Int
ern
at
i
onal
Journ
al of Ele
ctrical
an
d
Co
mput
er
En
gin
eeri
ng
(IJ
E
C
E)
Vo
l.
15
,
No.
1
,
Febr
uary
20
25
, pp.
12
29
~
1241
IS
S
N:
20
88
-
8708
, DO
I: 10
.11
591/ij
ece.v
15
i
1
.
pp
1229
-
12
41
1229
Journ
al h
om
e
page
:
http:
//
ij
ece.i
aesc
or
e.c
om
Multi
-
ob
jective
optimiz
ed
ta
sk
sch
eduli
ng in cogniti
ve inte
rnet
of vehi
cles:
towards
ener
gy
-
effici
ency
M.
Divy
as
hre
e
1,2
,
H. G.
R
anga
r
aju
3
,
C. R
.
Revan
na
1
1
Dep
artm
en
t of
E
l
ectron
ics an
d
Co
m
m
u
n
icatio
n
E
n
g
in
eering
,
Go
v
ernment
Sr
i
Krish
n
araje
n
d
r
a Silver Ju
b
ilee
Te
ch
n
o
lo
g
ical I
n
stitute,
Ben
g
alu
ru,
Aff
iliat
ed
to Visv
esv
araya T
echn
o
lo
g
ical Univ
ersity
,
Belg
au
m
,
Ind
ia
2
Dep
artm
en
t of
E
l
ectron
ics an
d
Co
m
m
u
n
icatio
n
E
n
g
in
eering
,
RV
Ins
titu
te of T
e
ch
n
o
lo
g
y
an
d
M
an
ag
em
en
t,
Be
n
g
alu
ru, I
n
d
ia
3
Dep
artm
en
t of
E
l
ectron
ics an
d
Co
m
m
u
n
icatio
n
E
n
g
in
eering
,
Go
v
ernment
E
n
g
in
eering
Co
lleg
e,
KR Pet
e
,
Ind
ia
Art
ic
le
In
f
o
ABSTR
A
CT
Art
ic
le
history:
Re
cei
ved
A
pr
5,
2024
Re
vised A
ug 29, 2
024
Accepte
d Oct
1,
2024
The
r
ise
of
inte
ll
ige
n
t
and
c
on
nec
t
ed
veh
icles
has
le
d
to
new
vehicula
r
appl
i
ca
t
ions,
bu
t
vehicl
e
co
mp
uti
ng
c
apa
b
il
i
ti
e
s
rem
ai
n
limit
e
d.
Mobil
e
edge
com
pu
ti
ng
(MEC)
ca
n
mi
t
iga
t
e
thi
s
by
off
loa
ding
co
mput
a
ti
on
ta
sks
to
the
ne
twork's
edge.
How
ev
er,
l
im
ited
co
mput
ational
ca
p
ac
i
ti
es
in
v
ehi
c
le
s
l
ea
d
to
i
ncr
ea
sed
l
at
en
cy
and
ene
rgy
cons
umpt
ion.
To
add
ress
thi
s,
roa
dside
uni
ts
(RSU
s)
with
cl
oud
serve
rs,
kn
own
as
edge
c
omput
in
g
devi
c
es
(ECDs),
ca
n
be
exp
and
ed
to
provide
e
ner
gy
-
eff
i
cient
s
che
dul
ing
for
ta
sk
com
pu
t
at
ion
.
A
new
e
ner
gy
-
eff
i
cient
s
che
dul
ing
meth
od
ca
l
le
d
mul
ti
-
ob
jecti
v
e
opti
mization
en
erg
y
co
mput
a
tion
(MO
EC)
is
proposed,
base
d
on
mul
t
i
-
obje
c
ti
ve
par
ti
c
l
e
sw
arm
optimi
za
t
ion
(MO
PS
O
)
to
r
educe
ECDs'
ene
rgy
usage
and
e
xec
ut
ion
t
ime.
Simul
at
ion
re
sults
using
MA
TL
AB
show
tha
t
MO
EC
ca
n
b
alanc
e
th
e
tra
d
e
-
off
b
et
w
ee
n
ene
rgy
usage
and
ex
ec
u
ti
on
t
ime,
leadin
g
to more
eff
i
cie
nt
offlo
adi
ng
.
Ke
yw
or
ds:
Ed
ge
c
ompu
ti
ng
dev
ic
es
M
obil
e e
dg
e
c
ompu
ti
ng
M
ulti
-
ob
je
ct
iv
e opti
miza
ti
on
M
ulti
-
ob
je
ct
iv
e p
a
rtic
le
swarm
-
opti
miza
ti
on
Road
side
unit
This
is an
open
acc
ess arti
cl
e
un
der
the
CC
BY
-
SA
l
ic
ense
.
Corres
pond
in
g
Aut
h
or
:
M
.
Divyas
hree
Dep
a
rtme
nt of
Ele
ct
ro
nics
and C
om
m
unic
at
ion
En
gin
ee
rin
g,
G
ov
e
r
nm
e
nt Sr
i
Kr
is
hnaraj
endra
Sil
ver
Jubil
ee
Tech
no
l
og
ic
al
In
sti
tute
K.
R.
Ci
rcle
-
5600
01, Be
ng
al
uru, Ka
rn
at
a
ka
, In
dia
Emai
l:
m.
di
vyashr
ee
4@g
mail
.co
m
1.
INTROD
U
CTION
With
t
he
prolif
erati
on
of
intel
li
gen
t
tra
nsport
at
ion
s
ys
te
m
s,
the
vo
l
um
e
of
data
produce
d
by
veh
ic
le
s
and
thei
r
as
soc
ia
te
d
se
ns
ors
ha
s
i
ncr
ease
d
dramat
ic
al
ly.
Nev
e
rtheless
,
m
os
t
ve
hicle
s
la
ck
the
r
eq
uisit
e
capab
il
it
y
for
local
data
pro
cessi
ng
an
d
stora
ge.
Co
ns
e
quently
,
c
omp
ut
at
ion
al
res
ponsi
bili
ti
es
need
to
be
trans
ferred
t
o
distant
cl
oud
data
cente
rs.
This
is
ac
hiev
ed
th
rou
gh
r
oa
ds
ide
unit
s
ut
il
iz
ing
the
ve
hi
cl
e
-
to
-
infr
a
struct
ur
e
(V2I)
c
onnecti
on
m
ode
[1]
.
Ther
e
are
s
ome
draw
bac
ks
t
o
t
his
a
ppro
ac
h,
in
t
he
i
nter
net
of
veh
ic
le
s
(Io
V)
netw
ork w
hich
experie
nces
hi
gh tra
ns
missi
on
delays
and i
nc
onsist
ent con
necti
ons
[
2],
[3]
.
Roadsi
de
unit
s
(RS
Us)
ar
e
st
rategica
ll
y
pla
ced
al
ongs
id
e
ro
a
d
netw
orks
an
d
high
ways
to
pro
vid
e
commu
nicat
io
n
ser
vices
for
co
nnect
ed
ve
hi
cl
es.
T
o
e
nhance
t
he
ef
fec
ti
ven
ess
of
co
mputat
ion
al
ta
sk
s
f
or
veh
ic
le
s
in
co
gn
it
ive
i
nterne
t
of
veh
ic
le
s
(
CIoV),
the
fun
ct
ion
of
RS
Us
has
been
exte
nd
e
d
to
ser
ve
as
ed
ge
com
pu
ti
ng
de
vices
(ECD
s)
,
offe
rin
g
proc
essing
a
nd
st
orage
ca
pab
il
it
ie
s
[4]
.
The
m
ob
il
e
e
dge
co
mputi
ng
(MEC)
[5]
–
[
13]
is
an
ap
proac
h
dep
l
oy
s
cl
ou
d
se
rv
ic
es
cl
ose
r
to
t
he
ra
dio
acce
ss
net
work’s
e
dg
e
,
facil
it
at
ing
the
c
ompu
ta
ti
on
offloa
ding
[
14]
of
ta
s
ks
t
o
ne
arby
ECDs
,
l
ocated
in
cl
os
e
vicinit
y
to
the
ve
hicle
s,
i
ns
te
ad
of
relyin
g
on
dist
ant
cl
oud
i
nfr
ast
ru
ct
ure
[15]
.
I
n
the
fr
a
me
work
of
the
C
IoV,
as
sig
ning
ta
sk
s
to
ECD
s
can
ind
ee
d
en
ha
nc
e
the
qual
it
y
of
the
dri
ve
r's
e
xp
e
rien
ce
by
a
ddressi
ng
dela
ys
in
tra
ns
mis
sion
a
nd
im
pr
ov
i
ng
connecti
on
sta
bili
ty
[
16]
.
H
oweve
r,
wh
e
n
offloa
de
d
co
mputat
ion
al
ta
sk
s
accomm
odat
ed
in
EC
Ds,
it
is
cru
ci
al
to
pri
ori
ti
ze
th
e
li
mit
ed
res
ources
of
t
hese
de
vices.
Sp
ec
ific
al
ly,
w
he
n
operati
ng
on
an
EC
D,
t
her
e
is
a
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
S
N
:
2088
-
8708
In
t J
Elec
&
C
omp E
ng,
V
ol.
