Int
ern
at
i
onal
Journ
al of
El
e
ctrical
an
d
Co
mput
er
En
gin
eeri
ng
(IJ
E
C
E)
Vo
l.
8
, No
.
6
,
Decem
ber
201
8,
pp. 4
713~
4723
IS
S
N: 20
88
-
8708
,
DOI: 10
.11
591/
ijece
.
v8
i
6
.
pp
4713
-
47
23
4713
Journ
al h
om
e
page
:
http:
//
ia
es
core
.c
om/
journa
ls
/i
ndex.
ph
p/IJECE
Hybrid
Genetic
Algo
rithms
an
d Si
mulated Ann
ea
lin
g fo
r
Mul
ti
-
trip
Vehicl
e
R
ou
ting Pro
blem
w
it
h Time W
ind
o
w
s
Ama
li
a
K
art
i
ka
Ariy
an
i
1
,
Wayan Fi
rd
aus M
ahmud
y
2
, Yusu
f
Pri
yo
Anggod
o
3
1,2
Dep
ar
t
m
ent
of
Inform
at
ic
s
,
Fa
c
ulty
of
Com
puter Sci
en
ce,
Br
awi
jay
a
Univer
sit
y
,
Indone
sia
3
Data
Sc
ie
nt
ist, I
lmuone
Data, In
donesia
Art
ic
le
In
f
o
ABSTR
A
CT
Art
ic
le
history:
Re
cei
ved
Ma
r 28
, 201
8
Re
vised
Ju
l
13
,
201
8
Accepte
d
J
ul
27
, 2
01
8
Vehic
l
e
routi
ng
proble
m
with
ti
m
e
windows
(V
RP
TW)
is
one
of
NP
-
har
d
proble
m
.
Multi
-
tri
p
is
appr
oa
c
h
to
solve
the
VRP
TW
tha
t
looki
ng
tri
p
sche
duli
ng
for
g
et
s
best
result.
E
ven
though
the
r
e
are
var
ious
al
g
orit
hm
s
for
the
probl
em,
there
is
opportunit
y
to
impr
ove
th
e
exi
sting
al
gori
th
m
s
in
orde
r
gai
ning
a
b
et
t
er
result
.
In
th
is
re
sea
rch
,
g
enetic
al
gorit
m
is
h
y
br
idi
z
ed
with
sim
ula
te
d
annea
li
ng
al
gori
tm
to
solve
the
proble
m
.
Gene
tic
al
gorit
m
is
emplo
y
ed
to
exp
lore
global
sea
rc
h
are
a
and
sim
ula
te
d
ann
ea
l
ing
is
emp
lo
y
e
d
to
expl
o
it
lo
ca
l
sea
rch
area.
Four
combinat
ion
t
y
p
es
of
genetic
al
g
orit
hm
and
sim
ula
te
d
anneal
ing
(GA
-
SA
)
are
te
st
ed
to
get
th
e
b
est
soluti
on.
Th
e
computat
ion
al
e
xper
iment
show
s
tha
t
GA
-
SA
1
and
GA
-
SA
4
ca
n
produc
ed
the
m
ost
opti
m
al
fit
n
ess
ave
ra
ge
val
ues
with
ea
ch
v
al
ue
was
1.
0888
and
1.
0887
.
How
ever
GA
-
SA
4
ca
n
found
the
best
f
itness
chr
om
osome
faste
r
tha
n
GA
-
SA
1.
Ke
yw
or
d:
Gen
et
ic
al
gorithm
Hybr
i
d
al
gorithm
Mult
i
-
trip
Rou
ti
ng
pro
blem
Si
m
ulate
d
an
ne
al
ing
al
gorith
m
Copyright
©
201
8
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
:
Wayan
Fir
dau
s
Mahm
ud
y
,
Faculty
of Com
pu
te
r
Scie
nc
e,
Bra
wij
ay
a
U
niv
e
rsity
,
Vetera
n
Roa
d,
Ma
la
ng
,
Indo
ne
sia
.
Em
a
il
:
wayanfm
@u
b.
ac.id
1.
INTROD
U
CTION
The
gr
ow
t
h
of
the
nu
m
ber
of
tourist
s
in
tourist
destinat
ion
s
l
ocated
in
Ba
nyuw
a
ngi,
East
Java
,
Ind
on
esi
a,
is
increasi
ng
as
th
e
infr
ast
ru
ct
ure
i
m
pr
ov
em
ents
by
the
gove
r
nm
ent.
W
it
h
t
he
increa
sin
g
nu
m
be
r
of
visit
s,
the
re
are
s
om
e
pr
oblem
s
fo
r
t
ourist
s
com
ing
f
ro
m
ou
t
of
to
wn
or
from
abr
oa
d.
So
m
e
of
them
fin
d
it
diff
ic
ult
to
determ
ine
wh
ic
h
sigh
ts
to
visit
with
the
tim
e
l
i
m
i
t
they
hav
e
set
du
e
to
lim
ited
acce
ss
to
av
ai
la
ble
inf
or
m
at
ion
.
T
hey
can
not
de
ci
de
the
trip
base
d
on
their
tim
e
lim
it
wi
th
m
ini
m
u
m
c
os
t
a
nd
ar
rive
to
th
e
destinat
io
n
on
op
e
ning
hours
[1
]
.
That
pro
ble
m
us
ually
cal
le
d
as
sche
duli
ng
pr
ob
le
m
.
Schedulin
g
pro
bl
e
m
is
to
m
ake
a
sche
du
le
to
visit
to
ur
places
that
t
otal
tim
e
of
th
e
sche
du
le
is
ba
la
nced
unde
r
m
any
con
trai
ns
[2
].
Sche
du
li
ng
pr
ob
le
m
is
com
plex
pr
ob
le
m
becau
se
t
o
ge
t
best
s
olu
ti
on
m
us
t
be
balan
ci
ng
al
l
ocati
on
the
visit
that
to
pass
al
l
co
ntrains
.
T
he
m
a
in
f
ocu
s
of
sche
duli
ng
is
to
m
ini
m
iz
e
the
total
durati
on
of
the
visit
[
3].
I
n
add
it
io
n,
the
de
la
y
of
sche
du
li
ng
durin
g
ch
ang
e
the
pa
rt
is
ve
ry
im
po
rtant
becau
se
ge
t
in
flue
nce
the
tota
durati
on
of
t
he
sche
du
li
ng
[
4]
.
The
researc
h
f
oc
us
to
get
m
axi
m
u
m
so
lu
ti
on
with
hav
i
ng
m
ini
m
u
m
c
os
t
as
low
as possible
.
The
t
ourist
m
us
t
ha
ve
a
sc
he
du
le
tri
p.
T
o
s
ol
ve
this
pro
ble
m
need
s
a
ce
rtai
n
m
od
el
.
Anggo
do
et
al
[1
]
had
fi
nish
e
d
this
pro
blem
with
sa
m
e
data.
It
ca
n
so
lve
t
he
pro
bl
e
m
with
veh
i
cl
e
routing
pro
ble
m
appr
oach.
Veh
i
cl
e routin
g pro
blem
is p
ro
ce
dure
for sol
ving
sche
du
li
ng
pro
blem
in
veh
ic
l
e r
ou
te
.
Veh
ic
le
routin
g
prob
le
m
with
ti
m
e
wind
ows
is
a
com
bin
at
ion
sc
hedul
e
visit
ing
m
any
places
wit
h
tim
e
con
strai
ne
d
ser
vice
pro
visio
n
[
5].
V
RPT
W
is
us
e
d
to
fi
nd
op
ti
m
al
ro
ute
of
veh
ic
le
w
hich
le
ave
a
centrali
zed
de
po
t.
T
he
s
olu
t
ion
of
r
oute
ge
ts
fr
om
geog
raphical
ly
disp
erse
d
c
us
tom
ers.
Fi
nd
i
ng
ve
hicle
sche
du
li
ng
an
d
routing
plan
a
nd
m
ini
m
iz
ing
total
ro
ute
distance
are
the
focus
of
the
probl
e
m
[6
]
.
VRPT
W
is
us
e
d
in
lo
gisti
cs
m
anu
factur
i
ng
with
the
c
om
plexit
y
pr
obl
e
m
.
As
m
any
as
co
ns
trai
n
m
akes
the
pro
ble
m
as
diff
ic
ult
as.
In
fact,
to
so
lve
VRPT
W
pro
ble
m
is
easy
wh
en
just
hav
i
ng
on
e
de
pot
an
d
few
nodes
but
th
e
pro
blem
wil
l
beco
m
e
co
m
plex
w
he
n
ha
ving
m
any
depots
and
node
s.
I
n
add
it
io
n,
the
pro
blem
will
be
m
or
e
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
S
N
:
2088
-
8708
In
t J
Elec
&
C
om
p
En
g,
V
ol.
8
, N
o.
6
,
Dece
m
ber
2
01
8
:
4713
-
4723
4714
com
plex
wh
e
n
be
f
ound
the
fl
exible
pat
h
[7
]
.
Ba
sed
on
that,
VRPT
W
is
a
non
-
determ
insti
c
po
ly
nom
inal
-
tim
e
hard
(
HP
-
ha
r
d)
prob
le
m
[6
]
.
P
rev
i
ou
s
resea
rc
h
that
pr
opos
e
d
A
nggo
do
et
al
[1
]
ha
ve
co
nst
rain
day
of
tr
ip
the
tourist
.
Eve
ry
finish
in
g
the
tri
p
on
once
day
the
tourist
m
us
t
go
bac
k
to
th
e
ho
te
l
or
ce
ntr
al
iz
ed
depot.
I
n
this
stud
y,
the
t
ourist
will
visit
fo
r
m
any
days
so
the
pro
blem
has
m
a
ny
trips.
It
is
cal
le
d
m
ul
ti
-
trip
VRPT
W
that
diff
e
re
nce
pro
blem
with
on
l
y
VRPT
W
.
M
ulti
-
trip
VRPT
W
m
akes
pro
blem
to
be
com
plex
becau
se
on
ly
VRPT
W wit
hout m
ulti
-
trip ju
st nee
d on
e
leave a
nd go bac
k t
o
the
cent
rali
zed
depot.
