ans
te
c
hn
i
qu
e
is b
et
te
r
t
o use t
o detec
t r
oad c
ongestio
n.
The
c
om
par
is
on
betwee
n
th
ese
te
chn
iq
ues
in
te
r
m
s
of
aver
a
ge
com
pu
t
ing
ti
m
e
is
sh
own
in
the
Figure
6.
By
seei
ng
the
se
co
m
par
ison
res
ul
ts,
it
m
ay
be
s
afely
sta
te
d
that
the
cl
us
te
r
f
or
m
at
ion
sp
ee
d
of
-
m
eans
cl
ust
ering
al
go
rithm
is
m
or
e
t
han
FCM
al
gorith
m
and
Fu
z
zy
-
m
eans.
But
in
FCM
a
nd
F
uzzy
-
m
eans
te
chn
i
qu
e
s
eac
h
point
has
a
prob
a
bili
ty
of
belo
ngin
g
t
o
eac
h
c
luster
rathe
r
t
ha
n
belo
ngin
g
t
o
just
on
e
cl
us
te
r
as
in
-
m
eans.
Be
cause
of
this
only
we
can
prefer
fu
zzy
te
c
hn
i
qu
e
s
to
ou
r
prob
le
m
of
tr
aff
ic
congesti
on
det
ect
ion
as
v
e
hic
le
s ar
e
dynam
i
c in
natu
re.
Table
2
.
T
he
A
ver
a
ge
C
om
pu
ti
ng
Tim
e (in
S
econds
) for
-
Me
ans,
FCM
a
nd
Fu
zzy
-
Me
an
s
U
sin
g
the D
at
a
Set,
D
Metho
d
s
k
(
Nu
m
b
er
of
clus
ters)
2
3
4
5
m
eans
0
.06
6
4
0
.06
8
2
0
.10
4
6
0
.12
0
1
FCM
0
.06
7
1
0
.06
9
1
0
.17
8
7
0
.21
6
9
Fu
zzy
-
m
e
an
s
0
.54
1
0
0
.64
9
7
0
.65
5
8
0
.68
3
8
Table
3
.
T
he
D
ist
ance
et
wee
n t
he
Ce
ntre
s in
-
Me
ans
Tec
hniqu
e
Cen
tres of
clu
sters
Clu
ster1
Clu
ster2
Clu
ster3
Clu
ster4
Clu
ster5
Clu
ster1
0
5
5
6
.2
1
1
8
.8
7
4
3
2
.8
836
Clu
ster2
5
5
6
.2
0
4
3
7
.4
6
8
7
6
.7
2
7
9
.8
Clu
ster3
1
1
8
.8
4
3
7
.4
0
7
3
1
4
.1
7
1
7
.2
Clu
ster4
7
4
3
2
.8
6
8
7
6
.7
7
3
1
4
.1
0
6
5
9
6
.9
Clu
ster5
836
2
7
9
.8
7
1
7
.2
6
5
9
6
.9
0
Table
4
.
T
he
D
ist
ance b
et
wee
n
the
Centre
s i
n
FCM
Tech
ni
qu
e
Cen
tres of
clu
sters
Clu
ster1
Clu
ster2
Clu
ster3
Clu
ster4
Clu
ster5
Clu
ster1
0
5
7
2
0
.0
8
4
7
8
.1
3
7
6
6
.5
2
8
4
6
.8
Clu
ster2
5
7
2
0
.0
0
1
4
1
9
8
9
4
8
6
.5
2
8
7
3
.2
Clu
ster3
8
4
7
8
.1
1
4
1
9
8
0
4
7
1
1
.6
1
1
3
2
5
Clu
ster4
3
7
6
6
.5
9
4
8
6
.5
4
7
1
1
.6
0
6
6
1
3
.4
Clu
ster5
2
8
4
6
.8
2
8
7
3
.2
1
1
3
2
5
6
6
1
3
.4
0
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
S
N
:
2502
-
4752
Ind
on
esi
a
n
J
E
le
cEng& C
omp Sci,
V
ol.
13
, No
.
3
,
Ma
rch 2
019
:
884
–
891
890
Table
5
.