15
, No
.
1
,
Febr
uary
20
25
:
1229
-
1241
1230
necessit
y
to
re
stric
t
the
qu
a
nt
it
y
of
c
on
c
urre
ntly
act
ive
ta
s
ks
.
In
s
uc
h
cas
es,
there
mig
ht
be
instanc
es
wh
e
re
com
pu
ti
ng
ta
s
ks
withi
n
t
he
area
of
co
ve
rag
e
of
one
ECD
necessit
a
te
trans
fer
t
o
a
di
ff
e
ren
t
E
CD
for
processi
ng.
En
han
ci
ng
al
l
EC
Ds'
res
pons
e
ti
mes
a
nd
lo
weri
ng
their
e
nerg
y
us
a
ge
are
c
r
ucial
f
or
facil
it
at
ing
the comp
utati
on
of offl
oad
i
ng b
et
wee
n EC
D
s.
W
h
e
n
a
c
o
n
n
e
c
t
e
d
v
e
h
i
c
l
e
s
e
n
d
s
i
n
f
o
r
m
a
t
i
o
n
t
o
a
n
E
C
D
,
t
h
e
E
C
D
c
a
l
c
u
l
a
t
e
s
t
h
e
e
n
e
r
g
y
r
e
q
u
i
r
e
d
t
o
p
r
o
c
e
s
s
t
h
e
i
n
f
o
r
m
a
t
i
o
n
a
n
d
a
l
s
o
c
a
l
c
u
l
a
t
e
s
t
h
e
e
n
e
r
g
y
r
e
q
u
i
r
e
d
b
y
n
e
i
g
h
b
o
r
i
n
g
E
C
D
s
t
o
p
r
o
c
e
s
s
i
t
.
B
a
s
e
d
o
n
t
h
i
s
c
a
l
c
u
l
a
t
i
o
n
,
t
h
e
E
C
D
d
e
c
i
d
e
s
w
h
i
c
h
n
e
i
g
h
b
o
r
i
n
g
E
C
D
s
h
o
u
l
d
p
r
o
c
e
s
s
t
h
e
i
n
f
o
r
m
a
t
i
o
n
w
i
t
h
l
e
s
s
e
n
e
r
g
y
c
o
n
s
u
m
p
t
i
o
n
.
T
h
e
i
n
f
o
r
m
a
t
i
o
n
i
s
t
h
e
n
r
e
d
i
r
e
c
t
e
d
t
o
t
h
e
s
e
l
e
c
t
e
d
E
C
D
f
o
r
p
r
o
c
e
s
s
i
n
g
[17]
.
T
h
i
s
l
o
a
d
-
b
a
l
a
n
c
i
n
g
t
e
c
h
n
i
q
u
e
[
1
8
]
,
[
1
9
]
h
e
l
p
s
w
i
t
h
t
a
s
k
s
c
h
e
d
u
l
i
n
g
[
2
0
]
,
[
2
1
]
b
y
d
i
s
t
r
i
b
u
t
i
n
g
t
a
s
k
s
a
c
r
o
s
s
t
h
e
n
e
t
w
o
r
k
,
p
r
e
v
e
n
t
i
n
g
a
n
y
s
i
n
g
l
e
E
C
D
f
r
o
m
b
e
c
o
m
i
n
g
o
v
e
r
l
o
a
d
e
d
.
B
y
a
v
o
i
d
i
n
g
o
v
e
r
l
o
a
d
i
n
g
,
E
C
D
s
c
a
n
a
c
h
i
e
v
e
h
i
g
h
e
r
o
p
e
r
a
t
i
o
n
a
l
e
f
f
i
c
i
e
n
c
y
a
n
d
c
o
n
s
u
m
e
l
o
w
e
r
a
m
o
u
n
t
s
o
f
e
n
e
r
g
y
.
E
C
D
s
c
a
n
f
u
n
c
t
i
o
n
w
i
t
h
i
n
c
r
e
a
s
e
d
e
f
f
i
c
i
e
n
c
y
,
r
e
d
u
c
e
d
e
n
e
r
g
y
c
o
n
s
u
m
p
t
i
o
n
,
a
n
d
e
x
t
e
n
d
e
d
l
i
f
e
t
i
m
e
.
T
h
i
s
,
i
n
t
u
r
n
,
h
e
l
p
s
t
o
o
p
t
i
m
i
z
e
t
h
e
o
v
e
r
a
l
l
n
e
t
w
o
r
k
’
s
e
n
e
r
g
y
u
s
a
g
e
,
m
a
k
i
n
g
i
t
m
o
r
e
s
u
s
t
a
i
n
a
b
l
e
a
n
d
c
o
s
t
-
e
f
f
e
c
t
i
v
e
.
T
h
e
k
e
y
c
o
n
t
r
i
b
u
t
i
o
n
p
r
o
v
i
d
e
d
b
y
t
h
i
s
p
a
p
e
r
i
s
i
m
p
l
e
m
e
n
t
a
t
i
o
n
o
f
m
u
l
t
i
-
o
b
j
e
c
t
i
v
e
p
a
r
t
i
c
l
e
s
w
a
r
m
o
p
t
i
m
i
z
a
t
i
o
n
(
M
O
P
S
O
)
t
o
a
c
h
i
e
v
e
m
u
l
t
i
-
o
b
j
e
c
t
i
v
e
o
p
t
i
m
i
z
a
t
i
o
n
,
l
e
a
d
i
n
g
t
o
d
e
c
r
e
a
s
e
d
e
n
e
r
g
y
u
s
a
g
e
i
n
E
C
D
s
a
n
d
d
e
c
r
e
a
s
e
e
x
e
c
u
t
i
on
t
i
m
e
f
o
r
c
o
m
p
u
t
i
n
g
t
a
s
k
s
.
T
h
i
s
n
o
v
e
l
a
p
p
r
o
a
c
h
a
d
d
r
e
s
s
e
s
t
h
e
e
x
i
s
t
i
n
g
g
a
p
s
i
n
e
n
e
r
g
y
e
f
f
i
c
i
e
n
t
c
o
m
p
u
t
a
t
i
o
n
f
o
r
c
o
m
p
u
t
a
t
i
o
n
a
l
t
a
s
k
a
l
l
o
c
a
t
i
o
n
i
n
C
I
o
V
,
o
f
f
e
r
i
n
g
a
m
o
r
e
e
f
f
i
c
i
e
n
t
a
n
d
s
u
s
t
a
i
n
a
b
l
e
s
o
l
u
t
i
o
n
f
o
r
i
n
t
e
l
l
i
g
e
n
t
t
r
a
n
s
p
o
r
t
a
t
i
o
n
s
y
s
t
e
m
s
.
N
u
m
e
r
o
u
s
s
t
u
d
i
e
s
h
a
v
e
e
x
p
l
o
r
e
d
e
n
e
r
g
y
-
e
f
f
i
c
i
e
n
t
s
t
r
a
t
e
g
i
e
s
i
n
e
d
g
e
c
o
m
p
u
t
i
n
g
/
M
E
C
i
m
pl
e
m
e
n
t
a
t
i
o
n
,
w
i
t
h
a
f
o
c
u
s
o
n
t
a
s
k
s
c
h
e
d
u
l
i
n
g
/
o
f
f
l
o
a
d
i
ng
i
n
r
e
l
a
t
e
d
p
a
p
e
r
s
.
N
i
n
g
e
t
a
l
.
[17]
p
r
o
p
o
s
e
d
a
n
M
EC
-
ena
bled
energ
y
-
e
ff
ic
ie
nt
sche
du
li
ng
(
M
EE
S)
meth
od
in
I
oV,
w
hic
h
in
cl
udes
del
ay
est
imat
io
n,
energ
y
c
onsumpti
on
est
imat
ion
,
ta
s
k
sc
hedulin
g,
proces
sin
g,
a
nd
resu
lt
f
eedi
ng
back.
T
he
fr
a
m
ewor
k
ai
ms
t
o
minimi
ze
the
e
nerg
y
consu
mp
ti
on
of
RSU
s
w
hile
consi
der
i
ng
ta
sk
la
te
nc
y
c
on
s
trai
nts.
T
hey
de
velo
ped
a
he
ur
ist
ic
al
gorith
m
that
jointl
y
c
onside
rs
ta
s
k
sc
he
du
li
ng
a
mon
g
MEC
ser
vers
a
nd
dow
nlin
k
e
ne
rgy
c
ons
umpti
on
of
RS
Us
.
The
p
er
forma
nce
e
valuati
ons
de
m
onstrat
ed
the
eff
ect
ive
ness
of
t
he
f
rame
w
ork
in
te
rms
of
ene
r
gy
c
ons
umpti
on,
la
te
ncy
, a
nd ta
sk
blo
c
king
po
ssibil
it
y.
Liu
et
al.
[22
]
i
ntrod
uce
d
t
wo
c
ompu
ta
ti
on
offloa
ding
al
gorithms,
bi
nary
offl
oad
i
ng
a
nd
pa
rtial
offloa
ding,
i
n
order
t
o
ha
ndle
the
issue
of
ta
sk
s
bein
g
di
vide
d
into
in
div
isi
ble
an
d
div
isi
bl
e
ta
sk
s.
