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.
10
,
No.
4
,
A
ugus
t
2020
,
pp. 359
6~36
0
4
IS
S
N: 20
88
-
8708
,
DOI: 10
.11
591/
ijece
.
v10
i
4
.
pp3596
-
36
04
3596
Journ
al h
om
e
page
:
http:
//
ij
ece.i
aesc
or
e.c
om/i
nd
ex
.ph
p/IJ
ECE
Conflict
-
free dyn
amic r
oute
mu
lti
-
agv usin
g dijkst
ra
Floyd
-
war
shall h
ybrid
algorith
m
with tim
e w
ind
ows
So
li
chudin
, A
ri
s Tri
w
iya
tno
, Mun
awar A.
R
iy
ad
i
Dep
a
rtm
ent o
f El
ect
rical
Eng
i
neer
i
ng, F
ac
ulty
o
f
Enginee
ri
ng, Dip
on
e
gor
o Un
i
ver
sit
y,
I
ndonesi
a
Art
ic
le
In
f
o
ABSTR
A
CT
Art
ic
le
history:
Re
cei
ved
Sep
9
, 2
019
Re
vised
Dec
2
9
,
20
19
Accepte
d
J
an
11
, 2
020
Autonom
ous
Gu
ide
d
Vehicle
is
a
m
obil
e
robot
tha
t
ca
n
m
ove
au
tonomous
l
y
on
a
route
or
lane
in
an
indoo
r
or
outdoor
env
ironment
while
per
form
ing
a
serie
s
of
ta
sks
.
Dete
rm
ina
ti
o
n
of
the
shortest
route
on
an
aut
onom
ous
guide
d
vehicl
e
is
one
of
the
o
pti
m
iz
atio
n
pro
ble
m
s
in
handl
i
ng
conf
li
c
t
-
fre
e
rout
es
that
have
an
inf
l
uenc
e
on
the
distri
buti
on
of
goods
in
the
m
anuf
a
ct
uri
ng
industr
y
'
s
ware
house.
Picku
p
and
d
el
iv
er
y
proc
esses
i
n
the
distr
ibut
ion
on
AG
V
go
ods
such
as
sche
duli
ng
,
shi
pping,
an
d
det
ermining
the
r
oute
of
v
ehi
c
l
e
with
short
m
ileage
ch
ara
c
te
r
isti
cs,
is
v
e
r
y
poss
ibl
e
to
do
sim
ula
ti
ons
with
thre
e
AG
V
units
.
The
r
e
is
a
wi
ndows
ti
m
e
li
m
it
on
works
ta
ti
ons
tha
t
l
imits
shipping.
The
proble
m
of
d
et
erminin
g
the
route
in
this
study
is
conside
red
n
ec
essar
y
as
a
m
ult
i
-
ve
hic
l
e
rout
e
proble
m
with
a
t
ime
window.
Thi
s
stud
y
ai
m
s
to
desc
ribe
the
com
bina
ti
on
of
al
gorit
hm
s
written
base
d
on
d
y
n
amic
progra
m
m
ing
to
over
come
the
proble
m
of
conf
l
ic
t
-
f
ree
AG
V
route
s
using
ti
m
e
window
s.
The
combined
appr
oa
ch
o
f
the
Dijkstr
a
a
n
d
Flo
y
d
-
W
arsha
ll
a
lgori
thm
r
esult
s
in
the
op
tim
iz
at
ion
of
the
cl
osest
distance
in
ov
erc
om
i
ng
conf
l
ic
t
-
f
ree r
oute
s.
Ke
yw
or
d
s
:
Confli
ct
-
fr
ee
rou
te
Dynam
ic
p
rogra
m
m
ing
Hybr
i
d
al
gorithm
Mult
i
-
AGV
Ti
m
e w
indo
ws
Copyright
©
202
0
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
:
So
li
ch
ud
i
n,
Dep
a
rtm
ent o
f El
ect
rical
En
gi
neer
i
ng,
Dipone
gor
o U
niv
e
rsity
,
Pr
of.
S
oed
a
rto
,
SH
. S
tre
et
, Te
m
balang
, S
em
arang,
12
69, In
donesia.
Em
a
il
:
su
din3007@
gm
ail.co
m
1.