Most resea
rch
e
rs
ha
ve bee
n
gr
ow
i
ng
i
nterest
to u
se c
om
bin
a
ti
on
a
nd
de
velop
i
ng
he
ur
ist
ic
s algorit
hm
to
so
lve
VR
PT
W
pro
blem
.
T
he
hybri
d
he
ur
ist
ic
al
go
rithm
gets
ver
y
str
ong
searc
hing
an
d
the
so
luti
on
qu
al
it
y
of
m
et
ho
d
is
high
[8
]
.
Ge
ose
iri
&
Gh
a
nnadpo
ur
[
5]
propose
d
goal
appr
oach
to
s
olv
e
ro
utin
g
pro
blem
,
bu
il
di
ng
a
nd
im
ple
m
ented
ge
netic
al
gorith
m
(G
A)
t
o
so
l
ve
the
pr
ob
le
m
.
Re
centl
y,
m
any
researchers
us
e
d
gen
et
ic
al
gorithm
to
so
lve
the
prob
le
m
.
Ur
sa
ni
et
al
[9
]
us
e
d
ge
ne
ti
c
al
go
rithm
in
routin
g
prob
le
m
fr
am
ewo
r
k.
G
eneti
c
al
go
rit
hm
can
get
m
ax
im
u
m
so
lutio
n
for
S
olo
m
on
benchm
ark
prob
le
m
,
research
pro
blem
,
and
E
-
Com
m
erce
Supp
li
er
[
7],
[
10
]
,
[
11
]
.
H
oweve
r,
ge
netic
al
gorithm
has
wea
kness
i
n
local
searchi
ng
s
olu
t
ion
.
S
olv
i
ng
th
e
r
ob
le
m
m
us
t
i
m
pr
ove
to
ef
f
ect
ive
an
d
ef
fici
ent
searc
hing
m
axi
m
u
m
so
luti
on.
Wang
et
al
[1
2]
propose
d
com
bin
at
ion
ge
netic
al
go
rith
m
with
par
ti
cle
swar
m
op
ti
m
ition
.
T
he
f
ocus
of
Gen
et
ic
al
gorithm
find
s
glob
al
so
luti
on
a
nd
pa
rtic
le
swarm
op
tim
iz
a
ti
on
sea
rch
local
so
luti
on.
It
is
balanc
e
and
ex
cel
le
nce
but
need
i
ng
m
or
e
tim
e
fo
r
co
m
pu
ta
ti
on.
Im
pr
ov
i
ng
cr
osso
ver
ge
netic
al
gorithm
us
e
cy
cl
ic
sh
ift
cr
os
s
ov
e
r
with
hill
-
cl
i
m
bing
m
echan
ism
to
gen
erate
a
child
in
init
ia
li
zi
ton
proses
[
13
]
.
It
is
go
od
m
od
el
to
so
l
ve
t
he
prob
le
m
bu
t
this
m
od
el
j
ust
th
e
f
ocu
s
f
or
sea
rch
i
ng
local
a
r
ea
beca
us
e
c
rosso
ver
m
od
ific
at
ion
m
akes ch
a
nge
fo
c
us
s
earc
h
a
r
ea.
Ba
sed
on
it
,
w
e
nee
d
m
od
el
to
s
olv
e
w
hich
balance
t
o
sea
r
ch
s
olu
ti
on
a
nd
fa
st
com
pu
ta
ti
on
.
In
fact,
gen
et
ic
al
gorit
hm
sh
ow
s
on
e
m
od
el
can
to
get
m
axi
m
u
m
so
luti
on
with
f
ocus
in
gl
ob
al
search
.
Pr
e
vi
ou
sly
,
gen
et
ic
al
go
rithm
had
c
ou
l
d
i
m
ple
m
entat
ion
in
this
ca
se
[
1]
.
Ma
ny
proce
dures
do
to
get
good
l
ocal
sea
rch.
I
n
this
stud
y,
ad
di
ng
c
om
bin
at
ion
one
he
ur
ist
ic
m
et
ho
d
w
hich
fo
c
us
i
n
local
searchi
ng.
Cr
osso
ver
m
od
ific
at
ion
is
no
t
done
be
cause
ca
n
cha
ng
e
f
ocu
s
g
e
ne
ti
c
al
go
rithm
from
glo
bal
se
arch
t
o
local
s
earch
.
Ma
hm
ud
y
[
14
]
,
i
m
pr
oved
sim
ulate
d
an
neali
ng
(SA)
f
or
r
ou
ti
ng
pr
ob
le
m
gets
the
best
so
luti
on
and
fast
com
pu
ta
ti
on.
Si
m
ulate
d
an
ne
al
ing
fo
c
us
on
local
sea
rc
h,
w
hen
us
i
ng
la
rg
e
data
sim
ulate
d
a
nn
eal
i
ng
te
nds
to
ge
t
poor
so
luti
on.
In
t
his
researc
h,
we
pro
po
se
ne
w
c
om
bin
at
ion
m
od
el
of
ge
netic
al
gorithm
and
si
m
ulate
d
ann
e
al
in
g
(GA
-
SA)
to
sol
ve
the
pro
blem
.
The
pro
po
s
ed
m
et
ho
d
will
te
st
with
oth
e
r
m
od
el
[
15
]
.
Fo
c
us
this
rese
arch
i
s
bu
il
di
ng
new
com
bin
at
ion
of
ge
n
et
ic
al
gorithm
and
sim
ulate
d
ann
eal
in
g
an
d
com
par
e
al
l
G
A
-
S
A
com
bin
at
ion
t
o get m
axi
m
u
m
so
luti
on.
2.
RELATE
D
W
ORK
Pr
e
visious
res
earch
a
bout
r
ou
ti
ng
prob
le
m
hav
e
done
to
get
m
axi
m
um
so
luti
on
.
I
t
is
no
t
ne
w
pro
blem
ho
we
ver
ve
ry
intere
sti
ng
t
o
fi
nd
t
he
m
od
el
w
hic
h
s
olv
i
ng
in
m
any
co
ndit
ion
s
.
T
he
sam
e
data
ha
d
been
fi
nish
e
d
by
A
nggo
do
[
1],
it
was
fi
nis
hed
usi
ng
genet
ic
al
go
rithm
.
It
propose
d
one
-
c
ut
-
po
i
nt
cr
os
s
ove
r
and
reciprocal
exch
a
ng
e
m
utati
on
to
get
the
so
l
ution.
to
get
m
axi
m
u
m
so
luti
on
it
was
us
ed
per
m
utat
ion
m
od
el
fo
r
ch
ro
m
os
om
e.
Pre
vious
st
ud
ie
s
,
G
A
ca
n
so
l
ve
routin
g
prob
le
m
[9
]
.
T
hat
stu
dy
pro
po
s
e
d
com
bin
at
ion
m
uta
ti
on
ope
r
at
or
a
nd
rec
om
bin
at
ion
ope
rator
f
or
t
he
gen
et
ic
s
al
gor
itm
op
erato
r.
In
t
hat
pro
blem
,
chr
om
os
o
m
e
rep
re
sentat
ion
was
us
e
d
pe
rm
utatio
n
m
od
el
.
G
A
was
us
e
d
to
s
olv
e
r
outi
ng
pro
ble
m
with
m
any
co
nst
rains
i.e
.
vehi
cl
e
with
li
m
ited
ca
pacit
y,
de
po
t
c
onstrai
ns,
an
d
ti
m
e
windows
co
ns
trai
ns
[
5].
that
researc
h
got
m
axi
m
u
m
resu
lt
beca
us
e
th
e
m
ai
n
fo
c
us
w
as
the
te
ch
niqu
es
gen
et
ic
al
go
rithm
that
us
ed
be
st
cost
-
best
r
oute
crosso
ve
r,
se
qu
e
nce
best
m
utati
on
,
an
d
pe
rm
utati
on
m
o
del
f
or
c
hrom
os
om
e
represe
ntati
on.
Ba
sed
on
previ
so
us
stu
dies, GA
is goo
d
m
eth
od
t
o
fi
nd
so
l
ution
in g
lo
bal search
i
ng
an
d
t
he
best
ch
r
om
os
om
e
represe
ntati
on
for ro
ute
prob
l
e
m
is p
erm
ut
at
ion
m
od
el
.
Rou
ti
ng
pro
blem
will
get
m
axi
m
u
m
so
luti
on
wh
e
n
have
m
eth
od
wh
ic
h
balance
to
fin
d
s
ol
ution
.
It
is
fo
c
us
on
gl
ob
a
l
and
l
ocal
sear
ch.
GA
can
do
fin
ding
s
olu
ti
on
in
global
sea
rch,
so
nee
ding
m
e
tho
d
ca
n
find
i
n
local
search
.
Ma
hm
ud
y
[1
4]
prosese
d
i
m
prov
e
d
sim
ulated
an
neali
ng
to
so
lve
r
outi
ng
pro
blem
.
It
go
t
go
od
so
luti
on
with
fast
com
pu
ta
tio
n.
Com
bin
at
ion
GA
a
nd
S
A
will
get
best
so
luti
on
whic
h
rather
tha
n
othe
r
m
et
ho
d.
G
A
do
fi
nd
i
ng
s
ol
ution
in
f
ocus
gl
ob
al
sea
rc
h
an
d
S
A
do
fin
ding
s
olu
ti
on
in
f
ocus
local
sear
ch.
T
o
get
bette
r
s
olu
t
ion
will
be
us
e
d
ada
ptive
c
hrom
os
e
[5
]
.
It
is
good
m
et
ho
d
howe
ver
to
ne
ed
one’
s
proce
dure
to
com
bin
at
ion
G
A
&
SA.
Faty
anosa
et
al
[
15
]
had
finis
hed
hybri
d
GA
an
d
SA
t
o
s
olv
e
be
nch
m
ark
s
f
unct
ion
.
In
that
resea
rc
h
us
e
d
exte
n
de
d
interm
ediat
e
cro
ss
over
a
nd
rando
m
m
utatio
n
f
or
operato
r
G
A.