T
he
bi
nary
offloa
ding
met
hod
tra
nsfers
t
he
e
ntire
ta
sk
to
the
M
EC
serv
e
r
a
nd
us
e
s
an
e
nh
a
nce
d
meth
od
for
uppe
r
confide
nce
bo
unds
to
ch
oose
the
be
st
offloa
ding
sit
e.
T
he
par
ti
al
offloa
d
ing
al
go
rithm
di
vid
es
co
mp
le
x
ta
sk
s
into
ti
me
sl
ots
proces
sed
by
dif
fer
e
nt
M
E
C
ser
ver
s
,
us
ing
the
Q
-
le
ar
ni
ng
al
gorithm
to
est
ablis
h
th
e
most
eff
ect
ive
offlo
adin
g
strat
eg
y.
The
ou
tc
om
es
of
t
he
simulat
ion
im
ply
that
the
bin
a
r
y
offl
oad
i
ng
al
gorith
m
has
lowe
r
dela
y
c
os
t
an
d
e
ne
rgy
us
e
durin
g
processi
ng
t
he
com
puta
ti
on
a
l
intensive
ta
s
ks
,
w
hile
the
par
ti
al
offloa
ding al
gorithm si
gn
i
ficantl
y
im
pro
ves r
eal
-
ti
me p
e
rformance a
nd c
on
serv
e
s m
obil
e terminal
e
ne
rgy
.
Xu
e
t
a
l
.
[23]
pr
op
o
se
d
an
edg
e
co
mp
u
t
i
ng
en
ab
l
ed
comp
u
t
a
t
i
on
o
ff
lo
adi
ng
t
e
chn
iqu
e
c
a
l
l
e
d
edg
e
c
om
pu
t
ing
of
f
lo
ad
in
g
(
E
CO)
.
I
t
r
edu
c
es
com
pu
t
ing
t
a
sk
en
erg
y
us
ag
e
and
ex
e
cu
t
i
on
t
im
e
w
h
i
l
e
add
re
s
s
in
g
pr
i
v
ac
y
con
f
l
i
c
t
s
.
F
ir
s
t
,
to
a
cqu
i
r
e
th
e
r
ou
t
ing
ve
h
i
c
l
es
fr
om
t
h
e
or
i
g
in
ve
h
i
c
l
e
in
wh
i
ch
t
he
c
om
pu
t
i
ng
task
i
s
l
oc
a
t
ed
to
t
h
e
de
s
t
in
a
t
i
on
v
eh
i
c
l
e
,
v
eh
i
c
le
-
to
-
v
eh
i
c
l
e
(V
2
V)
co
mmu
n
i
c
a
t
io
n
-
b
a
sed
ro
u
t
i
ng
f
or
a
v
eh
i
c
l
e
i
s
d
ev
e
lop
ed
.
T
h
en,
non
-
d
om
i
n
a
t
ed
so
r
t
ing
g
en
e
t
i
c
a
l
go
r
i
t
hm
II
(N
SG
A
-
II
)
i
s
u
t
i
liz
e
d
to
a
ch
i
ev
e
th
e
mu
lti
-
ob
j
ec
t
i
ve
op
t
i
m
i
za
t
i
on
.
Su
b
s
equ
en
t
exp
er
i
m
en
t
a
l
ev
a
lu
a
t
i
on
s
v
er
i
fy
t
he
ef
f
i
c
i
enc
y
and
e
ff
e
c
t
iv
en
e
s
s
of
E
CO.
Be
hb
e
ha
ni
et
al
.
[
24]
de
vel
oped
a
mixe
d
i
nteger
li
nea
r
pro
gr
a
mmin
g
(
M
I
LP)
m
od
el
that
opti
mize
s
distrib
ution
of
proces
sin
g
de
man
ds
t
hat
c
omp
rise
veh
ic
le
s,
c
ompu
ti
ng
i
n
the
e
dg
e
al
s
o
in
t
he
cl
oud.
T
he
model
inte
nds
to
le
sse
n
powe
r
us
a
ge,
a
nd
c
ompa
red
to
co
nventio
nal
cl
ouds
,
the
fin
ding
s
sho
w
powe
r
savin
gs
ov
e
r
70%
–
90%
for
lo
w
w
orkl
oad
s
.
H
owe
ver,
f
or
me
dium
an
d
la
rg
e
de
man
d
siz
es,
t
he
res
ults
ind
i
cat
e
a
li
mit
ed
am
ount
of
cl
oud use
due
t
o
ca
pacit
y
l
imi
ta
ti
on
s o
n
t
he
ve
hicular
a
nd
e
dge
no
des,
le
adin
g
t
o
20%
-
30%
powe
r
sa
vings.
The
st
ru
ct
ur
e
of
t
his
pap
e
r
i
s
orga
nized
as
fo
ll
ows
.
Sect
i
on
2
prese
nts
the
met
hodolo
gy,
pro
vid
i
ng
the
mathe
mati
cal
modeli
ng
for
prob
le
m
form
ulati
on,
the
sc
heduling
co
mpu
ta
ti
on
f
or
CI
oV
i
n
e
dg
e
com
pu
ti
ng
bas
ed
on
m
ulti
-
ob
je
ct
ive
pa
rtic
le
s
war
m
opti
miza
ti
on
,
a
nd
the
sim
ulati
on
e
nvir
onment.
Sec
ti
on
3
pr
ese
nts a
nd
discuss
es
the
res
ults. Fi
nally,
se
ct
ion
4 pr
ese
nt
s conclusi
on a
nd futu
re s
c
op
e.
2.
METHO
D
This
sect
ion
a
nalyzes
mat
he
mati
cal
models
f
or
pro
blem
f
ormulat
io
n,
in
cl
ud
in
g
offloa
ding
ti
me
a
nd
energ
y
us
age
est
imat
ion
.
It
discusse
s
a
sc
hedulin
g
c
ompu
ta
ti
on
strat
e
gy
f
or
CI
oV
in
ed
ge
c
omp
uting
env
i
ronme
nts,
base
d
on
MO
PSO.
T
he
strat
egy
incl
ud
e
s
V2V
tra
ns
miss
ion
te
ch
niques
f
or
e
ff
ic
ie
nt
offloa
d
path
ac
quisi
ti
on
a
nd comp
uta
ti
on
sc
he
du
li
ng
, v
al
idate
d
t
hro
ugh
a
sim
ulati
on setu
p.
Evaluation Warning : The document was created with Spire.PDF for Python.
In
t J
Elec
&
C
omp E
ng
IS
S
N:
20
88
-
8708
Multi
-
obje
ct
iv
e opti
mized
task
sch
e
duli
ng in c
ogniti
ve inte
r
net o
f ve
hicle
s:
towar
ds
…
(
M
.
Divy
ashree
)
1231
2
.
1
.
F
o
r
m
u
l
a
t
i
o
n
o
f
p
r
o
b
l
e
m
s
a
n
d
t
h
e
s
y
s
t
e
m
m
o
d
e
l
T
h
i
s
s
u
b
s
e
c
t
i
on
p
r
e
s
e
n
t
s
a
m
o
d
e
l
f
o
r
C
I
o
V
i
n
c
l
o
u
d
-
e
d
g
e
c
o
m
p
u
t
i
n
g
s
y
s
t
e
m
s
,
a
d
d
r
e
s
s
i
n
g
a
m
u
l
t
i
-
o
b
j
e
c
t
i
v
e
o
p
t
i
m
i
z
a
t
i
o
n
i
s
s
u
e
o
f
c
o
m
p
u
t
a
t
i
o
n
s
c
h
e
d
u
l
i
n
g
.
I
n
c
l
o
u
d
-
e
d
g
e
c
o
m
p
u
t
i
n
g
,
F
i
g
u
r
e
1
d
e
p
i
c
t
s
a
c
o
m
m
u
n
i
c
a
t
i
o
n
s
t
r
u
c
t
u
r
e
f
o
r
C
I
o
V
[
2
5
]
.
T
a
b
l
e
1
c
o
n
t
a
i
n
s
e
s
s
e
n
t
i
a
l
t
e
r
m
s
a
n
d
t
h
e
i
r
c
o
r
r
e
s
p
o
n
d
i
n
g
d
e
s
c
r
i
p
t
i
on
s
.
Figure
1. Com
munica
ti
on str
uctu
re fo
r
CI
oV
Table
1.
Def
i
ne
s the
fo
ll
owin
g key t
erms
Key
ter
m
Definitio
n
Nu
m
b
er
o
f
E
CDs
Co
llectio
n
of E
CD
,
wh
ere
=
{
1
,
2
,
.
.
,
}
Co
llectio
n
of RSU
,
wh
ere
=
{
1
,
2
,
.
.
,
}
Co
llectio
n
of
serv
ers
,
wh
er
e
=
{
1
,
2
,
.
.
,
}
The to
tal cou
n
t of
v
eh
icles
Co
llectio
n
of
v
eh
icles
,
wh
ere
=
{
1
,
2
,
.
.
,
}
All
-
serv
er
capacity
Co
m
p
u
tin
g
task
,
wh
ere
=
{
1
,
2
,
.