INTROD
U
CTION
Veh
ic
le
trans
portat
io
n
syst
em
s
based
on
AGVs
ha
ve
been
de
velo
ped
for
deca
de
s.
They
are
te
chn
ic
al
ly
adv
ance
d
an
d
c
om
plex.
Ther
e
are
m
any
tacti
cal
and
op
e
rati
on
al
issues
wh
ic
h
hav
e
to
be
addresse
d.
Dec
isi
on
s
relat
ed
t
o
the
de
sig
n
st
age
hav
e
a
very
la
rg
e
im
pa
ct
on
fu
t
ur
e
syst
e
m
per
f
orm
ance
[
1].
This a
rea incl
udes iss
ues
s
uc
h as:
gu
i
de
-
path
d
esi
gn
est
i
m
ating
the
nu
m
ber
a
nd lo
cat
ion
of p
a
rk
i
ng, pic
k
-
up a
nd
delivery
poin
ts,
est
i
m
ating
the
nu
m
ber
of r
e
quire
d veh
ic
le
s
The
sec
ond are
a of
prob
le
m
s r
el
at
ing
to
syst
em
m
anag
em
ent
inclu
des
t
he fo
ll
ow
in
g:
po
sit
io
ning
of idle ve
hicle
s,
veh
ic
le
disp
at
c
hing,
veh
ic
le
routin
g,
veh
ic
le
sc
he
duli
ng
,
colli
sion
a
nd
de
adloc
k
a
vo
i
da
nce
On
e
s
uc
h
inter
pr
et
at
io
n
in
vo
l
ves
the f
ollo
wi
ng
p
r
oble
m
:
if
a
delivery
de
ci
sion
is
m
ade,
the
r
ou
te
a
nd
sche
du
le
m
us
t
be
plan
ne
d
for
an
oth
e
r
A
GV,
that
the
purpo
se
of
A
GV
sc
he
du
li
ng
is
to
s
end
a
n
AGV
s
et
an
d
the
r
ou
te
m
issio
n
is
t
o
fi
nd
a
su
it
able
r
oute
[
2]
.
But
on
a
no
ther
inte
rpretat
ion
,
determ
ining
t
he
veh
ic
le
route
m
us
t
be
in
ac
corda
nce
with
the
order
of
sta
ti
on
s
that
m
us
t
be
visit
ed
by
this
veh
i
cl
e.
Sche
duli
ng
al
so
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
Con
fl
ic
t
-
fre
e
d
ynamic
r
oute
m
ulti
-
A
GV
u
si
ng
dijkst
ra f
loy
d
-
w
ar
s
ha
ll
…
(
So
li
ch
udin
)
3597
determ
ines
the
veh
ic
le
'
s
tim
e
wh
e
n
it
ha
s
to
ta
ke
a
nd
se
nd
loa
ds
[3
]
.
S
om
eti
m
es
disp
at
ching
,
r
outi
ng
a
nd
sche
du
li
ng
are
handled
se
pa
ratel
y.
However,
this
pr
ob
l
e
m
is
cl
os
el
y
relat
ed
an
d
m
us
t
be
con
si
der
e
d
si
m
ultaneou
sly
.
Co
ns
ide
rati
on
of
the
pro
bl
e
m
carried
ou
t
depen
ds
on
the
f
or
m
of
m
et
hods
a
nd
al
go
rithm
s
app
li
ed
. T
he p
aram
et
ers
us
ed
in
this
stu
dy a
re as f
ollow
s
.
Veh
ic
le
disp
at
chin
g
This
pro
blem
i
s
co
ns
ide
re
d
f
r
om
two
points
of
vie
w.
First,
a
n
a
vaila
ble
l
oad
is
assi
gn
e
d
to
an
idle
veh
ic
le
.
S
eco
nd,
a
n
idle
AGV
is
assig
ne
d
to
a
wait
in
g
lo
ad.
Disp
at
c
hing
m
et
ho
ds
a
re
base
d
on
the
r
ul
es
by
wh
ic
h
t
he
idle
A
GVs
ar
e
sel
ect
ed
a
nd
assi
gn
e
d
t
o
e
xec
ute
the
tr
ans
port
at
ion
ta
s
ks
,
e.
g.
nea
rest
or
s
hortest
veh
ic
le
tra
vel t
i
m
e ru
le
[4].