T
he
D
ist
ance Bet
wee
n
the
Centre
s i
n
F
uzzy
-
Me
an
s
Tech
nique
Cen
tres of
clu
sters
Clu
ster1
Clu
ster2
Clu
ster3
Clu
ster4
Clu
ster5
Clu
ster1
0
6
0
.38
0
5
0
.02
6
9
3
0
.59
7
2
0
.00
9
5
Clu
ster2
6
0
.38
0
5
0
6
0
.40
7
0
3
0
.70
3
3
6
0
.38
9
8
Clu
ster3
0
.02
6
9
6
0
.40
7
0
0
3
0
.62
4
0
0
.01
7
4
Clu
ster4
3
0
.59
7
2
3
0
.70
3
3
3
0
.62
4
0
0
3
0
.60
6
6
Clu
ster5
0
.00
9
5
6
0
.38
9
8
0
.01
7
4
3
0
.60
6
6
0
Figure
6. Com
par
is
on b
et
wee
n nu
m
ber
of clusters
and a
verage c
om
pu
ta
ti
on
ti
m
e fo
r dif
fe
ren
t t
ec
hn
i
qu
e
s
6.
CONCL
US
I
O
N
Fr
om
our
resul
ts,
we
c
on
cl
ude
that
FCM
and
F
uzzy
-
m
eans
pro
du
ce
cl
os
e
res
ults
t
o
-
m
eans
cl
us
te
rin
g
in
t
he
process
of
de
te
ct
ion
of
c
onge
sti
on
on
a
bus
y
ro
a
d
but
sti
ll
they
require
m
or
e
execu
ti
on
ti
m
e
than
-
m
eans
cl
us
te
rin
g
beca
use
of
the
in
volvem
ent
of
f
uz
zy
m
easur
es
c
al
culat
ion
s
in
the
al
gorithm
.
And,
ou
t
of
t
hese
f
uz
zy
te
chn
iq
ues
,
F
uzzy
-
m
eans
is
bette
r
as
t
he
dista
nce
bet
ween
cl
us
te
r
c
enters
is
le
sse
r
than
FCM
techn
i
que g
ivi
ng the
id
ea that
c
ongest
ion
is
m
or
e pro
m
inently
d
et
ect
ed
in
Fu
zzy
-
m
eans.
REFERE
NCE
[1]
Dhan
y
a
P.
M,
Ananth
S.
Eff
icient
Tr
aff
ic
Co
ngesti
on
Detect
i
on
Method
i
n
Vane
t.
Int
ernational
Journal
for
Technol
ogi
cal
R
ese
arch
in
Eng
in
ee
ring.
2013;
1(
3).
[2]
Jos
hi
A,
Mishra
D.
Revi
ew
of
Tr
aff
ic
D
ensity
An
aly
s
is
Techni
qu
e
s.
Inte
rnationa
l
Journal
of
Ad
va
nce
d
R
ese
arch
i
n
Computer
and
C
omm
unic
ati
on
E
ngine
ering
.
201
5;
4
(7)
.
[3]
Ahm
ed
H,
Pierre
S,
Quinte
ro
A.
A
Flexi
ble
Te
st
bed
Archi
t
ec
tur
e
for
VA
NET.
Vehic
ular
Comm
unic
ati
on
.
2017;
9:
15
-
126.
[4]
Yoo
H,
Kim
D.
Repe
tition
-
b
ase
d
coope
ra
ti
ve
bro
ad
ca
st
ing
for
ve
hic
ul
ar
ad
-
hoc
n
et
works
.
Comput.
Comm
un.
2011
;
34
(5):1870
-
188
2.
[5]
Fußler
H,
Mau
ve
M,
Hart
enst
ei
n
H,
Kasem
a
nn
M,
Vollmer
D.
Locati
on
-
b
ased
routing
for
ve
hic
ular
ad
-
hoc
net
works.
Mobi
l
e
Com
puti
ng
an
d
Com
m
unic
at
io
ns Re
vie
w,
SIG
MO
BILE
,
New
York,
US
A,
200
2.
[6]
W
ei
Q,
Yang
B
.
Adapt
abl
e
v
eh
ic
l
e
detec
ti
on
an
d
spee
d
est
imation
for
cha
ng
ea
b
le
urba
n
tra
ff
ic
with
ani
sotrop
ic
m
agne
to
resist
iv
e
sensors
.