.
,
}
The n
th
co
m
p
u
tatio
n
al task
in
Req
u
ested
qu
an
tit
y
of the
reso
u
rce u
n
its
Tim
e
con
su
m
ed
to im
p
le
m
en
t
Bas
elin
e energy
us
ag
e f
o
r
all
the serv
ers
Energy
us
ag
e by
r
eso
u
rce
u
n
its
em
p
l
o
y
ed
Energy
us
ag
e by
un
em
p
lo
y
ed
r
eso
u
rce
u
n
its
Total en
ergy
us
ag
e by
all
th
e serve
rs
2.1.1
.
M
od
el
of ex
ecuti
on
ti
me
It
is
cru
ci
al
to
consi
der
exec
ution
ti
me
,
fee
db
ac
k
per
i
od
f
or
se
ndin
g
bac
k
the
e
xec
utio
n's
fi
nd
i
ngs
back
to
ve
hicle
,
an
d
ve
hicle
t
o
EC
D
offloa
di
ng
ti
me
w
hile
us
i
ng
a
ve
hicle
to
pe
rform
c
ompu
ta
ti
ons
[
23]
.
As
the
ve
hicle
s
move
al
on
gs
id
e
the
r
oa
d,
th
ey
tra
ver
se
m
ulti
ple
ECDs
base
d
on
t
heir
curre
nt
locat
ion.
T
o
determi
ne
wh
e
ther
a
pa
rtic
ular
ve
hicle
,
wh
e
re
(
=
{
1
,
2
,
.
.
.
.
,
}
)
is
a
pa
rt
of
the
ser
vice
se
ct
or
of
the
th
ECD at a
give
n i
ns
ta
nt i
n
ti
me
, a fla
g i
s u
ti
li
zed. The
f
la
g
i
s meas
ur
e
d usi
ng
:
(
)
=
{
0
,
ℎ
ℎ
1
,
ℎ
(1)
To
tra
ns
mit
th
e
com
puta
ti
onal
assignme
nt
t
o
the
i
nten
de
d
destinat
io
n
se
gme
nt's
veh
ic
le
,
V2V
te
c
hnology
sh
oul
d be
util
i
zed. The
ti
me re
qu
i
red f
or tra
ns
missi
on of th
e
co
mputi
ng t
ask, can
b
e
calc
ulate
d usin
g:
(
)
=
∑
∑
′
=
1
=
1
(
)
.
′
(
)
.
(
1
−
′
(
)
)
.
2
.
(
.
′
+
1
)
(2)
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
S
N
:
2088
-
8708
In
t J
Elec
&
C
omp E
ng,
V
ol.
15
, No
.
1
,
Febr
uary
20
25
:
1229
-
1241
1232
wh
e
re
,
′
in
t
he
e
qu
at
io
n
is
the
count
of
ve
hicle
s
that
wer
e
r
oute
d
t
o
′
fro
m
rate
of
data
t
ransmi
ssio
n
us
in
g
V2V
te
c
hnolog
y
is
represente
d
by
2
,
wh
il
e
bin
a
ry
va
riable
′
(
)
is
us
e
d
to
dete
rmin
e
,
if
i
s
de
li
ver
e
d from
to
′
at
a
giv
e
n t
ime i
ns
ta
nt
. T
he
cal
culat
io
n of
′
(
)
is give
n by
:
′
(
)
=
{
0
,
′
1
,
ℎ
(3)
Fo
r
the
ℎ
(
=
{
1
,
2
,
.
.
.
,
}
)
c
ompu
ti
ng tas
k
,
th
e
durati
on
of offloa
ding is
establi
sh
e
d by
:
(
)
=
∑
=
1
(
)
.
2
(4)
The
rate
of
data
trans
missi
on
in
V
2I
te
ch
no
l
ogy
is
de
no
te
d
by
2
.
The
durat
ion
require
d
for
ta
sk
e
xec
utio
n
reli
es
on
both
the
ta
sk
le
ngth
and
res
ource
unit
s’
pe
rfo
r
ma
nce.
I
f
rep
re
se
nts
the
al
l
-
ser
ve
r
capaci
ty
a
nd
represe
nts
res
ource
un
it
s
qu
antit
y
re
qu
est
e
d
f
or
.
T
he
a
moun
t
of
ti
me
need
e
d
for
th
e
execu
ti
on
of
the
ta
sk
is:
(
)
=
∑
=
1
(
)
.
.
(5)
Every
resou
rce
un
it
posses
ses
a
proc
essin
g
powe
r
de
note
d
by
.
A
fter
the
execu
ti
on
of
a
ta
sk
,
the
ve
hi
cl
es
mu
st
receive
f
e
edb
ac
k re
gardi
ng the
res
ults.
The
ti
me
re
qu
i
red f
or this
fee
db
ac
k
ca
n be
c
al
culat
ed by
(6)
:
(
)
=
′
2
(6)
Data’s
siz
e
i
n
t
he
outp
ut
ge
ne
rated
f
rom
t
he
execu
ti
on
of
is
denoted
by
′
.
Fo
r
t
he
im
ple
mentat
io
n
of
,
the ove
rall
ti
me r
e
qu
ire
d
is:
(
)
=
(
)
+
(
)
+
(
)
+
(
)
(7)
Subseque
ntly,
the ove
rall
dura
ti
on
require
d
t
o
im
pleme
nt all
the tasks
in
volving
co
mputat
ion
[
23]
is:
=
∑
(
)
=
1
(8)
2.1.2
.
M
od
el
f
or
ener
gy u
sage
The
am
ounts
of
e
nerg
y
us
e
d
by
EC
Ds
are
pri
maril
y
at
trib
uted
to
the
RS
Us
an
d
the
ser
ver
s
.
As
the
amo
un
t
of
ene
rgy
us
ed
by
RSUs
is
dyna
mica
ll
y
re
gu
la
te
d
base
d
on
their
em
ploym
ent
sta
tus
wh
i
le
the
y
remain
in
ope
rati
on
al
m
od
e
,
our
at
te
ntio
n
is
mai
nly
on
the
se
r
vers.
Amount
of
e
ne
rgy
us
e
d
by
serv
e
rs
include
s
baseli
ne
e
ne
rgy
co
nsume
d
w
hile
t
hey
a
re
r
unning,
ene
r
gy
us
e
d
by
the
unocc
up
ie
d
res
ource
unit
s
al
so
en
e
rgy uti
li
zed b
y res
our
ce un
it
s that ar
e o
ccu
pied
[
23]
.
T
he
ser
ve
rs’
serv
ic
e ti
me i
s
the primar
y
fac
tor
in
determi
ning e
ne
rgy usa
ge.
Th
e
ser
vice ti
me
can
be
c
ompu
t
ed by
(
9)
:
(
)
=
=
1
(
(
)
.
(
)
)
(9)
The bina
ry v
a
r
ia
ble
L
n
m
(
i
)
assesse
s
if
is carried
ou
t on
.
(
)
=
{
0
,
1
,
ℎ
(10)
Ba
sel
ine en
e
rgy
c
onsumpti
on
of all
ECDs'
s
erv
e
rs'
is:
=
∑
=
1
(
)
.
(11)
The
EC
D
ser
ve
rs
ha
ve
a
pow
er
rate
de
note
d
by
t
he
va
riabl
e
α
.
T
he
pr
oce
ss
of
deter
mini
ng
t
he
ene
r
gy
us
a
ge
for reso
urce
unit
s u
ti
li
zed em
ployed
is:
=
∑
∑
=
1
=
1
(
)
.
(
)
.
(12)
Evaluation Warning : The document was created with Spire.PDF for Python.
In
t J
Elec
&
C
omp E
ng
IS
S
N:
20
88
-
8708
Multi
-
obje
ct
iv
e opti
mized
task
sch
e
duli
ng in c
ogniti
ve inte
r
net o
f ve
hicle
s:
towar
ds
…
(
M
.
Divy
ashree
)
1233
The
var
ia
ble
β
represe
nts
the
powe
r
rate
of
r
eso
ur
ce
unit
s
that
are
e
mp
l
oyed.
T
he
proces
s
of
dete
rmin
i
ng
the
energ
y usag
e
for res
ource
unit
s that are
une
mp
lo
ye
d
is ca
r
ried o
ut by
:
=
∑
(
−
∑
=
1
=
1
(
)
)
.
(
)
.
(13)
The
va
riable
γ
represe
nts
the
powe
r
rate
of
r
eso
ur
ce
unit
s
t
hat
a
re
unem
plo
ye
d.
F
ollo
wing
t
hat,
t
he
a
ggr
egat
e
of all
servers'
e
nerg
y usag
e
[
23]
is dete
rmin
e
d usin
g
:
=
+
+
(14)
2.2
.
Sche
duli
ng
c
omput
at
i
on b
as
ed
on t
he
multi
-
obj
ecti
ve
p
ar
ticl
e swarm
optimi
zati
on
Unde
r
t
his
s
ub
sect
ion
,
Fir
stl
y,
the
offloa
ding
path
for
the
com
pu
ta
ti
onal
ta
sk
s
is
ac
qu
i
r
ed
t
hro
ugh
the
ad
opti
on
of
V
2V
tra
ns
mis
sion.