Veh
ic
le
routin
g
a
nd sc
heduling
Af
te
r
dis
patchi
ng,
a
c
ertai
n
r
ou
te
sho
uld
be
plan
ne
d
a
nd
assigne
d
t
o
t
he
A
GV.
On
ce
the
routin
g
decisi
on
is
m
a
de,
t
he
ap
pro
pri
at
e
sche
dule
m
us
t
be
set
.
W
it
h
sc
heduli
ng,
al
l
arr
i
val
and
de
par
t
ur
e
tim
es
of
the
AGV
at
each
sig
nificant
po
int
on
the
route
are
deter
m
ined.
This
is
necessary
to
ens
ur
e
colli
sio
n
an
d
dead
l
ock f
ree
r
ou
ti
ng
[
5].
Dynam
ic
p
at
h t
opology
Op
ti
m
al
ro
ute
search
gen
e
rall
y
ta
kes
a
lon
g
tim
e.
This
is
becau
se
the
al
gorithm
is
base
d
on
glo
ba
l
inf
or
m
at
ion
an
d
cal
c
ulate
s
al
l
routes
for
ea
ch
AGV.
T
he
dynam
ic
m
eth
o
d
of
fin
ding
r
ou
te
s
is
bas
ed
on
real
-
ti
m
e
info
r
m
at
ion
ab
out
traf
fic.
F
or
e
xa
m
ple,
in
the
S
un
an
d
Wa
nk
researc
h
al
gori
thm
the
m
et
ho
d
is
cal
le
d
Dynam
ic
Tim
e
-
windows
for
Sm
art I
ndoor
(
DT
WS)
[
6]
.
C
olli
sion
-
fr
ee
routes
occur
i
n
m
any
app
li
c
at
ion
s
rangin
g
from
robo
t
m
ov
em
ent
to
sc
hedulin
g
of
sever
al
c
ra
nes
in
lo
gisti
cs
to
con
ta
ine
r
tra
nsporta
ti
on
by
a
uto
m
at
ed
gu
i
de
d
veh
ic
le
s
(AGV)
[7
]
.
Ty
pical
ly
,
AGV
is
not
eq
uipped
with
a
sm
art
local
coll
isi
on
av
oid
a
nc
e
syst
e
m
and
only
reli
es
on
c
entral
co
ntr
ol.
This
is
us
ua
ll
y
done
with
a
sta
ti
c
a
ppr
oach,
bu
t
i
n
this
stu
dy
w
e
de
velo
ped
a
dynam
ic
app
r
oa
ch.
The
first
s
te
p
is
to
cal
culat
e
the
r
ou
te
i
n
the
un
der
ly
in
g
gr
a
ph
an
d
the
n
us
e
add
it
io
nal
m
eth
ods
duri
ng
r
oute
exec
utio
n
t
o
a
vo
i
d
a
colli
sion
.
T
hese
r
ules
are
us
ually
heurist
ic
[8
]
wh
ic
h
can
le
ad
to
un
e
xpect
ed
ar
rival
tim
es
an
d
eve
n
dead
l
ocks. T
o ov
e
rc
om
e this we devel
op
e
d usin
g
cl
assic
al
ru
le
s.
Pr
e
vious
resea
rch
on
r
ou
ti
ng
al
go
rit
hm
s
ca
n
be
cl
assifi
e
d
in
to
tw
o
cl
asses,
nam
el
y
gen
eral
path
topolo
gy
an
d
s
pecial
path
t
opology
[9
]
.
This
stud
y
pro
po
se
s
the
de
vel
op
m
ent
of
a
gen
e
ra
l
path
to
polo
gy
wit
h
a
gr
id
m
od
el
us
in
g
the
dijks
tra
al
gorithm
.
The
ge
ne
ral
pa
th
he
re
m
eans
that
the
pat
h
netw
ork
is
a
regular
gr
a
ph.
B
ro
a
db
ent
et
al
.
[10]
was
t
he
fi
rst
r
esearche
r
t
o
prov
i
de
the
c
on
c
ept
of
A
GV
r
outi
ng
in
a
s
hor
t
tim
e
fr
ee
of
c
onflic
t
and
propos
e
routin
g
proc
edures
based
on
t
he
sho
rtest
path
al
gorith
m
.
W
h
ile
Ha
m
zee
i
et
al
.