I
EE
E
Sensors
Journal.
2017;
17(7)
:
20
21
-
2028
.
[7]
Sen
R,
Siria
h
P,
Raman
B.
RoadSoundSe
nse:
A
cousti
c
sensing
based
road
con
gesti
on
monitori
ng
in
dev
el
opin
g
regions
.
Sensor,
Mesh
and
Ad
Hoc
Com
m
unic
at
ions
an
d
Netw
orks,
SECON
’2
011,
Salt
L
ake
Cit
y
,
Utah
,
US
A,
2011:
125
–
133.
[8]
Ro
y
A,
Gal
e
N,
Hong
L.
Auto
m
at
ed
tr
aff
ic
surveil
l
ance
using
fusion
of
Dopp
le
r
r
ada
r
and
vi
deo
informat
ion
.
Mathe
mati
cal
an
d
Computer
Mo
del
li
ng
.
2011
;
54
:
531
–
543.
[9]
Manda
l
K,
Sen
A,
Chakra
bort
y
A,
Ro
y
S,
B
at
ab
y
a
l
S,
Band
y
op
a
dh
y
a
y
S
.
Road
t
raffi
c
cong
esti
on
monitoring
and
measur
eme
nt
using ac
ti
v
e
rfid
an
d
gsm
te
chnol
og
y
.
IE
EE
Int
el
l
igent
Tra
nsport
at
io
n
S
y
stems
,
ITSC’11,
W
ashingt
on
DC,
US
A.
2011:
1375
–
1379.
[10]
Jiang
G,
Niu
S,
Chang
A,
Men
g
Z,
Zha
ng
C
.
Aut
omatic
traff
i
c
cong
esti
on
id
e
nti
ficati
on
meth
od
of
e
xpre
ss
way
based
on
gain
a
mplif
ie
r
the
ory
.
Advanc
ed
Com
p
ute
r
Con
trol
,
IC
ACC’10,
Shen
yang,
Ch
ina,
201
0;2:
648
–
651
.
Evaluation Warning : The document was created with Spire.PDF for Python.
Ind
on
esi
a
n
J
E
le
cEng& C
omp Sci
IS
S
N:
25
02
-
4752
Tra
ff
ic
con
gestion
detect
io
n
i
n a city
usi
ng c
lusteri
ng tech
ni
qu
es
in
V
ANE
Ts
(
Anita
Moh
an
ty
)
891
[11]
Zhu
W
,
Bart
h
M.
Ve
hi
cle
traj
ec
tory
-
based
road
ty
pe
and
co
ngesti
on
rec
ogn
it
ion
using
w
a
v
el
e
t
analy
sis
.
IE
EE
Inte
lligen
t
Tr
ans
porta
ti
on
S
y
s
te
m
s Confe
ren
c
e, I
T
SC
’06,
Paris,
Franc
e
,
2006
:
879
-
884.
[12]
Vaqa
r
S,
Basir
O.
Tra
ff
ic
p
at
t
er
n
det
e
ct
ion
in
a
par
tially
depl
o
y
ed
vehicul
ar
a
d
hoc
ne
twork
of
vehi
c
le
s.
IEEE
Wirel
ess Comm
unic
ati
ons
.
2009;
16
(6)
:
40
–
46
.
[13]
Li
u
C,
Chig
an
C.
Struct
ure
-
le
s
s
mess
age
aggr
egat
ion
(
SLMA):
Re
l
iabl
y
and
ef
fici
ent
l
y
improve
in
formation
prec
ision
and
c
e
rtaint
y
for VAN
ETs
.
IEEE
GLO
BECOM’10,
Mi
ami,
FL
,
US
A.
2
010:
1
–
6.
[14]
Bauz
a
R,
Go
za
lv
ez
J.
Tra
f
fic
conge
st
ion
det
ection
in
la
rge
-
sc
a
le
sce
nar
ios
using
vehi
cle
-
to
-
veh
ic
l
e
comm
unic
at
ions
.
Journal
of
Networ
k
and
Comp
ute
r A
pp
li
ca
ti
on
s
.
2013;
36
:
129
5
-
1307.
[15]
EmreCe
le
b
i
M,
Hass
an
Kingra
vi
A,
Patri
ci
o
Vel
a
A.