F
ur
t
herm
or
e
,
we
util
iz
e
mu
lt
i
-
obje
ct
ive
opti
miza
ti
on
i
n
order
to
dete
rmin
e
the
m
os
t
ef
f
ec
ti
ve
sche
du
li
ng
strat
e
gy
f
or
these
ta
sk
s
,
ba
la
ncing
obje
ct
ives
li
ke
mini
mizi
ng
e
xec
ution
ti
me
and re
duci
ng e
nerg
y
c
on
s
ump
ti
on
. T
his a
ppr
oach ma
ximize
s over
al
l
perfor
mance i
n
CI
oV
.
2.2.1
.
V
2V t
r
ansmi
ssion
to
offlo
ad p
at
h
ac
quisi
tion
In
order
to
off
load
co
mputi
ng
ta
s
ks
to
the
go
al
ECD
us
in
g
V
2V
c
om
m
unic
at
ion
,
a
vehi
cl
e
path
to
the
final
ve
hic
le
fr
om
the
ini
ti
al
veh
ic
le
must
be
desig
ne
d
[26]
.
The
offload
i
ng
process
can
be
op
ti
mi
zed
thr
ough
strat
e
gic
path
desi
gn.
This
ma
ximize
s
the
pot
entia
l
of
V2V
co
mm
un
ic
at
ion
in
CI
oV,
r
edu
ci
ng
la
te
ncy
a
nd e
nhanci
ng com
puta
ti
on
al
ef
fici
ency.
Algorith
m
1
.
O
ff
loa
ding
pat
h ac
qu
isi
ti
on
Inp
ut: Ve
hicle
s (
V
), EC
D
Ou
t
pu
t:
Pat
h (
P)
Step
1:
W
hile
veh
ic
le
s a
re c
overe
d u
nd
e
r
th
e init
ia
l EC
D d
o
Step
2: Cal
cula
te
d
ist
ance
bet
ween t
w
o veh
i
cl
es
Step
3: Cal
cula
te
d
ist
ance
bet
ween t
he g
oal
veh
ic
le
s a
nd th
e v
e
hicle
s
Step
4:
E
nd
while
Step
5: Ch
oose
the mi
nim
um
distance
betwe
en
init
ia
l ve
hic
le
an
d t
he g
oal
v
ehicl
e
Step
6:
Re
tur
n
P
2.2.2
.
C
omp
u
t
at
i
on
sc
heduli
ng
ba
se
d
on
m
ulti
-
objec
tive opt
im
iz
at
i
on
e
nergy c
omput
at
i
on
(MOE
C
)
This
s
ub
sect
i
on
pro
poses
a
sche
du
li
ng
meth
od
for
cl
oud
-
e
dge
com
pu
ti
ng,
a
ddressi
ng
t
he
mu
lt
i
-
obje
ct
ive
opti
miza
ti
on
pro
blem
with
mu
lt
ipl
e
go
al
s.
To
ad
dress
t
hi
s,
we
em
ploy
MOPS
O
[
27]
–
[
31]
wh
ic
h
is
an
a
ccur
at
e
a
nd
r
obus
t
strat
e
gy
in
ha
ndli
ng
com
plex
opti
miza
ti
on
ta
s
ks.
MOP
SO
e
ffec
ti
vely
balances
m
ul
ti
ple
co
nf
li
ct
in
g
obje
ct
ives,
m
akin
g
it
ideal
for
op
ti
mizi
ng
cl
oud
-
e
dge
c
ompu
ti
ng
sc
he
du
le
s
.
Her
e
,
we
a
do
pt
M
O
EC
to
so
lve
the
ene
r
gy
-
relat
ed
as
pe
ct
s
of
th
e
opti
miza
ti
on
pr
oble
m.
By
inte
gr
at
in
g
M
O
PS
O,
the
propose
d
sc
hedulin
g
met
hod
M
O
EC
ens
ur
es
e
ff
ic
ie
nt
an
d
bala
nced
ta
s
k
dis
tribu
ti
on,
consi
der
i
ng f
ac
tors
s
uc
h
as
exe
cution t
ime a
nd en
e
r
gy cons
umpti
on.
a.
In
it
ia
li
ze
p
a
ra
mete
rs:
Def
i
ne
the
swa
rm
or
popula
ti
on
siz
e
,
ma
xim
um
it
erati
on
s
a
nd
othe
r
par
a
mete
rs
li
ke
the ine
rtia
l, co
gn
it
ive,
s
ocial
co
ef
fici
ents a
nd m
utati
on r
at
e
,
et
c
.
b.
In
it
ia
li
ze
popu
la
ti
on
a
nd
r
ep
os
it
ory:
Pop
ulati
on
re
pr
ese
nt
s
the
colle
ct
io
n
of
par
ti
cl
es
with
t
heir
posi
ti
on
s
and
velocit
ie
s
i
n
sea
rch
s
pa
ce.
It
deter
m
ines
eac
h
par
t
ic
le
's
fitness
value
in
the
i
niti
al
popula
ti
on.
Re
po
sit
or
y
m
ai
ntains
no
n
-
dominate
d
s
olut
ion
s
ob
ta
ine
d
in
the
proce
ss
of
opti
miza
ti
on
.
In
it
ia
li
ze
a
popula
ti
on
sw
arm
of
pa
rtic
le
s,
eac
h
of
wh
ic
h
represe
nts
a
s
olu
ti
on
,
en
c
odin
g
th
e
distrib
utio
n
o
f
com
pu
ti
ng tas
ks
to
the se
r
vers
.
F
or eac
h par
ti
cl
e:
−
Ra
ndom
l
y
assi
gn comp
utin
g
t
asks
t
o
se
rv
e
rs, res
pecti
ng the
al
l
-
serv
e
r
ca
pa
ci
ty
q
.
−
In
it
ia
li
ze veloc
it
y
an
d posi
ti
on
vecto
rs
to
gu
ide searc
h
i
n
s
olu
ti
on s
pace.
Fit
ness
e
valuat
ion
:
f
or eac
h p
arti
cl
e, calc
ulate
the
ob
je
ct
ive
s:
−
:
t
otal t
ime usa
ge fo
r
al
l t
asks
assigne
d.
−
E
:
c
ompu
te
tot
al
energy
c
ons
umpti
on
us
i
ng
,
,
an
d
co
ns
ideri
ng
t
he
dist
rib
ution
of
ta
s
ks
and res
ource
u
t
il
iz
at
ion
.
Creat
e a r
e
posit
ory
that st
or
es
the
best
non d
om
inate
d
s
olu
t
ion
s
am
ong
t
he
sw
a
rm
t
o
f
orm
a “Pa
reto fr
ont”
that f
ound s
o f
ar. N
on
-
domi
na
te
d
mea
ns
no
oth
e
r
s
olu
ti
on i
s b
et
te
r
i
n bo
t
h o
bject
ives
an
d
.
c.
Wh
il
e
(stop
c
onditi
on
=
=
false):
Co
ntin
ue
unti
l
a
te
rmi
natio
n
c
onditi
on
is
met
(Lik
e
ge
tt
ing
a
go
od
enou
gh s
olu
ti
on
qu
al
it
y o
r
r
ea
chin
g
a
maxi
m
um
c
ount
of it
erati
on
s
).
d.
Choose
le
a
der
“
ℎ
”
:
Ch
oose
a
le
ader
f
r
om
t
he
popula
ti
on
or
re
posit
ory
t
o
direct
t
he
mo
ti
on
of
eac
h
par
ti
cl
e.
To
ma
intai
n
a
div
e
rs
e
colle
ct
ion
of
so
luti
ons,
t
he
le
ader
s
houl
d
be
cho
s
en
acc
ordin
g
to
it
s
Pare
t
o
op
ti
mali
ty a
nd
typ
ic
al
ly a
d
i
ve
rsity mec
ha
ni
sm.
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
S
N
:
2088
-
8708
In
t J
Elec
&
C
omp E
ng,
V
ol.
15
, No
.
1
,
Febr
uary
20
25
:
1229
-
1241
1234
e.
Update
posit
io
ns
a
nd
veloc
it
ie
s
:
Update
ve
locit
y
f
or
ea
ch
pa
rtic
le
accor
ding
to
it
s
current
veloc
it
y,
distance
to
tha
t’s
pe
rs
on
al
be
st
po
sit
io
n
a
nd
distance
t
o
th
e
ch
os
en
le
ade
r
‘
ℎ
’
.
Eac
h
par
t
ic
le
’s
posit
ion
sh
oul
d be
upda
te
d
acco
r
ding t
o
the
n
e
w velo
ci
ty.
T
he new
po
sit
io
n rep
res
en
ts i
ts ne
w sol
ution.
f.
Perfo
rm
m
utati
on
(i
f
necess
a
ry):
A
pp
l
y
mut
at
ion
by
intr
oduci
ng
rand
om
per
t
urbati
on
s
or
al
te
rati
ons
to
so
me
par
ti
cl
es
to
e
xp
l
or
e
ne
w
re
gions
in
s
earch
s
pace
if
desire
d
to
ad
d
di
ver
sit
y
t
o
t
he
po
pu
la
ti
on
a
nd
pr
e
ve
nt to
o
ra
pi
d
co
nver
ge
nce
.
g.