[11]
intr
oduce
d
a
n
al
gor
it
h
m
to
route
AGV
ve
hic
le
s
in
tw
o
-
way
net
works
based
on
a
nn
eal
ing
al
gorithm
m
et
ho
ds
,
the
n
Nish
i
an
d
Ta
na
ka
[
12
]
disc
usse
d
dynam
ic
routes
to
ove
r
com
e
con
flic
t
-
fr
ee
disp
at
c
h
ing
an
d
AGV ro
utes.
Re
search rela
te
d
to t
he AG
V veh
ic
le
c
onflic
t
-
fr
ee
r
ou
te
whic
h
was
scruti
ni
zed b
y R
a
hn
a
m
a et
al
[1
3]
,
Ba
har
i
[
14]
,
A
ntackly
et
.al
[8]
,
Ma
lop
ols
ki
[
1]
,
an
d
Ku
m
ar
et
.al
[15
]
,
w
he
n
ope
rated
only
1
or
2
AGV
s
that
can
acce
pt
assi
gn
m
ents
so
tha
t
wh
e
n
a
nother
A
GV
is
opera
te
d,
the
pre
vious
AGV
m
us
t
wait
for
a
n
ord
er
or
com
m
and
.
Fur
therm
or
e,
wh
e
n
acce
pting
th
e
ne
xt
assig
nm
ent,
each
A
GV
was
only
able
to
av
oid
fro
nt
colli
sion
s,
a
nd
dea
dlo
c
ks
bet
ween
inter
sect
ion
s
as
i
n
Jae
Pil
'
s
researc
h
[
16
]
a
nd
M.
Strzel
ecki
[17]
only
avo
i
dan
ce
of
st
at
ic
bar
rie
rs,
but
not
disc
us
s
l
ivelock
ha
n
dling
[
18
]
,
s
o
t
hat
the
s
peed
of
AGV
decr
ea
se
s
w
hen
appr
oach
i
ng
obsta
cl
es
as
res
earch
Arutsel
va
n
[
19
]
an
d
Y.
Wan
et
.al
[20]
res
ulted
in
the
em
erg
en
ce
of
a h
yst
eresis e
ffec
t beca
us
e ac
cur
acy
decre
as
es whe
n
A
G
V spee
d
inc
rease
s.
The
pr
ob
le
m
of
A
GV
m
ov
e
m
ent
t
hat
app
e
ars
in
the
li
te
r
at
ur
e
re
view
a
bove
is
that
re
searche
rs
pay
le
ss
at
te
ntion
to
the
distance
f
un
ct
io
n,
the
prob
le
m
of
ha
ndli
ng
un
ce
rtai
nty
betwee
n
A
GV,
fin
ding
the
s
hortest
r
oute
,
the
c
onflic
t
-
fr
ee
r
ou
te
,
a
nd
opti
m
al
sched
ulin
g,
an
d
pa
rk
i
ng
ef
fici
en
c
y.
Va
rio
us
kin
ds
of
def
ic
ie
ncies
w
hich
e
xp
la
i
ned
in
relat
ed
res
earch
,
so
i
n
th
is
pap
e
r,
th
e
r
esearche
r
proposes
a
n
opti
m
i
zat
ion
so
luti
on
t
o
ha
nd
le
c
onflic
t
f
ree
r
oute
s
wit
h
a
dynam
ic
ro
ute
a
ppr
oac
h
us
in
g
a
hy
br
i
d
al
gorit
hm
dij
kst
ra
floyd
-
wa
rs
hall wh
ic
h
incl
udes
sch
e
duli
ng p
a
r
a
m
et
ers,
shor
te
st route sea
rch,
and c
onflic
t fre
e r
ou
te
.
2.
RESEA
R
CH MET
HO
D
The
resea
rc
h
m
et
ho
d
to
ove
rco
m
e
ob
sta
cl
e
fr
ee
r
oute
s
in
m
ul
ti
-
AGV
use
d
is
dynam
ic
pro
gr
am
m
ing
us
in
g
the
flo
yd
-
w
ars
hall
al
gorithm
with
a
gr
id
to
po
l
ogy
route
m
odel
.