I
n
this
s
tud
y,
we
will
us
e
hy
br
id
GASA
wh
ic
h
pro
pose
d
by
Faty
an
osa
.
oth
e
r
tha
n
that,
will
pro
posed
proc
ed
ur
e
hybri
d
GASA
and c
om
par
e w
it
h
Fat
ya
no
sa
’s GAS
A
m
et
ho
d.
Evaluation Warning : The document was created with Spire.PDF for Python.
In
t J
Elec
&
C
om
p
En
g
IS
S
N: 20
88
-
8708
Hyb
ri
d Ge
netic
Alg
or
it
hms
and Si
mu
l
ated A
nn
e
alin
g
fo
r
.
.. (
Ama
li
a K
ar
ti
ka
Ariy
an
i
)
4715
3.
DESCRIPTI
ON OF THE
PROBLE
M
Veh
ic
le
Ro
uting
P
r
ob
le
m
wi
th
Ti
m
e
W
in
dows
(V
RP
T
W)
is
on
e
of
the
m
os
t
stud
ie
d
VRP
issues
.
The
c
oncept
of
VRPT
W
i
s
th
e
de
velo
pm
ent
of
VRP
pro
bl
e
m
s
with
t
he
a
dd
it
io
n
of
new
co
ns
trai
nts
th
at
is
a
tim
e
wind
ows.
Thu
s
,
the
se
rvi
ce
of
each
c
ust
om
er
m
us
t
be
sta
rted
an
d
te
r
m
inate
d
base
d
on
t
he
tim
e
wi
ndow
s
that
ha
ve
bee
n
determ
ined.
I
f
the
ve
hicle
di
stribu
te
s
up
be
fore
the
c
us
t
om
er'
s
window
s
tim
e,
it
is
su
bj
ect
to
the
wait
in
g
poi
nts,
wh
il
e
the
distrib
ution
ve
hicle
that
com
es
after
th
e
cus
tom
er
'
s
windows
ti
m
e
exp
ires,
th
e
n
it
is su
bject
t
o pen
al
ty
poi
nts
[16].
The
pur
pose
of
VRP
T
W
'
s
op
ti
m
iz
ation
pro
blem
is
to
determ
ine
the
shortest
dista
nce
route
to
m
ini
m
iz
e
trave
l
cost
a
nd
the
nu
m
ber
of
ve
hi
cl
es
without
breaki
ng
the
ti
m
e
windows
a
nd
ve
hicle
ca
pa
ci
ty
.
The
j
ou
rn
ey
st
ar
ts
from
a
sing
le
de
pot
then
go
e
s
to
eac
h
of
the
scat
te
re
d
custom
ers.
T
he
distrib
utio
n
veh
ic
le
m
us
t
reach
th
e
cust
om
er'
s
place
betwee
n
the
c
us
tom
er's
tim
e
windows.
The
total
weig
ht
of
th
e
good
s
trans
ported
s
houl
d
not
excee
d
the
capaci
ty
of
the
veh
ic
le
.
Veh
i
cl
es
that
hav
e
finish
e
d
visit
ing
the
cu
stom
er
node
m
us
t
return
to
the
init
ia
l
dep
ot
withi
n
the
prede
fin
ed
tim
e
wind
ows
de
pot
[17].
Mult
i
-
Trip
V
ehicl
e
Rou
ti
ng
P
r
ob
l
e
m
(MT
-
VRP
)
is
one
of
the
dev
el
op
m
ent
of
cl
assic
VRP
pro
blem
s
in
wh
ic
h
a
ve
hicle
travels
on
m
ulti
ple
routes
within
a
c
ertai
n
tim
e
fr
am
e
[1
8].
T
he
MT
-
VRP
is
a
s
pecial
veh
ic
le
route
issue
where
each
veh
ic
le
ca
n
s
er
ve
m
or
e
tha
n
one
tri
p
wh
e
re
e
ach
tri
p
sta
rts
f
ro
m
the
de
pot,
passes
seve
ral
custom
ers
an
d
en
ds
up
in
t
he
depot
an
d
the
total
c
us
tom
er
dem
a
nd
sho
uld
not
exceed
cap
aci
ty
[19].
Mult
i
-
Trip
VRPT
W
app
li
es
MT
-
VRP
conc
ept but the
re is
addit
ion
of ti
m
e lim
it
o
f
wi
ndows in
it
.
In
t
his
stu
dy,
hybri
dizat
io
n
of
ge
netic
al
gorithm
and
s
i
m
ulate
d
ann
e
al
ing
is
us
ed
to
so
l
ve
the
pro
blem
of
Mult
i
-
Trip
Ve
hicle
Rou
ti
ng
P
rob
lem
with
Ti
m
e
W
in
dows
w
hich
is
ab
out
sch
edu
li
ng
to
uri
st
rout
e
in Bany
uw
a
ngi R
ege
ncy. T
here are se
ve
ral c
onditi
ons that l
i
m
i
t t
he
stu
dy.
a.
Ther
e
are
19 t
ourist
data an
d 1
4 ho
te
l
data
use
d
in
this
resea
rch
b.
Trips
are se
t f
or
3 days an
d oc
cup
y
only
1
ho
te
l as de
po
t
c.
Trav
el
is
ass
um
ed
us
i
ng the
car a
nd road
cond
it
io
ns
sm
oo
t
hly
d.
The ro
und t
rip
tim
e is assum
e
d
to
b
e
the
sam
e
e.
Ti
m
e w
indo
ws
tourist
s start at
05.00
-
19.00
f.
Fit
ness value
de
rive
d
f
r
om
tot
al
trav
el
ti
m
e a
nd total
to
ur
ist
at
tract
ion
ca
n be
visit
ed w
it
hi
n
3 days
g.
The
m
or
e to
uri
st o
bj
ect
s a
nd t
he
le
ss t
rav
el
ti
m
e required wi
ll
r
esult i
n g
rea
te
r
fitness
v
al
ue
h.
Penalty
value
is
the
total
va
lue
of
vio
la
ti
on
a
gainst
ti
m
e
wind
ow
s
to
uri
st
obj
ect
an
d
tim
e
wind
ows
tourist
s
4.
ALGO
RITH
M USED
4.1.
Genetic
Algor
ithm
Gen
e
ti
c
Algor
it
h
m
is
on
e
of
searc
h
a
nd
optim
iz
at
ion
al
gorithm
s
base
d
on
natur
al
s
el
ect
ion
a
nd
gen
et
ic
m
echan
ism
s
that
hav
e
bee
n
de
velop
e
d
as
eff
ect
ive
op
ti
m
iz
ati
on
a
ppro
a
ches
to
so
lve
co
m
plex
pro
blem
s [
20
]
. G
eneti
c alg
or
i
thm
ap
pr
oac
h ca
n
hel
p
fin
d
a
g
oo
d
so
l
ut
ion
for
com
plex
m
at
hem
atical
p
ro
blem
s
su
c
h
as
the
V
RP
pro
blem
[21].
Mult
i
-
T
rip
Pr
oble
m
s
VRP
T
W
is
a
c
om
bin
at
or
ia
l
optim
iz
at
ion
prob
le
m
by
form
ing
the
app
r
opriat
e
repre
pen
ta
ti
on
so
l
ution
that
will
get
op
tim
al
so
luti
on
in
the
us
e
of
ge
netic
al
gorithm
for
M
ult
i
-
Trip
VRPT
W
pro
blem
.
In
t
his
stud
y,
we
use
proce
dure
of
ge
netic
al
gori
thm
that
is
show
n
i
n
Figure
1 [1
]
.
Init
iali
za
t
ion
p
ar
amete
r
GA
GA
(stopping
co
ndit
ion)
{
Cal
culat
ing
fi
tnes
s of
chr
om
osom
e
//
doing
rep
rodu
ct
ion
to
g
et ne
w
chr
om
osom
e
Cross
over
using
One
-
Cut
-
Point
C
ross
over
Mutat
ion
using
Resiproc
a
l
Ex
ch
ange
Mut
at
ion
Sele
c
ti
on
using
Repl
a
ce
m
ent Sel
ec
t
ion
}
Figure
1
.
Proce
dure
of g
e
netic
algorit
hm
4.1.1.
Ch
r
om
osom
Repre
sen
tation
In
case
a
bout
VRPT
W,
c
hro
m
os
o
m
e
is
an
encode
of
de
sti
nations
perm
uta
ti
on
that
are
visit
ed
by
tourist
.
Usi
ng
per
m
utati
on
w
il
l
include
al
l
of
t
ourism
destinat
ion
s
s
o
ge
netic
al
goritm
can
sea
rch
optim
u
m
so
luti
on.
Ch
rom
os
o
m
r
epr
ese
ntati
on
c
an
b
e
seen i
n
Fi
gure
2.
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
S
N
:
2088
-
8708
In
t J
Elec
&
C
om
p
En
g,
V
ol.
8
, N
o.
6
,
Dece
m
ber
2
01
8
:
4713
-
4723
4716
Figure
2
.
Ch
rom
os
o
m
e rep
res
entat
ion
Af
te
r
getti
ng
t
he
c
hrom
os
ome
res
ults,
ne
xt
ste
p
is
to
cal
cu
la
te
the
fitness
value
of
the
c
hrom
os
om
e
that
has
bee
n
f
or
m
ed.
T
he
fitness
cal
c
ulati
on b
egi
ns
as
if
it
inserts
a h
otel i
nd
e
x
on
a
c
hrom
os
om
e.
In
t
he
first
gen
e
of
the
c
hrom
os
om
e
inser
te
d
the
ind
e
x
of
the
sel
ect
ed
ho
te
l,
f
or
exa
m
ple
the
ind
ex
of
the
sel
ect
ed
hote
l
is
4,
t
hen
t
he
hote
l
ind
e
x
4
is
inserted
at
the
beg
i
nn
i
ng
of
t
he
c
hrom
os
ome
.
O
nce
th
e
ho
te
l
ind
ex
is
ins
erted,
the
ne
xt
ste
p
is
to
insert
the
tourist
obj
ect
gen
e
one
by
one
on
the
ch
r
om
os
o
m
e,
if
the
total
du
rati
on
of
the
tourist
visit
ha
s
excee
ded
14.