Boun
dary
c
hec
ks
on
posit
ion
an
d
vel
ocity:
Check
t
hat
the
ne
w
posit
io
ns
an
d
vel
ociti
es
sta
y
within
t
he
def
i
ned
bounda
ries.
B
rin
g
a
ny
el
eme
nt
of
a
posit
io
n
or
ve
locit
y
vecto
r
ba
ck
i
ns
ide
the
bounda
ries
i
f
i
t
go
e
s
beyo
nd.
h.
Update
t
he
opti
mal
posit
io
ns
:
Upd
at
e
pe
rsonal
best
po
sit
io
ns
an
d
gl
ob
al
best
po
sit
io
ns
base
d
on
domina
nce
or
f
it
ness
crit
eria.
Use
obje
ct
ives
‘
’
an
d
‘
’
to
e
va
luate
the
fitne
ss
of
new
posit
ion
s
.
Update
a
par
ti
cl
e's
per
s
onal
best
in
te
r
ms
of
Pareto
dom
inance
if
,
it
s
new
posit
ion
i
s
bette
r
tha
n
i
ts
pr
e
vious
one.
Add
new
no
n
-
dominate
d
so
l
utions
to
the
r
eposi
tor
y,
e
nsu
rin
g
that
it
el
imi
nates
du
plica
te
s
and maintai
ns
i
ts represe
ntati
on
of the c
urre
nt.
i.
Terminati
on
c
heck
:
The
al
gorith
m
c
on
ti
nues
it
erati
ng
un
ti
l
set
c
ount
s
of
it
erati
ons
ha
ve
passe
d,
n
o
sign
ific
a
nt im
pro
veme
nt is
observ
e
d, or a
nother
sto
pp
i
ng c
onditi
on is sati
sfied.
j.
Re
su
lt
:
Ou
t
pu
t
the
Pa
reto
fron
t
t
hat
re
pr
e
sents
the
best
trade
-
offs
f
ound
betwee
n
t
ime
co
ns
umpt
ion
an
d
total
e
nergy co
nsum
ptio
n
E
f
or exec
uting co
mputi
ng t
asks o
n
se
rv
e
rs
.
As
a
res
ult,
th
e
M
O
PS
O
al
gorith
m
updates
pa
rtic
le
velocit
ie
s,
po
sit
io
ns
and
t
he
re
posit
ory
to
fi
nd
non
-
domi
nated
so
l
utio
ns
for
mu
lt
i
-
obje
ct
ive
opti
miza
ti
on
pro
blems.
T
he
it
erati
ve
proc
ess
c
on
ti
nues
,
with
par
ti
cl
es
e
xp
l
ori
ng
a
nd
e
xp
l
oi
ti
ng
t
he
s
earc
h
sp
ac
e,
unti
l
te
rmin
at
io
n
co
nd
it
ion
s
a
re
met.
Upo
n
meet
ing
these
conditi
ons,
the
al
gorith
m
c
oncl
ud
es
by
ide
ntify
in
g
opti
mal or
nea
r
-
opti
ma
l
so
luti
ons
that
bala
nce
t
he
m
ulti
ple
ob
je
ct
ives
of t
he op
ti
miza
ti
on
pro
blem.
2.3
.
Simul
at
io
n setup
This
s
ubsect
io
n
i
nclu
des
s
e
ver
al
e
xte
ns
iv
e
simulat
io
ns
carried
ou
t
to
eval
uate
ef
fe
ct
iveness
of
su
ggest
e
d
sc
he
du
li
ng
meth
od
for
e
dge
c
omp
uting,
ref
e
r
red
as
MOEC
by
us
in
g
MATL
AB
2020
w
hich
introd
uces
t
he
M
A
TLAB
9.8
runtime
.
It
ena
bles
par
al
le
l
c
ompu
ti
ng
f
or
hundre
ds
of
f
unct
ions.
This
f
eat
ur
e
al
lows
us
e
rs
to
us
e
l
ocal
m
ulti
cor
e
proce
ssor
s
an
d
gra
ph
ic
s
processi
ng
un
it
s
(
GPUs
)
,
an
d
scal
e
co
mputat
ion
s
to comp
ute cl
ust
ers,
im
pro
ving
performa
nce
and pr
oductivit
y
in
lar
ge
-
scal
e
d
at
a
processi
ng tas
ks
.
Fo
r
ou
r
sim
ulati
on
,
we
c
onsider
va
ry
i
ng
nu
mb
e
rs
of
ve
hicle
s,
sp
eci
fical
l
y
20,
40,
60,
80,
100,
an
d
120.
The
rate
of
data
tran
sm
issi
on
us
in
g
V
2V
te
c
hnolog
y
denoted
a
s
λ
V2V
,
an
d
the
rate
of
data
tran
sm
issi
on
us
in
g
V
2I
te
c
hnolog
y,
de
no
te
d
as
λ
V2I
,
a
r
e
both
set
to
1
G
bp
s
a
nd
600
M
bps
r
esp
ect
ively,
f
ollo
wing
t
he
values
mentio
ned
i
n
[
1],
[
12]
.
The
si
mu
l
at
ion
pa
ramet
er
set
ti
ng
s
us
e
d
are
represe
nt
ed
in
Ta
ble
2.
T
he
Simulat
io
n
set
up
wa
s
us
e
d
t
o
te
st
t
he
s
ys
t
em
m
odel
s
f
or
pro
blem
f
orm
ulati
on
an
d
sc
hedulin
g
c
omp
utati
on
strat
egies.
S
ubseq
uen
tl
y,
the
pro
po
se
d
M
O
EC
performa
nc
e
is
assessed
on
ve
hicle
scal
es
in
te
rms
of
t
ime
and
energ
y usag
e
.
Table
2.
Simul
at
ion
par
a
mete
r
set
ti
ngs
Descripti
o
n
of the
p
aram
et
er
Valu
e
Data tran
smissio
n
r
ate us
in
g
V2V tec
h
n
o
lo
g
y
λ
V
2
V
1
Gbp
s
The ent
ire
n
u
m
b
e
r
o
f
ECDs is
M
20
The ECDs s
ervers
h
av
e a
p
o
wer
r
ate
o
f
α
3
0
0
W
Em
p
lo
y
ed
r
eso
u
rc
e un
its h
av
e a
p
o
wer
ra
te of
β
5
0
W
Un
em
p
lo
y
ed
r
eso
u
rce
u
n
its h
av
e a
p
o
wer rate
γ
3
0
W
Data tran
smissio
n
r
ate us
in
g
V2I tech
n
o
lo
g
y
λ
V
2
I
6
0
0
M
b
p
s
Every reso
u
rce
un
it po
ss
ess
es a proces
sin
g
po
wer
2
0
0
0
M
Hz
3.
RESU
LT
S
AND DI
SCUS
S
ION
The
pr
opos
e
d
sche
du
li
ng
or
offloa
ding
MO
EC
method
is
com
par
e
d
to
th
e
existi
ng
EC
O
meth
od
f
or
a
com
prehe
ns
i
ve
c
omparati
ve
analy
sis.
T
o
il
lustrate
the
va
riat
ion
s
a
nd
e
ff
ic
acy
of
t
hes
e
meth
ods,
a
t
hor
ough
com
par
is
on
is
pro
vid
e
d.
T
hi
s
pa
per
e
mp
l
oys
a
c
omparat
ive
ap
proac
h,
wh
ic
h
is
detai
le
d
in
the
f
ollow
i
ng
su
bse
ct
ions.
3.1
.
Ex
isting
ECO
The
pr
e
viousl
y
de
velo
ped
EC
O
[
15]
meth
od'
s
pur
pose
is
t
o
strike
a
bala
nc
e
betwee
n
opti
mizi
ng
t
he
us
e
of
ti
me
a
nd
lo
we
rin
g
e
nerg
y
c
ons
umpti
on.
The
me
thod
i
nvol
ves
cond
ucting
m
ulti
ple
e
xp
e
ri
ments
to
evaluate
it
s
pe
rformance
a
nd
i
de
ntify
op
t
imal
so
l
utions
.
T
o
dete
rmine
a
set
of
c
ompa
rati
vely
s
up
e
rio
r
so
luti
ons,
m
ulti
ple
-
crit
eria
de
ci
sion
-
maki
ng
(
M
C
D
M
)
w
it
h
simple
a
dd
it
ive
weig
htin
g
(
SAW
)
te
ch
niques
Evaluation Warning : The document was created with Spire.PDF for Python.
In
t J
Elec
&
C
omp E
ng
IS
S
N:
20
88
-
8708
Multi
-
obje
ct
iv
e opti
mized
task
sch
e
duli
ng in c
ogniti
ve inte
r
net o
f ve
hicle
s:
towar
ds
…
(
M
.
Divy
ashree
)
1235
wer
e
em
ploye
d.
These
te
ch
ni
qu
es
hel
p
in
a
ssessin
g
a
nd
s
el
ect
ing
the
be
st
offloa
ding
st
rategies
base
d
on
the
fitness
of the
s
o
luti
ons
reg
a
r
di
ng
ti
me
an
d
e
nerg
y ob
je
ct
iv
es.