Dynam
ic
routes
us
e
dy
nam
ic
pro
gr
am
m
ing
with
ti
m
e
wind
ows
base
d
on
pro
blem
s
in
determ
ining
ve
h
ic
le
ro
utes
,
wit
h
s
olu
ti
ons
to
so
lve
pro
blem
s
that
app
ea
r
i
n
the
f
or
m
of
C
onflic
t
-
fr
ee
Ve
hicle
Rou
te
Prob
le
m
s
with
Tim
e
W
i
ndow
(CV
RPT
W)
.
Dynam
ic
ro
uting
m
od
el
is
pr
opos
e
d
as
a
s
ol
ution
to
overc
om
e
the
pr
oble
m
of
co
nf
li
ct
-
f
ree
r
ou
te
s
i
n
fi
nd
i
ng
the
shortest
pa
th
to
the
desti
nation
(
wor
ks
ta
ti
on
)
i
n
the
A
GV
pick
-
up
/
drop
-
off
distri
buti
on
process,
s
o
that
each
AGV
ope
rati
on
r
uns
sm
oo
t
hly
acco
rd
i
ng
to
the
sp
eci
fied
ti
m
e.
The
m
ul
ti
-
AGV
a
uto
nom
ou
s
na
vig
at
io
n
syst
e
m
con
sist
s
of
th
ree
pa
rts,
nam
el
y
the
input
par
t
(
wi
th
pa
ram
et
ers:
delivery,
c
onfl
ic
t
fr
ee
routes,
and
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.
10
, No
.
4
,
A
ugus
t
2020
:
3596
-
3604
3598
sche
du
li
ng)
,
th
e
process
par
t
(d
at
a
processin
g)
a
nd
t
he
outp
ut
pa
rt
(d
ist
ri
buti
on
proces
s:
pick
up
a
nd
del
ivery
)
on the
w
orksta
ti
on
. T
he
AG
V
rou
te
t
opology
syst
e
m
b
ased
on the
pa
ram
eter
s a
bove
is
shown i
n
F
i
gure
1.
Figure
1. G
rid t
opology r
oute
Pr
oble
m
s
associat
ed
with
th
e
desig
n
of
fl
ow
paths
s
uc
h
as
disp
la
cem
ent
an
d
buf
fers,
fleet
siz
e
est
i
m
at
es,
and
ta
ke
-
up
an
d
drop
-
off
sta
ti
ons
hav
e
bee
n
ag
re
ed
by
Vim
ar
and
Selva
[
15]
,
Klei
n
a
nd
Kim
[
21
]
,
Liu
et
al
.
[5
]
,
Um
ar
et
al
.
[2
2],
and
C
orrea
et
al
.
[23]
Thr
e
e
ty
pes
of
bi
-
di
recti
on
al
netw
ork
r
ou
te
m
od
el
s
are
widely
us
e
d
on
A
GV
routes,
su
ch
as
sin
gle
la
ne,
double
la
ne,
a
nd
m
ixed
m
od
el
s
[24
]
.
W
e
us
e
a
tw
o
-
way
sing
le
path
d
es
ign
us
i
ng a
gr
i
d
to
polo
gy
m
od
el
w
it
h di
recti
on
al
grap
hs
.
The
FMS
syst
e
m
gr
id
to
pol
og
y
route
net
work
sim
ulatio
n
i
n
Fi
gure
1
us
es
24
point
s/nodes
al
l
of
wh
ic
h
are
cal
culat
ed
to
fin
d
the
cl
os
est
dista
nce
to
12
node
s
as
wo
r
ks
ta
ti
ons
cal
culat
ed
in
the
P/D
proc
ess
of
distrib
ution
of
goods
in
t
he
m
anufactu
rin
g
in
du
st
ry
wa
reho
us
e,
ass
um
ing
that
node
1
(
w
areho
us
e)
is
ori
gin
al
node,
no
de
2
(
DMD
),
node
3
(
Die
cast
in
g)
,
node
4
(Machi
ning)
,
node
5
(
Painti
ng),
no
de
7
(P
la
sti
c
I
nje
ct
ion
)
,
node
18
(
Weld
ing),
node
20
(
Assem
bly),
no
de
21
(Re
pair
),
node
22
(
Fina
l
In
s
pecti
on)
,
node
24
(PPIC
),
an
d
node
15
(
Par
ki
ng).