45
t
he
n
the
i
nse
rtion
of
the
hote
l
ind
e
x
4
ret
urns.
A
fter
the
insertio
n
of
t
he
ho
te
l
ind
e
x
the
proc
ess
of
e
nterin
g
the
ret
urn
to
uri
st
ge
ne
is
done
with
the
sa
m
e
pr
oce
ss
unti
l
the
day
lim
it
is
e
ntere
d.
I
n
thi
s
stud
y
it
is
assum
ed
that
the
trip
is
done
within
3
days
so
that
the
pr
oc
ess
of
e
nterin
g
the
chrom
os
om
e
gen
e
is
done
i
n
3
trips
.
I
f
the
proces
s
of
enter
ing
the
ge
ne
ha
s
been
c
om
plete
d,
the
n
the
end
of
the chr
om
os
ome
is re
-
inse
rt
ed
the
hotel
in
dex 4.
The
cal
culat
io
n
of
fitne
ss
va
lue
is
influ
e
nc
ed
by
the
pa
ra
m
et
ers
of
tra
ve
l
tim
e
(
ti
me
),
penalty
,
an
d
nu
m
ber
of
t
ourist
destinat
io
ns
that
can
be
visit
ed.
In
t
he
tim
e
par
am
e
ter
the
gr
eat
e
r
t
he
value
of
ti
m
e,
the
sm
a
ll
er
the
fitness
value
obt
ai
ned
so
eq
uat
ion
1
is
use
d
di
vid
ed
by
1
+
tim
e,
the
ad
diti
on
of
value
1
in
th
e
tim
e
par
a
m
et
e
r
is
us
ed
t
o
avo
i
d
infi
nity
value
if
there
i
s
an
occ
urrenc
e
of
ti
m
e
value
equ
al
to
0.
In
the
increasin
g
pe
na
lt
y
par
am
et
er
pen
al
ty
val
ue
will
reduce
th
e
fitness
value
s
o
th
at
t
he
pen
a
lt
y
equ
at
ion
is
us
e
d
m
ul
ti
plied b
y t
he value
-
1 f
or
the p
e
nalty
v
al
ue
to
b
e
m
inu
s,
if the
re is
no penalt
y t
hen
t
he penalt
y val
ue
e
qu
a
l
to
0.
I
n
the
pa
ram
et
er
of
the
nu
m
ber
of
at
t
racti
on
s
,
the
le
ss
num
ber
of
visit
ed
to
ur
ist
sit
es
will
red
uc
e
the
value
of
fitnes
s
so
that
the
va
lue
of
t
he
nu
m
ber
of
to
ur
ist
at
tract
ion
s
di
vi
ded
by
the
nu
m
ber
of
ideal
tourist
at
tract
ion
that
can
be
visit
ed
within
3
days
as
m
any
as
10
at
tract
ion
s
[
1].
In
this
stu
dy
Eq
uation
1
sho
ws
th
e
functi
on
to
cal
culat
e
the
fitn
ess
value
.
T
he
exam
ple
of
c
hrom
os
om
e’s
fitness
cal
cula
ti
on
can
be
se
en
in
Table
1.
=
1
1
+
+
(
−
1
)
+
10
,
(
1
)
Table
1
.
C
hro
m
os
o
m
e’s
Fit
ne
ss Cal
culat
ion
Ch
romos
o
m
e
Path
No
d
e
Starting
Ti
m
e
Tr
av
el
Ti
m
e
Arr
iv
e
Ti
m
e
Ear
liest
Ti
m
e
W
aitin
g
Ti
m
e
Start
Visit
Du
ration
Fin
ish
Latest
Ti
m
e
Pen
alty
1
4
,8
5
:0
0
2
:0
6
7
:0
6
5
:0
0
0
:0
0
7
:0
6
3
:0
0
1
0
:0
6
1
9
:0
0
0
:0
0
8
,17
1
0
:0
6
1
:1
2
1
1
:1
8
8
:0
0
0
:0
0
1
1
:1
8
3
:0
0
1
4
:1
8
1
7
:0
0
0
:0
0
1
7
,1
1
4
:1
8
1
:3
2
1
5
:5
0
5
:0
0
0
:0
0
1
5
:5
0
3
:0
0
1
8
:5
0
1
7
:0
0
1
:5
0
1
,4
1
8
:5
0
2
:2
1
2
1
:1
1
5
:0
0
0
:0
0
2
1
:1
1
0
:0
0
2
1
:1
1
1
9
:0
0
2
:1
1
2
4
,12
1
:0
0
0
:3
7
1
:3
7
5
:0
0
3
:2
3
5
:0
0
3
:0
0
8
:0
0
1
9
:0
0
0
:0
0
1
2
,2
8
:0
0
2
:0
2
1
0
:0
2
5
:0
0
0
:0
0
1
0
:0
2
3
:0
0
1
3
:0
2
1
9
:0
0
0
:0
0
2
,3
1
3
:0
2
1
:0
2
1
4
:0
4
5
:0
0
0
:0
0
1
4
:0
4
3
:0
0
1
7
:0
4
1
7
:0
0
0
:0
4
3
,4
1
7
:0
4
2
:5
7
2
0
:0
1
5
:0
0
0
:0
0
2
0
:0
1
0
:0
0
2
0
:0
1
1
9
:0
0
1
:0
1
3
4
,14
5
:0
0
0
:4
0
5
:4
0
8
:0
0
2
:2
0
8
:0
0
3
:0
0
1
1
:0
0
1
9
:0
0
0
:0
0
1
4
,10
1
1
:0
0
2
:1
9
1
3
:1
9
5
:0
0
0
:0
0
1
3
:1
9
3
:0
0
1
6
:1
9
1
9
:0
0
0
:0
0
1
0
,4
1
6
:1
9
2
:1
7
1
8
:3
6
5
:0
0
0
:0
0
1
8
:3
6
0
:0
0
1
8
:3
6
1
9
:0
0
0
:0
0
Total
1
9
:0
5
5
:0
6
f
itn
ess
-
4
.21
0
1
Af
te
r
g
et
ti
ng t
he result
of tra
vel tim
e and p
e
nalty
, f
it
ness
is cal
culat
ed by
us
in
g
E
quat
ion
2
.
=
1
1
+
19
.
05
+
(
5
.
06
−
1
)
+
8
10
(2)
=
−
4
.
2101
4.12.
Cro
s
sove
r
In
this
stu
dy
,
cro
ss
over
pr
oce
ss
us
es
one
-
c
ut
po
int
cr
os
s
ov
e
r
[22].
The
nu
m
ber
of
of
f
spring
ob
ta
ine
d
from
crosso
ver
ra
te
x
pops
ize
.
In
t
his
m
et
ho
d,
the
c
ut
point
value
will
be
ge
ner
at
e
d
ra
ndom
ly
with
a
range
of
values
0 to
c
hrom
os
om
e leng
th
-
1.
Evaluation Warning : The document was created with Spire.PDF for Python.
In
t J
Elec
&
C
om
p
En
g
IS
S
N: 20
88
-
8708
Hyb
ri
d Ge
netic
Alg
or
it
hms
and Si
mu
l
ated A
nn
e
alin
g
fo
r
.
.. (
Ama
li
a K
ar
ti
ka
Ariy
an
i
)
4717
4.1.3.
Mut
at
io
n
M
utati
on
proc
ess
us
i
ng
Re
ci
procal
E
xchan
ge
M
utati
on
[
22
]
.
The
num
ber
of
offs
pr
in
g
m
utati
ons
resu
lt
s
f
r
om
m
uta
ti
on
r
ate
x
pops
ize
.
In
t
he
process
of
m
utati
on
the
firs
t
ste
p
is
to
sel
ect
1
ind
i
vidu
al
at
rand
om
.
In
the
Re
ci
procal
Ex
change
M
utati
on
m
et
ho
d,
2
r
andom
ly
sel
ected
in
de
xes
will
be
sel
ect
ed
w
it
h
a
range
of
0
-
c
hrom
os
om
e
le
ng
th
-
1
a
nd e
xch
a
nge
bo
t
h values
of the selec
te
d
gen
e
in
dex.
4.1.4.
Sele
ctio
n
At
the
sel
ect
io
n
ste
p,
the
m
eth
od
t
o
be
us
ed
is
Re
placem
ent
Sele
ct
ion
bec
ause
ca
n
getti
ng
m
axi
m
u
m
so
luti
on
[
1]
.
T
he
pr
i
nciple
of
the
rep
la
cem
ent
sel
ect
ion
operator
is
th
at
if
the
off
sprin
g
re
producti
on
has
a
fitness
value
gr
ea
te
r
tha
n
t
he
par
e
nt
the
n
offs
pr
i
ng
wil
l
rep
la
ce
t
he
par
e
nt
Offspri
ng
obta
ine
d
from
the
m
uta
ti
on
proce
ss
will
rep
la
ce
the
par
e
nt
if
t
he
offs
pr
i
ng
ha
s
gr
eat
er
fitne
ss
than
the
pa
r
ent,
if
the
cros
so
ve
r
process
that
ha
s 2
pa
ren
t
s
, th
e
n
sel
ect
ed
1 pa
ren
t
with t
he
lo
west f
it
ne
ss
va
lue
[
23]
, [2
4]
.
4.2.
Simula
ted A
nnea
ling
Si
m
ulate
d
an
ne
al
ing
(
SA)
c
on
sist
s
of
optim
izing
the
“
m
el
t
ing
”
pr
oc
ess
with
the
appr
opriat
e
tem
per
at
ur
e
,
then
lo
wer
i
ng
the
te
m
per
at
ure
slow
ly
un
ti
l
it
"fr
eeze
s"
a
nd
no
furthe
r
process
[
25
]
.
At
t
he
beg
i
nn
i
ng
of
the
process
,
the
m
et
a
l
will
be
heated.
T
he
n,
the
tem
per
at
ur
e
is
lower
ed
slo
wly
to
avo
id
cr
acki
ng
to
m
ini
m
iz
e
the
ene
rg
y
us
e
d
[26].
S
A'
s
flo
w
ca
n
helps
G
A'
s
ind
i
vidual
to
get
co
nver
ge
nce
t
o
a
m
inim
u
m
global
[27].