A
compara
tive
stud
y
of
eff
i
c
ie
nt
initi
al
i
za
t
io
n
m
et
hods
for
th
e
k
-
m
ea
ns c
lust
ering a
lgor
it
hm
.
Expert
Syst
ems wi
t
h
Applications
.
2
013
;40
(1):
200
-
210
.
[16]
Kum
ar
Vishal,
Mishra
Shail
end
ra,
Chand
Narot
t
am.
Applic
ation
s
of
V
AN
ET
s:
P
rese
nt
&
Future
.
Comm
unic
ati
ons
and
Net
work
,
20
13;
5:
12
-
15.
[17]
Te
rroso
-
Saenz
F
,
Valde
s
-
Vel
a
M,
Sotom
a
y
or
-
Marti
ne
z
C,
To
l
edo
-
Moreo
R,
Gom
ez
-
Skarm
et
a
A.
A
coope
ra
t
ive
appr
oac
h
to
tr
af
fic
conge
st
ion
det
e
ct
ion
with
complex
eve
nt
proc
essing
and
VA
NET.
IEE
E
Tr
ansacti
ons
on
Inte
lligen
t
Tr
anspor
tat
ion
S
yste
m
s.
2012;
13
(2):
914
–
929.
[18]
htt
ps://
ww
w.sem
ant
i
csc
hola
r
.
org/
pape
r/R
el
i
abl
e
-
Routi
ng
-
in
-
Vehi
cul
ar
-
Adhoc
-
N
etw
ork
-
Us
ing
-
(
-
Rao
-
Kar
y
ems
et
t
y
/a
0d63df28b68b
b9ebd46e
0311a
14dc1958c
0917
3/fi
gure
/0
[19]
Bauz
a
R
,
Goz
alvez
J,
San
chez
-
Soriano
J.
Road
traff
ic
conge
st
io
n
det
e
ct
ion
thro
ugh
coope
rat
i
ve
ve
hi
cle
-
to
-
v
ehic
le
communic
ati
ons
.
IEE
E
Confer
en
ce
on
Loc
a
l
Com
pute
r
Networks,
LCN’10,
Denve
r,
Colora
do
,
US
A,
2010:
606
–
612.
[20]
https://
ww
w.ope
ns
tr
ee
tmap
.org/#
map
=16/20.
3147/85.8161
[21]
G
hosh
S,
Dube
y
S
K.
Com
par
at
ive
Anal
y
sis
of
K
-
Mea
ns
and
F
uzzy
CMe
a
nsA
lgori
thms
.
Inte
rnational
Journal
of
Adv
anc
ed
Comp
ute
r Sc
ie
nc
e
and
Applications
.
20
13;
4(4):
35
-
39.
[22]
Har
-
Pele
d
S,
Sadri
B
.
How
f
ast
is
th
e
k
-
m
ea
ns
Method
.
ACM
SIAM
Sy
m
posium
on
Dis
cre
t
e
Algor
it
hm
s
,
Vanc
ouver
.
200
5.
[23]
Bez
dek
J.
C
.
Pa
t
te
rn
R
ec
ogni
ti
on
with
Fuz
z
y
Obj
ec
t
ive
Func
ti
on
Algorit
hm
s.
Ne
w York:
Plenum Press
,
1981.
[24]
Yong
Y,
Chon
gxun
Z,
Pan
L.
A
Nove
l
F
uzzy
C
-
Me
ans
Cluste
ring
Al
gorit
hm
for
Im
age
Thre
sholdi
ng.
Me
asur
eme
nt
S
c
ie
nc
e
R
evie
w
.
20
04;
4(1).
[25]
Chang
Chih
-
Tang,
La
i
Jim
Z
C,
Jeng
Mu
-
Der.
A
Fuzz
y
K
-
me
ans
Cluste
rin
g
Algorit
hm
using
Cluste
r
Center
Displac
ement
.
J
ournal
of information
S
ci
en
ce an
d
Engi
n
ee
ring
.
2
011;
27:
995
-
10
09.
[26]
Math
works
htt
p
:/
/www
.
m
at
hworks
.
com
Evaluation Warning : The document was created with Spire.PDF for Python.