All
node
s
are
cal
c
ulate
d
us
in
g
the
dijkstra
a
nd
fl
oyd
-
wa
rshal
l
al
gorithm
in
find
i
ng
the
shortest
r
oute
with
dyna
m
ic
pr
ogram
m
ing
.
The
route
is
create
d
usi
ng
a
bi
directi
on
al
gr
id
t
opolog
y
m
od
el
to ove
rco
m
e the obstacl
e f
ree
route.
The
ty
pe
of
co
l
li
sion
-
fr
ee r
oute
s can
be
cl
assifi
ed
into four types [
25]
with thr
ee su
ggest
e
d
so
luti
ons
,
nam
ely
(a)
:
Sc
hedule
eac
h
A
GV
;
(
b
):
Sele
c
t
the
desire
d
r
ou
te
;
a
nd
(c)
AGV
go
e
s
acc
ordin
g
to
it
s
pur
pose.
Each
c
olli
sion
cl
assifi
cat
ion
is
fo
ll
o
we
d
by
on
e
or
tw
o
cas
es.
Ba
sed
on
this
case,
we
c
hose
the
rig
ht
st
rategy
for
the
di
ff
e
re
nt
colli
sio
n
cl
assifi
cat
ion
s
is
show
n
in
Fig
ur
e
2.
Re
ga
r
din
g
the
pla
nn
i
ng
a
nd
sc
heduling
of
the
pr
opos
e
d
r
ou
te
s
f
or
e
ach
AGV,
it
is
ex
pected
to
im
pr
ove
the
ef
fici
ency
of
the
A
utono
m
ous
Veh
ic
le
S
yst
e
m
.
Figure
2. Dy
na
m
ic
r
ou
te
c
olli
sion t
ype (a
)
he
ad
-
on c
olli
sion
,
(
b)
c
r
os
s
-
c
ol
li
sion
,
(c)
node
-
occup
ancy
colli
sio
n
,
(d)
s
helf
-
occ
up
ancy coll
isi
on
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
Con
fl
ic
t
-
fre
e
d
ynamic
r
oute
m
ulti
-
A
GV
u
si
ng
dijkst
ra f
loy
d
-
w
ar
s
ha
ll
…
(
So
li
ch
udin
)
3599
T
he
pur
pose
of
sc
hedulin
g
is
to
se
nd
a
seri
es
of
A
GV
ta
sk
s
i
n
ac
hievi
ng
go
al
s
in
pic
k
-
up/d
rop
-
off
work
(
or
P/
D
f
or
s
hort)
unde
r
certai
n
co
ns
tr
ai
nts
su
c
h
as
de
adlines,
pr
i
or
i
ti
es,
route
co
nfl
ic
ts
et
c.
Ob
j
ec
ti
ves
are
us
ually
rel
at
ed
to
proce
s
sing
tim
e
or
r
eso
ur
ce
util
iz
at
ion
,
su
c
h
a
s
m
ini
m
iz
ing
th
e
am
ou
nt
of
AGV
involve
d
w
hile
m
ai
ntaining
syst
e
m
thro
ughp
ut,
or
m
ini
m
iz
ing
the
total
travel
tim
e
of
al
l
veh
ic
le
s
[26]
.
The
prob
le
m
giv
e
n
can
be
sta
te
d
as
f
ollo
ws.
T
he
re
are
three
AGV
ve
hicle
s
with
Q
capaci
ty
an
d
will
distrib
ute
good
from
the
or
i
gin
node
to
th
e
wa
reho
us
e
wh
ic
h
is
re
pr
e
sented
by
a
ne
twork
G
=
(
V,
A)
,
V
{
0,1,...
,n,
n
+
1}
.
T
his r
ep
resen
ta
ti
on
is
t
he
set
of
nodes
or
A
as
t
he
set
of
a
rcs,
w
her
e
A
=
{i,
j)
|
i,
j
V}
.
Fo
r
i
V
is
tim
e
wi
ndows
[
a
i
b
i
]
an
d
ser
vi
ce
tim
e
S
i
,
fo
r
exam
ple
V`=
V
-
{
n=1}.
F
or
ar
c
(
i,
j
)
A
has
travel
tim
e
t
ij
and
co
st
(d
ist
ance
)
c
ij
.
The
or
i
gin
node
is
denoted
by
0
or
n
+
1
accor
ding
to
th
e
po
sit
io
n
in
dicat
i
ng
the
ori
gin
no
de
or
destinat
io
n
no
de
(e
nd)
with
S
0
=
S
n
+1
=
0,
q
0
=
q
n
+1
=
0.