T
he
highe
r
the
tem
per
at
ure
an
d
the
lowe
r
th
e
value
of
the
so
luti
on,
the
gr
eat
er
the
c
ha
nce
of
receivin
g
a
les
s tha
n op
ti
m
al
so
luti
on
[28].
Figure
3
s
hows
pro
ce
dure
of s
i
m
ulate
d
an
nea
li
ng
wh
ic
h
is
use
d.
Init
iali
za
t
ion
p
ar
amete
r
SA
SA
(stopping
co
ndit
ion)
{
If
(t
≥
final
te
m
per
at
ur
e
{
d
1
=
fit
n
ess (c
hro
m
oso
m
e)
S
n
=
Crea
t
ing
Nei
ghborhood
Solut
ion
using Scr
amble
Mut
at
ion
d
2
=
fit
n
ess (S
n
)
If
(d <
d
1
)
{
Init
ial
α
=
r
andom [0, 1]
Init
ial
p
=
Probabi
li
t
y
If
(α
<
p
)
{
S
=
S
n
}
}
el
se
{
S
=
S
n
}
}
}
Figure
2
.
Proc
udure
of
sim
ul
at
ed
an
neali
ng
Wh
e
re
S
is
m
a
xim
u
m
chr
omoso
m
e
w
hich
i
s
get
from
resul
t
of
G
A.
d
1
a
nd
d
2
a
re
fit
ness
of
S
a
nd
S
n
.
S
n
is
ne
w
chrom
os
om
e
fr
om
Neigh
bor
hood
scram
ble
m
uta
ti
on
proce
ss
and
resu
lt
s
olu
ti
on
of
SA
i
s
S
.
Pro
bab
il
ty
value
in S
A
is
ge
ner
a
te
d
usi
ng
E
qu
a
ti
on
3
.
=
1
(
2
−
1
)
1
(
3
)
4.2.1.
Neig
hb
or
hood Searc
h
In
this
st
ud
y,
SA
will
ge
ne
r
at
e
neih
bor
hood
s
olu
ti
on
by
us
i
ng
Sc
ram
ble
Muta
ti
on
[
27]
.
At
this
op
e
rato
r
two
r
andom
ly
sel
ected
points
are
s
el
ect
ed
and
s
w
it
ch
the
chrom
os
om
e
po
sit
ion
between
tw
o
po
i
nts
rand
om
l
y
fo
r
e
xam
ple sh
ow i
n
Fi
gure
4
.
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
S
N
:
2088
-
8708
In
t J
Elec
&
C
om
p
En
g,
V
ol.
8
, N
o.
6
,
Dece
m
ber
2
01
8
:
4713
-
4723
4718
Figure
3
.
Scra
m
ble
m
utatio
n process
4.3.
Hy
brid
Gene
t
ic
A
lg
orithm
an
d
Simula
te
d Anne
aling
Ba
sed
on
goodne
ss
an
d
we
akn
e
ss
each
al
gorithm
,
com
b
inati
on
gen
et
ic
al
gorithm
and
sim
ulate
d
ann
eal
in
g
will
furn
is
h
weakn
ess
each
al
gori
thm
.
Ho
we
ver,
nee
ding
one
c
om
bin
at
ion
to
get
m
axi
m
u
m
resu
lt
.
GA
-
S
A1
is
proses
ed
m
et
hod
in
t
his
stu
dy,
wh
il
e
GA
-
SA2,
G
A
-
SA3
,
a
nd
G
A
-
S
A4
are
c
om
par
is
on
m
et
ho
d [
15
]
.
4.3.1.
GA
-
S
A1
Com
bin
at
ion
GA
-
S
A1
is
sta
rted
by
i
niti
al
i
zat
ion
G
A
para
m
et
er.
The
se
cond
proce
ss
i
s
r
unning
S
A
with
init
ia
l
so
luti
on
us
in
g
ch
r
om
os
om
e
with
best
dan
worst
fitness.
It
is
stop
un
ti
l
it
erati
on
m
axi
m
u
m
i
n
G
A.
Pseudoc
ode
of
GA
-
S
A1 can
be seen
in
Fig
ure
5.
Init
iali
za
t
ion
p
ar
amete
r
GA
-
SA
GA
(stopping
co
ndit
ion)
{
//
running
GA
proc
ess
S
=
chr
om
osom
e
with
best
fit
n
ess
of
GA
SA
(stopping
co
ndit
ion)
{
S
=
chr
om
osom
e
with
best
fit
n
ess
of
GA
//
running
SA
proc
ess
}
S
=
chr
om
osom
e
with
wors
t
fi
tne
s
s of
GA
SA
(stopping
co
ndit
ion)
{
S
=
chr
om
osom
e
with
best
fit
n
ess
of
GA
//
running
SA
proc
ess
}
}
Figure
4
.
GA
-
SA1
process
4.3.2.
GA
-
S
A2
GA
-
S
A2
proc
ess
sta
rts
by
ru
nnin
g
G
A
first
un
ti
l
st
opping
co
ndit
ion
,
the
n
S
A
pr
oce
ss
is
execu
te
d
[15].
In
S
A
process,
the
num
ber
of
popula
ti
ons
use
d
is
obta
ine
d
base
d
on
rand
om
rate
of
the
best
ind
ivi
du
al
s
. P
s
eudoc
od
e
of
G
A
-
S
A
2
ca
n be
seen i
n
Fi
gure
6.
I
nit
iali
za
t
ion
p
ar
amete
r
GA
-
SA
GA
(stopping
co
ndit
ion)
{
//
running
GA
proc
ess
}
Chrom
oso
m
e
=
sele
c
t
(r
andom ra
t
e,
GA
chr
om
osom
e)
SA
(stopping
co
ndit
ion)
{
//
running
SA
proc
ess
}
Figure
5
.
GA
-
SA2
process
Evaluation Warning : The document was created with Spire.PDF for Python.
In
t J
Elec
&
C
om
p
En
g
IS
S
N: 20
88
-
8708
Hyb
ri
d Ge
netic
Alg
or
it
hms
and Si
mu
l
ated A
nn
e
alin
g
fo
r
.
.. (
Ama
li
a K
ar
ti
ka
Ariy
an
i
)
4719
4.3.3.
GA
-
S
A3
GA
-
S
A3
proce
ss
sta
rts
by
runn
i
ng
S
A
al
go
rithm
first
un
ti
l
stop
pi
ng
c
onditi
on
,
the
n
G
A
process
is
execu
te
d
[
15]
.
In
GA
proces
s
,
the
num
ber
of
popula
ti
ons
use
d
is
obta
ine
d
based
on
ra
nd
om
rate
of
the
best
ind
ivi
du
al
s
. P
s
eudoc
od
e
of
G
A
-
S
A
3
ca
n be
seen i
n
Fi
gu
re
7.
Init
iali
za
t
ion
p
ar
amete
r
GA
-
SA
SA
(stopping
co
ndit
ion)
{
//
running
SA
proc
ess
}
Chrom
oso
m
e
=
sele
c
t
(r
andom ra
t
e,
GA
chr
om
osom
e)
GA
(stopping
co
ndit
ion)
{
//
running
GA
proc
ess
}
Figure
6
.
GA
-
SA3
process
4.3.4.
GA
-
S
A4
GA
-
S
A2
proce
ss
sta
rts
by
runn
i
ng
GA
fi
rst
fo
r
ge
ne
rati
ons,
then
S
A
pro
cess
is
execu
te
d
an
d
afte
r
SA
pr
ocess
is
com
plete
d
it
w
il
l
ru
n
G
A
agai
n.
A
fter G
A
process
on
the
fi
r
st
it
erati
on
is
com
plete
d,
ind
i
vidual
is
ta
ken
as
m
uch
as
ra
ndom
rate
of
the
best
ind
i
vi
du
al
t
o
be
processe
d
by
SA
.
Af
te
r
S
A
process
is
c
omplet
ed
,
it
will
ta
ke
th
e
ind
ivid
ual
as
m
uch
as
the
ran
dom
rate
of
the
best
in
div
id
ual
from
the
SA
proces
s
to
be
processe
d
bac
k
by
G
A
[15].
The
recurre
nce
will
stop
after
stoppin
g
c
ondi
ti
on
is
f
ound.
Pseu
doco
de
of
G
A
-
SA4 ca
n be see
n
in
Fig
ure
8.
Init
iali
za
t
ion
p
ar
amete
r
GA
-
SA
W
hil
e
(Stopping
Condit
ion
)
{
If
(GA
condition
)
{
GA
()
{
//
running
GA
proc
ess
}
}
If
(SA
condition
)
{
SA
()
{
//
running
SA
proc
ess
}
}
Chrom
oso
m
e
=
sele
c
t
(r
andomrat
e
,
chr
om
osom
e)
}
Figure
7
.
GA
-
SA4
process
5.
E
X
PERI
MEN
TAL RES
UL
T AND
DI
SCUSSI
ON
5.1.
Experim
en
ta
l
Scenari
o
The
e
xp
e
rim
ental
scenario
wi
ll
be
done
by
c
om
par
ing
fitne
ss
res
ults
ge
ne
rated
f
r
om
eac
h
al
gorith
m
ie
GA
,
S
A,
G
A
-
S
A
1,
GA
-
S
A2,
G
A
-
SA3,
and
GA
-
S
A
4.
In
the
GA
-
S
A
2,
G
A
-
S
A3,
an
d
G
A
-
SA4
al
gorithm
s,
a
ran
dom
par
am
et
er
of
0.1
-
1
is
us
ed
to
sel
ect
the
nu
m
ber
of
ind
i
viduals
ra
ndom
ly
to
be
pr
oces
sed
f
ur
t
he
r.
I
n
GA
-
S
A2,
if
th
e
rando
m
rate
us
e
d
is
0.
1
t
he
n
only
10
%
of
the
ov
e
rall
GA
popu
la
ti
on
wil
l
be
re
-
op
ti
m
ized
by
SA
[
15]
.