3.2
.
A
na
ly
sis
of com
pa
ri
s
on
This
pa
per
pr
ov
i
des
a
de
ta
il
ed
c
omparis
on
of
the
pro
pose
d
M
O
EC
method
ology
with
t
he
EC
O
method
un
der
same
e
xp
e
ri
mental
c
onditi
on
s
.
Exec
utio
n
ti
me
a
nd
e
nerg
y
us
a
ge
are
t
he
t
wo
pr
ima
ry
par
a
mete
rs
use
d
t
o
e
valuat
e
the
pe
rfo
rm
ance.
T
o
il
lus
trat
e
the
real
res
ource
util
iz
at
ion
of
al
l
ECDs
par
ti
ci
patin
g
in
hosti
ng
the
co
mputi
ng
ta
sk
s
,
num
ber
of
em
ployed
ECD
s,
resou
rce
util
izati
on
,
a
nd
num
ber
o
f
com
pu
ta
ti
onal
ta
sk
s
offloa
ded acr
os
s t
he
EC
Ds
s
how
the
outc
om
e
s.
3.2.1
.
T
he us
age
of EC
Ds
-
a
comp
ariso
n
Both
EC
O
an
d
propose
d
M
O
EC
appr
oach
es
are
co
mp
a
red,
al
ong
with
t
he
amo
un
t
of
E
CDs
us
e
d.
Accor
ding
t
o
F
igure
2,
a
total
of
20
EC
Ds
a
r
e
use
d
in
this
e
xp
e
rime
nt.
The
MOEC
te
ch
ni
qu
e
em
ploys
E
CDs
more
f
re
qu
e
ntly
as
the
num
be
r
of
ve
hicle
s
increase
s,
as
il
lustrate
d
in
the
fig
ur
e.
N
otabl
y,
All
ECD
s
ne
ed
to
be
op
e
rati
onal
in
orde
r
to
me
et
the
re
qu
i
re
ments
for
the
dep
l
oyment
of
the
co
mputi
ng
ta
sks
once
t
her
e
a
re
100 ve
hicle
s.
M
O
EC
is
high
ly
scal
able
an
d
eff
ic
ie
nt
in
ha
nd
li
ng
higher
l
oads.
W
hile
E
CO
empl
oys
f
ewer
EC
Ds
acro
s
s
al
l
veh
i
cl
e
nu
m
be
rs,
i
nd
ic
at
in
g
po
te
ntial
li
mit
ations
in
ha
ndli
ng
l
arg
e
r
num
be
rs.
M
O
EC's
sig
ni
ficant
increase
in
E
CD
us
a
ge
as
veh
ic
le
num
be
rs
rise
s
upport
s
it
s
po
te
ntial
for
real
-
worl
d
ap
plica
ti
on
s
wh
e
re
dema
nd can
va
ry great
ly.
Figure
2. The
qu
a
ntit
y of EC
Ds
e
mp
l
oy
e
d
a
t var
i
ou
s
v
e
hic
le
size b
y
EC
O
and
pro
po
se
d MOEC
:
a compa
rison
3.2.2. T
he us
age
of reso
urce
s
-
a
c
ompari
s
on
It
is
gua
ran
te
ed
that
t
he
re
so
urce
un
it
w
il
l
be
occ
up
ie
d
once
al
l
c
omp
uting
j
ob
s
hav
e
bee
n
trans
ferred
to
t
he
EC
Ds
us
i
ng
the
a
pprop
riat
e
te
chn
i
qu
es
.
Fig
ur
e
3
c
ompares
the
res
ource
util
iz
at
ion
of
the
ECDs
by
t
he
E
CO
a
nd
th
e
propose
d
MOEC
at
diff
e
re
nt
ve
hicle
scal
es.
T
he
res
ource
uti
li
zat
ion
is
dete
rmin
e
d
by
est
imat
ing
the
numb
e
r
of
eng
a
ge
d
EC
Ds
an
d
the
am
ount
of
resou
rce
un
it
s
al
lott
ed
to
eac
h
EC
D.
Higher
resou
rce
util
iz
at
ion
res
ults
f
rom
al
locat
ing
mo
re
res
ourc
e
un
it
s
but
usi
ng
few
e
r
E
CDs.
T
he
fin
dings
il
lustrate
d
i
n
F
igure
3
de
mon
strat
e
that
MO
EC
co
ns
is
te
ntl
y
outpe
rforms
ECO
i
n
res
our
ce
util
iz
at
ion
a
cro
ss
var
i
ou
s
v
e
hicle
scale
s, wit
h
a
n est
imat
ed 80%
u
ti
li
zat
ion
r
at
e.
Eff
ic
ie
nt
resou
rce
util
iz
at
ion
is
essenti
al
for
addressi
ng
i
nc
reased
de
man
d
an
d
reducin
g
waste.
T
he
higher
res
ourc
e
util
iz
at
ion
ra
te
s
of
M
OEC,
com
par
e
d
t
o
E
CO,
validat
e
t
hat
MOEC
w
ould
be
more
e
f
fecti
ve
in
res
ource
m
anag
e
ment
.
MOEC'
s
c
onsist
ent
80%
util
iz
at
ion
in
dicat
e
s
it
s
capa
bili
ty
to
ha
ndle
va
rio
us
op
e
rati
onal
d
e
man
ds
e
ff
ic
ie
nt
ly.
3.2.3. Am
ou
n
t
of c
omp
u
t
at
i
onal
t
as
k
offl
oaded
a
m
ong E
CD
s
:
a
c
ompa
ri
so
n
In
c
om
m
on
pr
act
ic
e,
the
c
omp
uting
ta
s
k
is
delegated
t
o
the
neig
hbori
ng
ECD
.
H
ow
e
ver,
i
n
cases
wh
e
n
the
ve
hicle
siz
es
are
modest,
the
re
is
a
random
di
stribu
ti
on
of
t
he
ve
hicle
s
a
mong
di
ff
e
ren
t
ECD
ranges.
Under
these
c
onditi
on
s,
assi
gn
i
ng
al
l
co
mputat
ion
al
ta
sk
s
to
the
surr
oundin
g
EC
Ds
may
le
ad
to
ma
ny
ECDs
bei
ng
act
ive
simult
aneousl
y,
res
ul
ti
ng
in
e
xcess
ive
ene
rgy
c
onsum
ption.
W
e
so
lve
this
i
n
our
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
S
N
:
2088
-
8708
In
t J
Elec
&
C
omp E
ng,
V
ol.
15
, No
.
1
,
Febr
uary
20
25
:
1229
-
1241
1236
exp
e
rime
nt
by
offloa
ding
the
com
pu
ta
ti
onal
bur
den
from
t
he
su
r
rou
nd
i
ng
ECD
to
a
neig
hbori
ng
ECD
.
Wh
e
n
the
area
of
co
ve
rag
e
of
th
e
de
sti
nation
ECD
diff
e
rs
f
r
om
th
at
of
the
or
igi
n
ECD,
this
al
lo
ws
the
offloa
din
g
of
com
pu
ti
ng
ta
s
k
ac
ro
s
s
ECDs
.
Fig
ur
e
4
c
ompares
both
EC
O
an
d
pro
po
se
d
M
OEC
te
c
hniq
ues
f
or
the
amo
unt
of
c
omp
utati
ona
l
ta
sk
s
offl
oad
e
d
a
mon
g
ECDs.
It
s
hows
that
MO
EC
eff
ic
ie
ntl
y
util
iz
es
res
ources
by
distrib
uting
c
omp
utati
on
al
ta
sk
s
ac
r
os
s
EC
Ds
as
ve
hicle
siz
e
inc
rease
s.
Desp
it
e
both
EC
O
a
nd
M
O
EC
sh
owin
g
a
n
i
nc
rease
in
offlo
aded
ta
sk
s
,
EC
O
offl
oa
ds
m
ore
ta
sk
s
f
or
m
os
t
ve
hicle
co
un
ts
,
esp
eci
al
ly
w
he
n
veh
ic
le
s e
xcee
d 60.
ECO's
higher
offloa
ded
ta
sks
ind
ic
at
e
reli
ance
on
exte
rn
a
l
dev
ic
es
,
ca
usi
ng
highe
r
e
ne
rgy
a
nd
ti
me
consu
mp
ti
on.
M
O
EC's
fe
we
r
offl
oad
e
d
ta
sk
s
s
uggest
lo
cal
computat
ion
s
or
ef
fici
ent
ta
sk
ma
na
ge
ment
strat
egie
s,
re
duci
ng
offloa
di
ng.
As
the
syst
em
scal
es,
both
al
gorith
ms
'
offl
oad
e
d
ta
sk
s
re
flect
gr
ow
i
ng
com
pu
ta
ti
onal
dema
nd.
MOE
C's
supe
rio
r
e
f
fici
ency
in
ma
nag
i
ng
resou
rc
es
e
ns
ures
bett
er
perf
or
ma
nce
as
the
numb
e
r of ve
hi
cl
es increases
.