E
an
d
L
ind
ic
at
e
the
earli
est
dep
a
rtu
re
ti
m
e
from
the
or
i
gi
n
no
de
a
nd
t
he
la
st
ar
rival
tim
e
wh
en
re
achin
g
the
des
ign
at
ed
w
orkst
at
ion
.
The
pur
pose
of
this
prob
le
m
is
to
m
ini
m
ize
the
m
il
eage
in
serv
i
ng
the
pr
oce
ss
of
pick
-
up
and
dro
p
-
off
distrib
ution
of
AGV
goods
in
the
m
anu
fact
uri
ng
in
du
st
ry
war
eh
ouse
us
in
g
gr
id
to
po
l
og
y
by
m
eet
ing
the
tim
e
window c
onstr
ai
nts.
Be
fore r
e
so
l
ve a
conflic
t
-
f
ree
route,
we
i
ntrodu
cet
he follo
w
ing
e
quat
io
n n
otati
on
:
G
=
the
directed
gr
a
ph
V
= init
ia
l node
V
'
= n
e
xt no
de
A
= side /
bow
Q
= A
GV loa
d
ca
pacit
y
C
ij
=
the
distance
of origin
no
de t
o
de
sti
nation
node
D
= the
distance
betwee
n
tw
o
a
dj
ace
nt stat
io
n
X
ij
= d
eci
sio
n vari
able
t
ij
= travel ti
m
e
N
= num
ber
of
ve
rtic
es
k
= interm
ediat
e n
ode
S
0
=
q
0
= origin
no
de
[a
i
b
i
]
= tim
e w
indow
S
i
= ser
vice ti
m
e
Pr
oble
m
f
or
m
ulati
on
g
i
ve
n:
m
in
∑
∑
=
1
=
1
(1)
W
it
h
∑
=
∞
∀
∈
,
(2)
∑
=
∞
∀
∈
,
(3)
∑
=
1
,
(4)
∑
(
+
1
)
=
1
,
=
1
(5)
∑
−
∑
=
0
∀
∈
,
(6)
q
i
≤
u
i
≤
Q,
u
i
-
uj +
Qx
ij
≤
Q
-
q
j
(7)
∀
,
∈
with
q
1
+
q
j
≤ Q
(8)
S
i
+
t
ij
–
k(
1
-
x
ij
)
≤
S
j
∀
,
∈
,
(9)
a
i
≤ D
i
≤ b
i
,
∀
∈
(10)
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.
10
, No
.
4
,
A
ugus
t
2020
:
3596
-
3604
3600
X
ij
∈
{
0
,
1
}
∀
(
,
)
∈
(11)
In
(
2)
an
d
(
3)
are
sta
te
d
ea
ch
wor
ks
ta
ti
on
is
ap
pro
ve
d
and
acce
pted
on
dem
and
.
E
qu
at
io
n
s
(4)
and
(5)
re
pres
ent
the
arcs
t
ha
t
le
ad
to
an
d
le
ave
the
w
ork
sta
ti
on
.
E
qu
at
i
on
(6)
e
xpress
es
the
A
GV
fl
ow
at
a
node.
E
quat
ion
(7)
sta
te
s
th
at
the
dem
and
for
w
orkstat
io
ns
does
no
t
ex
ceed
the
A
G
V
c
apacit
y.
Eq
ua
ti
on
(
8)
sta
te
s
that
the
ro
ute
is
f
ree
of
ob
sta
cl
es.
T
he
n
(
9)
a
nd
(10
)
gu
a
ra
ntee
the
process
of
ta
ki
ng
a
nd
sen
ding
goods
accor
da
nce
with the
tim
e w
in
dows.
Thr
ee
poli
ci
es
wer
e
m
ade
regardi
ng
m
ov
em
ent
of
AGV
po
sit
ion
s,
na
m
ely
scheduli
ng,
se
nd
i
ng
,
an
d
confli
ct
fr
ee
r
ou
te
s
in
the
sim
ula
ti
on
.
I
n
the
first
case,
wh
e
n
the
ve
hicle
receives
de
li
ver
y
assignm
ents
to
the
pic
k
up
a
nd
dr
op
off
point,
t
hat
is
done
im
m
ediately
.