To
know
t
he
m
os
t
op
tim
al
ran
dom
rate,
it
will
be
te
ste
d
ra
ndom
rate
par
am
et
ers
in
ad
van
ce
s
o
i
t
can
be
known
ra
ndom
rate
valu
e
that
can
pro
du
ce
the
m
os
t
optim
al
fitness
value
.
T
he
pa
ram
et
ers
us
ed
in
this
exp
e
rim
ent can
be se
en
in
Ta
bl
e 2
.
Tab
le
2
.
E
cpe
rim
ent Par
am
et
e
rs
Para
m
eters
Valu
e
GA
Po
p
u
latio
n
Size
400
Maxi
m
u
m
Gene
rat
io
n
800
cros
so
v
er
rate
0
.3
m
u
t
atio
n
r
ate
0
.1
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
S
N
:
2088
-
8708
In
t J
Elec
&
C
om
p
En
g,
V
ol.
8
, N
o.
6
,
Dece
m
ber
2
01
8
:
4713
-
4723
4720
Tab
le
2
.
E
cpe
rim
ent Par
am
et
e
rs
Para
m
eters
Valu
e
SA
in
itial te
m
p
e
rature
0
.9
co
o
lin
g
r
ate
0
.9
n
ew so
lu
tio
n
acc
ep
tan
ce
p
rob
ab
ility
co
ef
f
ic
ien
t
200
The
ex
per
im
ents
will
be
carried
out
with
th
e
sa
m
e
env
iro
nm
ent
us
ing
J
ava
pro
gr
am
m
ing
la
ngua
ge
and
c
om
pu
ta
tio
n
ti
m
e
fo
r
ea
ch
al
gorithm
f
or
60
seco
nds
and
will
be
r
epeate
d
10
ti
m
es
to
pr
oduc
e
a
fair
com
par
ison.
5.2.
Result
and
Di
scussion
The
first
e
xp
e
r
i
m
ent
is
con
du
ct
ed
to
determ
ine
the
optim
a
l
ran
dom
rate.
The
res
ult
of
r
andom
rate
par
am
et
er
te
st
can
be
seen
i
n
Table
3
-
5.
Ta
bl
e
3
sho
ws
that
rand
om
rate
0.
8
yi
el
ds
the
hig
he
st
avera
ge
f
it
ness
value
of
1.0
323.
I
n
Ta
ble
4,
the
highest
fi
tness
a
v
era
ge
is
obta
ined
w
hen
the
ra
ndom
rate
is
1
w
hich
is
1.087
2.
I
n
Ta
bl
e 5
, t
he hig
hes
t avera
ge fit
nes
s v
al
ue
is
ob
ta
i
ned whe
n
the
ra
ndom
r
at
e is 0
.5
w
hich
is
1.0886.
Table
3
.
GA
-
S
A2’s
Ra
ndom
Rat
e Result
ra
n
d
o
m
ra
te
Fitn
ess
Value
Fitn
ess
Av
erage
Iter
atio
n
1
2
3
4
5
6
7
8
9
10
0
.1
1
.08
9
4
1
.00
6
7
1
.00
8
5
1
.00
6
4
1
.00
8
5
1
.00
9
3
1
.00
6
4
1
.01
0
1
1
.00
4
9
1
.00
8
5
1
.01
5
9
0
.2
1
.00
9
4
1
.00
9
4
0
.99
8
5
1
.00
5
9
1
.08
1
2
0
.99
8
8
1
.00
9
3
0
.99
8
8
1
.00
9
3
1
.08
9
1
1
.02
1
0
0
.3
1
.00
8
5
1
.08
9
1
1
.00
6
4
1
.00
7
4
1
.00
5
0
1
.00
9
3
1
.00
9
6
1
.00
9
3
1
.08
1
2
1
.00
8
5
1
.02
3
4
0
.4
1
.00
9
3
1
.00
5
9
1
.00
9
6
1
.00
9
3
0
.99
5
6
1
.00
6
0
1
.00
9
6
1
.08
9
7
1
.08
9
7
1
.00
9
3
1
.02
3
4
0
.5
1
.00
9
3
1
.00
9
4
1
.00
5
3
1
.00
4
9
1
.00
9
3
1
.00
5
2
1
.01
0
1
1
.01
0
1
1
.00
6
4
1
.01
0
1
1
.00
8
0
0
.6
1
.00
9
6
1
.00
6
0
1
.00
5
3
1
.00
9
3
1
.00
8
0
1
.01
0
1
1
.00
9
3
1
.01
0
1
0
.99
8
2
1
.00
5
9
1
.00
7
2
0
.7
1
.01
0
1
1
.01
0
1
1
.00
5
4
1
.00
8
5
1
.00
9
4
1
.00
9
4
1
.00
5
4
1
.00
7
4
1
.00
9
6
1
.00
8
0
1
.00
8
3
0
.8
1
.00
9
4
1
.00
7
4
1
.00
6
4
1
.08
6
8
1
.00
9
6
1
.08
9
1
1
.00
9
6
1
.00
6
7
1
.08
8
3
1
.00
9
4
1
.03
2
3
0
.9
1
.00
7
4
1
.00
8
5
1
.09
0
0
1
.01
0
1
1
.00
9
3
1
.01
0
1
1
.00
9
4
1
.00
5
2
1
.01
0
1
1
.00
6
4
1
.01
6
7
1
0
.99
8
5
1
.00
9
4
1
.00
9
4
1
.00
5
3
1
.00
5
2
1
.00
5
0
1
.00
5
4
1
.01
0
1
1
.08
8
0
1
.00
8
5
1
.01
4
5
Table
4
.
GA
-
S
A3’s
Ra
ndom
Rat
e Result
ra
n
d
o
m
ra
te
Fitn
ess
Value
Fitn
ess
Av
erage
Iter
atio
n
1
2
3
4
5
6
7
8
9
10
0
.1
0
.99
9
1
0
.99
5
9
0
.99
8
6
1
.00
4
7
0
.98
8
5
0
.76
6
2
1
.00
9
4
0
.99
7
3
1
.00
5
7
0
.99
8
4
0
.97
6
4
0
.2
0
.88
1
6
1
.00
9
3
0
.99
7
8
0
.99
8
3
1
.00
8
5
1
.08
2
7
0
.76
1
5
0
.85
5
1
0
.85
7
3
1
.08
8
8
0
.95
4
1
0
.3
1
.00
5
3
0
.88
0
3
0
.81
8
2
0
.86
9
6
0
.86
9
4
0
.85
7
1
1
.00
4
5
0
.86
1
3
0
.76
2
3
1
.00
6
2
0
.89
3
4
0
.4
1
.08
9
1
0
.80
0
0
0
.86
4
8
0
.67
0
6
1
.00
6
7
0
.86
0
9
0
.99
6
5
1
.00
6
4
1
.00
9
6
0
.86
1
4
0
.91
6
6
0
.5
0
.78
3
3
0
.77
3
9
1
.00
5
4
0
.86
4
4
1
.00
4
6
0
.86
5
8
0
.86
1
7
1
.00
5
7
1
.00
4
3
0
.85
8
3
0
.90
2
7
0
.6
1
.08
6
4
1
.00
6
7
0
.78
0
0
1
.00
5
3
0
.76
6
2
0
.78
7
6
0
.76
9
4
1
.00
7
6
0
.86
8
8
1
.00
9
3
0
.90
8
7
0
.7
0
.88
9
9
0
.86
6
4
0
.84
0
2
1
.00
8
0
0
.87
9
6
0
.88
1
0
1
.00
6
4
1
.00
5
2
0
.85
5
4
0
.87
0
2
0
.91
0
2
0
.8
1
.00
6
0
0
.86
4
9
1
.00
6
4
0
.91
5
3
0
.86
5
5
0
.87
5
1
1
.00
8
2
0
.85
4
3
0
.86
9
9
0
.86
6
1
0
.91
3
2
0
.9
0
.86
9
4
0
.95
7
0
1
.00
4
9
0
.96
5
4
0
.86
1
9
0
.99
9
3
0
.96
0
4
0
.86
6
6
0
.88
1
6
0
.87
0
0
0
.92
3
7
1
1
.08
8
8
1
.08
6
7
1
.09
0
0
1
.08
8
8
1
.08
6
7
1
.08
6
8
1
.08
6
6
1
.08
6
7
1
.08
8
8
1
.08
2
2
1
.08
7
2
Table
5
.
GA
-
S
A3’s
Ra
ndom
Rat
e Result
ra
n
d
o
m
ra
te
Fitn
ess
Value
Fitn
ess
Av
erage
Iter
atio
n
1
2
3
4
5
6
7
8
9
10
0
.1
1
.09
1
.08
1
2
1
.08
1
2
1
.08
6
8
1
.08
8
8
1
.08
8
8
1
.08
6
7
1
.08
8
8
1
.08
8
3
1
.08
8
8
1
.08
6
9
0
.2
1
.08
6
7
1
.08
6
8
1
.08
6
8
1
.08
8
8
1
.09
1
.08
8
3
1
.08
8
8
1
.08
2
6
1
.08
8
8
1
.09
1
.08
7
8
0
.3
1
.08
8
4
1
.09
1
.08
8
8
1
.08
9
7
1
.08
6
7
1
.08
8
4
1
.08
6
7
1
.08
9
1
1
.08
1
2
1
.08
6
7
1
.08
7
6
0
.4
1
.08
6
8
1
.08
6
7
1
.08
9
7
1
.08
6
8
1
.09
1
.08
6
7
1
.08
8
8
1
.08
6
7
1
.08
6
8
1
.08
9
4
1
.08
7
8
0
.5
1
.08
8
4
1
.08
9
7
1
.08
6
8
1
.08
9
7
1
.08
6
7
1
.08
6
8
1
.08
9
7
1
.09
1
.08
9
4
1
.08
9
4
1
.08
8
6
0
.6
1
.08
6
7
1
.08
6
8
1
.08
8
8
1
.08
8
8
1
.08
6
8
1
.08
6
8
1
.08
6
4
1
.08
9
4
1
.08
8
8
1
.08
9
7
1
.08
7
9
0
.7
1
.08
6
8
1
.08
8
4
1
.09
1
.08
6
8
1
.08
6
4
1
.08
9
4
1
.08
8
8
1
.08
9
4
1
.09
1
.08
6
7
1
.08
8
3
0
.8
1
.08
6
8
1
.08
8
8
1
.08
9
4
1
.08
1
2
1
.08
6
8
1
.08
9
4
1
.08
8
8
1
.08
1
2
1
.08
8
1
.08
6
8
1
.08
6
7
0
.9
1
.08
6
8
1
.08
9
4
1
.08
9
7
1
.08
6
8
1
.08
9
7
1
.08
8
8
1
.09
1
.08
8
8
1
.08
6
8
1
.08
9
7
1
.08
8
6
1
1
.08
9
4
1
.09
1
.08
6
7
1
.08
8
4
1
.08
6
8
1
.08
6
7
1
.08
8
8
1
.08
6
7
1
.08
9
4
1
.08
6
7
1
.08
7
9
Evaluation Warning : The document was created with Spire.PDF for Python.