Figure
3. Re
source
u
s
a
ge bet
ween EC
O
and
prop
os
ed
M
O
EC at
v
a
rio
us
veh
ic
le
size
:
a
c
omparis
on
Figure
4. A
moun
t
of c
ompu
ta
ti
on
al
task
offl
oad
e
d
a
mon
g EC
Ds by EC
O and p
r
opos
e
d MOEC
at
var
i
ou
s
veh
ic
le
size
:
a
com
par
is
on
3.2.4
.
E
nergy
usage
:
a
co
mp
ariso
n
As
sta
te
d
in
S
ect
ion
3,
t
her
e
are
t
hr
ee
par
t
s
to
e
nerg
y
co
ns
um
ption
:
the
ba
sel
ine
e
nerg
y
us
e
for
al
l
the
se
rv
e
rs
in
the
EC
Ds,
the
ene
r
gy
us
a
ge
of
res
ource
uni
ts
w
hich
a
re
i
n
us
e
,
a
nd
t
he
ene
r
gy
us
a
ge
of
the
un
it
s
t
hat
are not
in u
se.
T
he
r
esults
re
veal
th
ese
three
ene
r
gy u
sag
e
fact
or
s
for
EC
O
a
nd
t
he
propose
d
MOEC
te
chn
iq
ues
ac
r
os
s v
ario
us
v
e
hicle
scal
es
in
Figure 5
.
It
is
e
vid
e
nt
from
Figure
5(a
)
that wh
e
n
the v
e
hicle
scal
e
grows,
ECO
s
hows
a
n
i
ncr
e
ase
in
ba
sel
ine
en
er
gy
c
onsumpti
on
f
or
al
l
the
ser
vers
i
n
the
EC
Ds
esp
eci
al
ly
beyo
nd
60
ve
hi
cl
es,
reac
hing
up
to
a
bout
2
k
W
h
at
120
ve
hicle
s.
H
ow
e
ve
r,
the
M
O
EC
appr
oach
c
ons
um
es
Evaluation Warning : The document was created with Spire.PDF for Python.
In
t J
Elec
&
C
omp E
ng
IS
S
N:
20
88
-
8708
Multi
-
obje
ct
iv
e opti
mized
task
sch
e
duli
ng in c
ogniti
ve inte
r
net o
f ve
hicle
s:
towar
ds
…
(
M
.
Divy
ashree
)
1237
le
ss
energy
t
ha
n
ECO
si
nce
it
employs
few
e
r
ECDs
.
The
e
nerg
y
us
a
ge
of
employe
d
res
ource
unit
s
acr
os
s
the
var
i
ou
s
ve
hicle
s
siz
e
is
seen
in
Fi
gure
5(b
).
As
it
us
e
s
al
mo
st
the
s
ame
amo
unt
of
res
ource
un
i
ts
fo
r
com
pu
ti
ng
act
ivit
ie
s,
the
MOEC
te
ch
no
l
ogy
ac
hie
ves
c
ompara
ble
en
ergy
c
on
s
ump
ti
on
of
the
ut
il
iz
ed
resou
rce
un
it
s
acro
s
s
var
io
us
veh
ic
le
scal
es
.
I
n
Fi
gure
5(
c
)
the
MOEC
str
at
egy
yields
sli
gh
tl
y
hi
gher
e
nerg
y
consu
mp
ti
on
f
rom
idle
res
ou
rce
un
it
s
w
hich
ca
n
be
vie
w
ed
i
n
the
c
on
te
xt
of
it
s
over
al
l
opti
miza
ti
on
go
al
s
.
Energ
y
us
a
ge
com
par
is
on
in
F
igure
6,
un
de
rlines
t
he
M
O
EC
meth
od's
s
up
e
rio
r
perf
ormance
eve
n
f
urt
her.
As
an
i
ns
ta
nce
,
if
there
are
ove
r
100
ve
hicle
s,
le
ss
tha
n
1
k
W
h
of
e
nerg
y
is
us
e
d
by
the
MOEC
meth
od,
wh
e
rea
s
ECO
us
es
mor
e ene
rgy
t
han 1
k
Wh.
(a)
(b)
(c)
Figure
5. Pro
pose
d MO
EC
a
nd ECOs
c
omp
ariso
n of seve
r
al
en
er
gy c
on
s
umpti
on f
act
ors at va
rio
us
ve
hicle
siz
e (a)
b
asel
i
ne
en
e
rgy usa
ge
by all t
he se
r
ve
rs
in
the
ECD
s’
,
(b)
e
ne
rgy u
sage
by r
e
sour
ce u
nits em
ployed
,
and
(c
)
e
nerg
y usa
ge by i
dle re
source
unit
s
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
S
N
:
2088
-
8708
In
t J
Elec
&
C
omp E
ng,
V
ol.
15
, No
.
1
,
Febr
uary
20
25
:
1229
-
1241
1238
Figure
6. Ene
r
gy u
s
age
by pr
opos
e
d M
OEC
and the
ECO
a
t var
i
ou
s
v
e
hic
le
size
:
a
com
pa
rison
These
fi
nd
i
ngs
rev
eal
that
MOEC
is
more
e
nerg
y
-
e
ff
ic
ie
nt
and
scal
a
ble
than
EC
O,
high
li
gh
ti
ng
t
he
sign
ific
a
nce
of
energ
y
ef
fici
ency
in
e
dg
e
c
ompu
ti
ng.
MO
EC's
lowe
r
ba
sel
ine,
em
ploy
ed,
a
nd
une
mpl
oy
e
d
resou
rce
e
nerg
y
require
ments
ma
ke
it
s
uitab
le
for
la
rg
e
-
sca
le
impleme
ntati
on
s
.
Wh
il
e
E
CO's
higher
re
li
ance
on off
l
oad
i
ng tasks res
ults in i
ncr
ease
d
e
ne
rgy usa
ge.
3.2.5
.
Ti
me c
onsumed
:
a
c
om
pa
ri
s
on
The
offloa
di
ng
ti
me
a
nd
total
ti
me
co
ns
um
ption
bet
ween
th
e
pro
pose
d
MOEC
a
nd
the
E
CO
met
ho
d
are
co
mpa
re
d
at
var
io
us
ve
hi
cl
e
scal
es
in
F
igure
s
7
an
d
8
res
pecti
vely.
An
esse
ntial
sta
ti
sti
c
fo
r
cal
c
ulati
ng
ti
me
co
nsum
pt
ion
is
t
he
offl
oa
ding
ti
me.
Figure
7
cl
earl
y
sh
ows
that
M
OEC
meth
od
ha
s
qu
ic
ker
offload
i
ng
ti
mes
tha
n
t
he
ECO
met
hod
acr
os
s
al
l
ve
hicle
scal
es.
T
he
a
ve
rag
e
ti
me
f
or
offloa
ding
ta
s
ks
in
ECO
is
0.5
seco
nds
f
or
20
ve
hicle
s,
wh
il
e
MOEC
t
akes
0.6
seco
nds.
As
the
numb
e
r
of
ve
hicle
s
increases
,
the
ga
p
wide
ns
,
wit
h
M
O
E
C
ta
king
about
1
sec
ond
f
or
120
ve
hicle
s,
highli
ghti
ng
it
s
supe
rior
e
ff
ic
ie
nc
y.
I
n
Fi
gure
8,
we
c
ompare
d
t
he
total
ti
me
use
d
a
bout
t
he
t
wo
offl
oad
i
ng
te
chn
iq
ues
.
T
he
ECO
meth
od
re
qu
ire
s
ti
me
l
onge
r
than
the
MOE
C
so
luti
on
esp
eci
al
ly
beyond
80
veh
ic
le
s,
i
nd
i
cat
in
g
that
ECO
re
qu
i
res
more
ti
me
to
com
plete
ta
sk
s as
v
e
hicle
num
ber
s
gr
ow.
The
MOEC
m
et
hod
ou
t
performs
t
he
EC
O
method
i
n
offl
oad
i
ng
ti
me
a
nd
total
ti
me
c
on
s
umpti
on,
especial
ly
as
ve
hicle
co
unt
in
creases.
MOE
C's
eff
ic
ie
nt
sc
hedulin
g
a
nd
t
ask
mana
geme
nt
s
trat
e
gies
re
su
lt
i
n
lowe
r
offl
oad
i
ng
ti
mes
,
ma
intai
nin
g
ar
ound
1
sec
ond
even
with
120
ve
hicle
s,
com
par
e
d
to
ECO's
4.5
seco
nds.
MOEC
al
s
o
sho
ws
l
ower
total
ti
me
co
nsum
ption
,
in
dicat
in
g
su
pe
rio
r
scal
ab
il
it
y.
This
high
li
gh
ts
M
O
EC's
pote
ntial
for
pract
ic
al
tr
aff
ic
ma
nag
e
ment
a
ppli
cat
ion
s,
mai
nt
ai
nin
g
ef
fici
en
cy
a
nd
minim
iz
ing
delays
un
der
hi
gh comp
utati
onal
d
e
man
ds.
Figure
7. O
ff
l
oa
ding ti
me
us
a
ge of
ECO
and
prop
os
ed
M
O
EC at
v
a
rio
us
veh
ic
le
size
:
a
com
par
is
on
Evaluation Warning : The document was created with Spire.PDF for Python.