In
t J
Elec
&
C
om
p
En
g
IS
S
N: 20
88
-
8708
Hyb
ri
d Ge
netic
Alg
or
it
hms
and Si
mu
l
ated A
nn
e
alin
g
fo
r
.
.. (
Ama
li
a K
ar
ti
ka
Ariy
an
i
)
4721
The
ne
xt
ex
pe
rim
ent
is
to
com
par
e
the
fitness
res
ults
of
eac
h
al
gorithm
us
in
g
GA
an
d
S
A
par
am
et
ers
an
d t
he
ra
ndom
r
at
e p
aram
et
ers
th
at
h
ave
b
ee
n
te
ste
d
pre
viously
. Testin
g
is
done 10
ti
m
es o
n ea
ch
al
gorithm
. Th
e
co
m
par
iso
n of t
he
al
go
rithm
can
be
see
n
i
n Table
6.
T
he
first
te
st
w
as
perform
ed
on
a
gen
et
ic
al
gorithm
.
Af
te
r
te
sti
ng
10
tim
es
,
the
aver
a
ge
fi
tness
sco
re
of
1.038
4
was
ob
ta
ine
d.
T
he
second
ex
pe
rim
ent
was
perf
or
m
ed
on
the
si
m
ulate
d
anneal
ing
al
gorith
m
and
pro
du
ce
d
a
n
a
ver
a
ge
fitness
of
1.077
2.
T
he
t
hird
te
st
was
perf
or
m
ed
on
the
G
A
-
SA1
hybri
diza
ti
on
a
nd
ge
ne
rated
an
a
ver
a
ge
fitness
of
1.
0888.
I
n
GA
-
S
A2
te
st,
the
a
ver
a
ge
fitn
ess
gai
n
is
1.0
135.
I
n
the
G
A
-
S
A
4
te
st
the
aver
a
ge
fitness
gain
of
1.0
124
a
nd
the
G
A
-
SA4
t
est
ob
ta
ine
d
t
he
aver
a
ge
fitne
ss
value
of
1.0
86
3.
Fr
om
the six
t
h t
est
o
f
t
he
al
gorithm
, th
e b
est
aver
a
ge fit
ness
v
al
ue
is
gen
e
r
at
ed
by t
he GA
-
SA1.
Figure
9
s
how
s
the
fitness
re
su
lt
s
obta
ine
d
by
each
al
gorit
hm
within
1
m
inu
te
.
The
cha
r
t
sh
ows
that
us
es
of
the
sim
ulate
d
a
nn
eal
in
g
al
gorit
hm
ca
n
sig
nifica
n
tl
y
increase
t
he
fit
ness
value
bec
ause
e
t
he
al
go
rithm
eff
ect
ively
ex
plo
it
s
local
se
arch
a
rea,
pro
vid
in
g
a
n
op
port
un
it
y
to
fin
d
the
best
fitness
fa
ste
r.
F
r
om
all
al
gorithm
s
te
st
ed
in
this
rese
arch,
G
A
-
S
A
1
and
GA
-
S
A4
al
go
rithm
can
produce
the
m
os
t
op
tim
al
f
it
ness
a
m
on
g
ot
her
a
lgorit
hm
s,
bu
t
GA
-
S
A4
ca
n
f
ind
t
he
best
fitness
c
hrom
os
om
e
fo
r
26
sec
ond
faster
tha
n
G
A
-
SA1
that
requi
res
39
seco
nd
to
fin
d
the
be
st
fitness
ch
ro
m
osom
e.
ho
we
ver,
GA
-
S
A4
is
st
il
l
of
te
n
stuck
on
it
s
local
searc
h w
hile G
A
-
SA1 c
an fin
d
the
opti
m
u
m
p
oin
t eve
n
th
ough it
ta
ke
s a lit
tl
e lon
ge
r.
Table
6
.
Re
s
ult Com
par
iso
n o
f
Al
gorithm
Iter
atio
n
Alg
o
rith
m
GA
SA
GA
-
SA1
GA
-
SA2
GA
-
SA3
GA
-
SA4
1
1
.09
0
0
1
.08
1
6
1
.08
8
4
1
.00
8
5
1
.08
9
1
1
.08
6
7
2
1
.01
0
1
1
.08
7
6
1
.08
6
8
1
.00
9
3
1
.08
6
6
1
.08
6
8
3
1
.08
8
8
1
.08
1
6
1
.09
0
0
1
.00
6
4
1
.08
9
1
1
.08
9
4
4
0
.99
8
2
1
.08
6
7
1
.08
9
4
1
.00
9
4
1
.08
6
7
1
.09
0
0
5
1
.00
6
0
1
.08
6
8
1
.08
6
8
1
.00
6
7
1
.08
6
7
1
.08
9
4
6
1
.00
9
4
1
.08
1
2
1
.08
9
4
1
.00
5
7
1
.08
9
7
1
.08
6
8
7
1
.08
8
3
1
.08
6
8
1
.08
9
4
1
.09
0
0
1
.08
2
2
1
.08
9
7
8
1
.00
5
4
1
.00
6
4
1
.09
0
0
1
.00
5
4
1
.09
0
0
1
.08
8
4
9
1
.08
9
1
1
.08
6
8
1
.08
9
7
1
.00
4
7
1
.08
8
8
1
.08
9
7
10
0
.99
8
4
1
.08
6
7
1
.08
8
4
1
.09
0
0
1
.08
9
1
1
.08
9
7
Fitn
ess
Av
erage
1
.03
8
4
1
.07
7
2
1
.08
8
8
1
.02
3
6
1
.08
7
8
1
.08
8
7
Fitn
ess
Mini
m
u
m
0
.99
8
2
1
.00
6
4
1
.08
6
8
1
.00
4
7
1
.08
2
2
1
.08
6
7
Fitn
ess
Maxi
m
u
m
1
.09
0
0
1
.08
7
6
1
.09
0
0
1
.09
0
0
1
.09
0
0
1
.09
0
0
Figure
8
.
Com
par
is
on Result
6.
CONCL
U
S
ION
This
st
ud
y
c
om
par
es
the
fitness
valu
es
ge
ner
at
e
d
by
ge
netic
al
gorith
m
and
sim
ulated
a
nneal
ing
al
gorithm
and
four
it
s
hy
br
i
di
zat
ion
m
od
el
s.
From
the
sixth
te
st
of
the
al
gorithm
,
the
best
ave
rag
e
fitness
value
is
ge
ner
a
te
d
by
the
G
A
-
SA1.
Howe
ver,
the
rati
o
o
f
fitness
values
ge
ner
at
e
d
by
G
A
-
SA1
a
nd
G
A
-
SA4
can
be
sai
d
to
be
cl
os
e
to
get
her
s
o
that
bot
h
G
A
-
SA1
an
d
G
A
-
SA4
ca
n
pro
duce
the
m
os
t
op
tim
al
aver
a
ge
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
S
N
:
2088
-
8708
In
t J
Elec
&
C
om
p
En
g,
V
ol.
8
, N
o.
6
,
Dece
m
ber
2
01
8
:
4713
-
4723
4722
fitness
val
ues
bu
t
GA
-
S
A4
c
an
fin
d
the
be
st
fitness
ch
rom
os
o
m
e
faster
than
G
A
-
S
A1.
Determ
inati
on
of
popul
at
ion
nu
m
ber
of
ge
net
ic
al
gorithm
t
o
be
optim
iz
e
d
a
gain
by
sim
ula
te
d
an
nea
li
ng
al
gorithm
ca
n
influ
e
nce
al
go
rithm
sp
eed
in
find
i
ng
best
resu
lt
.
I
n
GA
-
SA1,
popula
ti
on
s
re
-
optim
ized
by
the
sim
ulate
d
ann
eal
in
g
al
go
rithm
are
j
us
t
the
best
a
nd
worst
fitness
ch
r
o
m
os
o
m
es,
bu
t
in
GA
-
S
A
4,
th
e
popu
la
ti
on
is
10
%
of
the
popula
ti
on
with
the
be
st
chr
om
os
om
es,
the
m
or
e
chrom
os
om
es
op
tim
iz
ed
by
the
si
m
uated
ann
eal
ing
al
gorithm
,
the
faster
the
c
ha
nc
e
of
t
he
al
gori
thm
to
fin
d
the
m
os
t
op
ti
m
al
f
it
ness.
F
uth
e
r
r
esearch
,
to n
ee
d
s
K
-
Me
ans for clu
s
te
ring no
des
th
at
can
im
pr
ove
co
m
pu
ta
ti
on ti
m
e [30].
REFERE
NCE
S
[1]
Y.
P.
Anggodo,
et
al.,
"O
pti
m
a
ti
on
of
Multi
-
T
r
ip
Vehic
l
e
Rout
ing
Problem
with
Ti
m
e
W
indows
using
Gene
ti
c
Algorit
hm
,
"
Jou
rnal
of
En
vi
ron
mental
En
gineering
and
Sustain
able
Te
chnol
ogy
,
vol
.
3
,
no.
2,
pp.
92
-
97,
Nov
2016.
[2]
X,
Ji
and
K.
Yao,
“
Unce
rt
ai
n
Projec
t
Sch
edul
i
ng
Problem
with
Resourc
e
Co
nstrai
nts,
”
Jour
nal
of
Intelli
g
e
nt
Manufac
turing
,
vol.
28
,
issue
3,
pp.
575
-
580
,
Oc
t
2014.
[3]
C.
Verbe
ec
k
,
e
t
al.
,
“
A
Meta
heu
risti
c
Solu
ti
o
n
A
pproa
ch
for
The
Ti
m
e
-
Constrai
n
ed
Proje
ct
Sch
e
duli
ng
Problem,
”
OR
Spectrum
,
v
ol
39,
issue
2
,
pp
.
353
-
371
,
Mar
2
016.
[4]
S.
Sundar,
e
t
al
.
,
“
A
Hy
brid
Artif
ic
i
al
Be
e
Colon
y
Algorit
hm
for
T
he
Job
-
Shop
Schedul
ing
Probl
em
with
No
-
W
ai
t
Constrai
nt
,
”
So
ft Computing
,
vol
.
12,
issue
5
,
pp
.
1193
-
1202,
Mar
2017.
[5]
K.
Ghos
ei
ri
and
S.
F.
Ghanna
dp
our.
,
“
Multi
-
Obj
ec
t
ive
Vehi
cle
R
outi
ng
Problem
with
Ti
m
e
W
ind
ows
Us
ing
Goal
Program
m
ing
and
Gene
t
ic Algori
thm,”
App
lied
S
oft
Computing
,
v
ol.
10
,
issue
4,
Sep
2010.
[6]
Y.
Bai,
et
al
.
,
“
S
tudy
of
Multi
-
V
e
hic
l
e
Rou
ti
ng
Proble
m
wit
h
Time
Window,
”
tw
el
f
th
Inte
rnat
ional
S
y
m
posium
on
Opera
ti
ons R
ese
arc
h
and it
s App
li
c
at
ions
in engi
nee
ring
,
te
chno
l
og
y
and
m
an
agem
ent
(ISO
RA), 2015.
[7]
K.
Heng,
et
al.,
“Appli
cation
of
Rout
e
Flex
ib
il
i
ty
in
Data
-
Star
ve
d
Ve
h
ic
l
e
Rou
ti
ng
Proble
m
wi
th
T
ime
Windows,”.
IEE
E
Congress
on
Evol
u
ti
onar
y
Com
puta
ti
on
(C
EC),
2016
.
[8]
R.
Chun
y
u
and
W
.
Xiaobo.,
“R
ese
arch
on
Multi
-
Ve
h
ic
l
e
and
M
ult
i
-
Depot
V
ehic
le
Rout
ing
Prob
le
m
wi
th
Time
Windows
for
El
ectronic
Comm
ere
,
”.
Int
ern
atio
nal
Conf
ere
n
ce
on
Artifi
ci
a
l
Inte
lligen
ce
and
Com
puta
ti
onal
Inte
lligen
ce,
201
0.
[9]
Z.
Urs
ani
,
et
al
.
,
“
Loc
al
ized
Gene
tic
Algorit
hm
for
Vehic
le
Rout
ing
Problem
wit
h
Ti
m
e
W
indow
s,
“
Appl
ie
d
Soft
Computing
,
vo
l.
11,
issue
8,
pp.
5
375
-
5390,
De
c
2
011.
[10]
S.
Kara
ka
ti
c
an
d
V.
Podgorel
e
c.
,
“
A
Surve
y
o
f
Gene
tic
Algor
it
hm
s
for
Solving
Multi
Depo
t
Vehic
l
e
Rout
in
g
Problem,”
Appl
i
ed
Sof
t
Computi
ng
,
vol
.
27
,
pp
.
5
19
-
532,
Feb
201
5.
[11]
S.
N.
Kum
ar
and
R.
Panneers
elvam
.
,
“
A
Ti
m
e
-
Depe
ndent
Veh
i
cl
e
Rou
ti
ng
Pro
ble
m
with
Ti
m
e
W
indows
for
E
-
Com
m
erc
e
Supplie
r
Si
te
Pi
ckups
Us
ing
Gene
tic
Algorit
hm
,
”
In
telli
gen
t
Informati
on
Manage
ment
,
vol.
7
,
no
.
4,
pp
.
181
-
194,
Jul
201
5.
[12]
Y.
W
ang,
et
al
.
,
“
Two
-
Ec
he
lo
n
Logi
stic
s
Distribut
ion
Regi
on
Parti
ti
oning
Pr
oble
m
Based
on
Hy
br
id
Sw
arm
Optimiza
ti
o
n
-
Ge
net
i
c
Algor
i
thm,
”
E
xpe
rt
Syste
ms
wit
h
Applicatio
ns
,
vol
42
,
issue
12,
Jul 20
15
.
[13]
D.
M.
Pierre
an
d
N.
Za
kar
i
a.,
“
Stocha
stic
Par
ti
a
lly
Op
ti
m
iz
ed
C
y
c
lic
Shift
Cross
over
for
Multi
-
Objec
t
ive
Gene
t
ic
Algorit
hm
s
for
The
Vehi
cl
e
Ro
uti
ng
Problem
with
Ti
m
e
-
W
indo
ws
,
”
Appl
i
ed
Sof
t
Computing
,
vol
.
52,
pp.
863
-
87
6,
Mar
2017.
[14]
W
.
F.
Mahm
udy
.
,
“
Im
prove
d
Sim
ula
te
d
Anne
al
ing
for
Opt
imization
of
Vehi
cl
e
Rou
ti
ng
Pro
ble
m
with
Ti
m
e
W
indows (VRP
TW),”
Kurs
or
,
v
ol.
7
,
no
.
3
,
pp
.
1
09
-
116,
2017
.
[15]
T.
N.
Fat
y
a
nosa,
et
al.,
“
H
y
brid
Gene
tic
Algorit
h
m
and
Sim
ula
te
d
Annea
li
ng
for
Functi
on
Optimi
za
t
ion,
”
Journa
l
of
Information
Technol
og
y
and
C
omputer
Scienc
e
.
Vol
.
1
,
No.
2,
p
p.
82
-
97
.
1
[16]
O.
Bahri
.
,
“
Multi
-
Obje
ct
iv
e
Optimiat
ion
of
Vehic
l
e
Routi
ng
Problem
with
unce
rtaint
y
:
An
Approac
h
Based
on
Pos
sibil
ity
Th
e
o
r
y
,
”
2012
.
<
h
tt
p
:/
/www
.
la
rod
ec.
com/download/
f
il
e/
f
id/
543
>
[Ac
ce
ss
on
Janu
ar
y
25
2017]
[17]
S.
W
.
Li
n,
e
t
al.
“
Vehic
le
Rout
in
g
Problems
with
Ti
m
e
W
indows
Us
ing
Sim
ula
te
d
Annea
li
ng
,
”
Sys
te
ms
,
Man,
and
Cybe
rnet
ic
s
,
pp.
645
-
650,
Jul
201
7.
[18]
N.
W
assan,
et
al
.
“
The
Multi
ple
Tri
p
Vehic
le
R
outi
ng
Problem
with
Bac
khaul
s:
For
m
ula
ti
on
an
d
a
Two
-
Le
ve
l
Vari
able
Ne
ighb
orhood
Sear
ch
,
”
Computer
&
Op
erati
ons R
ese
arc
h
,
vol
.
78
,
pp
.
45
4
-
467,
Feb
2017
.
[19]
Y.
Yang
and
L
.
Ta
ng.
,
“A
Filt
er
-
and
-
fan
Appr
oach
to
th
e
Mul
ti
-
Tr
ip
Ve
hi
cl
e
Rout
ing
Probl
e
m,”
Inte
rna
t
ional
Confer
ence
on
L
ogisti
cs
S
y
s
te
m
s a
nd
In
telli
g
ent
Mana
gement,
vo
l.
3
,
pp
.
1713
-
17
17,
Jan
2010.
[20]
Y.
Y.
Zha
ng,
e
t
al
.
,
“
Optimal
Re
servoir
Operation
Us
ing
a
Hybrid
Simulat
ed
Anne
ali
ng
A
lgorit
hm
-
Gene
t
i
c
Al
gorithm,
”
Internat
ion
al
Confer
enc
e
on
Bio
-
Inspire
d
Com
puti
n
g:
The
ori
es
and
Applic
ations,
p
p.
454
-
458,
Nov
2010.
[21]
B.
Minocha
,
e
t
al
.
,
“Solvi
ng
Ve
hi
cl
e
Rou
ti
ng
and
Sche
duli
n
g
Proble
ms
usi
ng
Hybrid
Gen
et
i
c
Al
gorithm”.
Inte
rna
ti
ona
l
Co
nfe
ren
c
e
on
E
le
c
troni
cs
Com
puter T
e
chnol
og
y
,
2
011,
pp
.
189
-
19
3.
[22]
V.
N.
W
ij
a
y
an
in
grum
and
W
.
F.
Mahm
udy
.
,
“
Op
ti
m
iz
ation
of
Shi
p’s
Route
Sched
uli
ng
Us
ing
Gene
tic
Algorit
hm
,
”
Indone
sian
Journal
of
El
ectric
a
l
Engi
nee
ring
an
d
Computer
Sci
enc
e
(
IJE
ECS)
,
vol
.
2,
no.
1,
p
p.
180
-
186,
Apr
2016.
[23]
W
.
F.
Mah
m
udy
,
et
al.
,
“
Rea
l
Coded
Gene
tic
Algorit
hm
s
for
Solving
Flexi
ble
Job
-
Shop
Schedul
ing
Problem
–
Part
I:
Model
ing,”,
A
dvanc
ed
Mat
eria
ls R
ese
arch
,
vo
l
.
701,
pp.
359
-
36
3,
Ma
y
2013.
[24]
W
.
F.
Mah
m
udy
,
et
al.
,
“
Rea
l
Coded
Gene
tic
Algorit
hm
s
for
Solving
Flexi
ble
Job
-
Shop
Schedu
li
ng
Problem
–
Part
II:
Optimiz
at
ion
,
”
Ad
vanced
Mat
erial
s R
ese
arch,
vol
.
701
,
pp
.
364
-
369,
Ma
y
2013.
Evaluation Warning : The document was created with Spire.PDF for Python.