The
seco
nd
case,
the
veh
ic
le
is
dire
ct
ed
t
o
a
centrali
zed
workst
at
ion
ar
ea
if
the
re
is
no
furthe
r
ta
sk
re
qu
e
st.
The
n
in
the
thir
d
case,
al
l
AGVs
ca
n
carry
out
the
process
of
loadi
ng
a
nd
unloa
ding
of
goods
f
r
om
the
place
of
ori
gin
to
the
de
sti
nation
without
r
ou
te
c
onflic
t.
Give
n
a
route
netw
ork
G
=
(
V,
A
)
wh
ic
h
is
a
dire
ct
ed
gr
a
ph.
A
s
et
V
=
{1,
2,
3
,
4,
5,
6,
7,
8,
9,
10, 1
1,
12
}
V'
= {
2,
3,
4,
5,
6,
7, 8, 9,
10, 11, 12}
with a total
of
24
nodes
. T
he
set
of
nodes
tra
versed is
S
⊆
V´
,
w
her
e
node
i
=
2,
3,
4,
5,
6,
7,
8,
9,
10,
11,
12
are
the
sel
ect
e
d
no
des
to
be
visit
ed.
F
ur
t
he
rm
or
e
n
ode
1
is
the
node
that s
hows
the
beg
i
nn
i
ng
and en
d o
f
the
trip
node.
In
a
ddit
ion
,
t
he
distance
c
i
j
is
the
t
ij
travel
tim
e
between
node
s
i
an
d
j
for
pair
i,
j
∈
V
and
t
he
ti
m
e
window [
a
i
, b
i
]
f
or
eac
h
no
de
i
. D
ist
ance and
trav
el
tim
e
m
e
et
the i
m
pr
eci
sion
of
tria
ngle
s
c
ij
≤ c
ik
+ c
kj
, t
ij
≤ t
ik
+
tk
j
for
i,
j,
k
∈
V
.
In
t
he
gri
d
to
po
l
og
y
use
d
as
a
si
m
ulatio
n,
the
dista
nc
e
betwee
n
ve
rtic
es
at
al
l
no
de
s
is
40
centim
et
ers
ass
um
ing
a n
et
work r
ou
te
of
24
nodes
for
m
a sq
ua
re
box o
n
e
ach
node.
2.1.
Hy
brid
al
go
ri
th
m
formul
a
dijkst
r
a
-
fl
oyd
w
arsh
all
2.1.1.
T
he dij
k
stra a
l
go
ri
th
m
for
CVRPT
W
Give
n
a
directi
on
al
gr
a
ph
G
=
(V,
E)
,
the
non
-
ne
gative
t
ran
sit
ti
m
e
fu
nc
ti
on
ce
(t)
f
r
om
each
side
with
e
=
(v,
w)
E
with
[ai
,
bj]
is
a
wind
ow
of
ti
m
e
wh
e
re
ai
≤
t
≤
bj,
t
∈
T
is
serv
ic
e
and
le
aves
ti
m
e
fo
r
node
v,
s
ourc
e
node
s,
de
s
ti
nation
node
d
a
nd
dep
a
rtu
re
ti
m
e
t0
at
the
s
ource
node
of
t
he
s
hortest
tim
e
-
dep
en
de
nt
path
pr
ob
le
m
with
a
tim
e
wind
ow
s
uch
that
F
IFO
(f
i
rst
-
in
fir
st
-
out)
[
27
]
with
d
can
be
ach
ie
ve
d from
the
di
j
kst
ra al
gorithm
. I
n t
abl
e
the
op
ti
m
al
s
olu
ti
on is
fou
nd if
G sat
isfie
s
three c
onditi
on
s:
Fo
r
all
v
e
rteks
v, w
∈
V,
a
nd
t
v
≤
tw
≤,
+
h
(
v
,
t
v
)
≤
t
w
+
h
(
w, t
w
)
is a
FIFO
conditi
on.
Fo
r
all
edges
e
= (
v,
w
)
∈
E
a
nd
t
∈ T
,
h
(
v, t
)
≤
c
e
(
t
)
+
h
(
w, t + c
e
(
t
)
)
is
a square
con
diti
on
.
Fo
r
all
v
e
rteks
v, w
∈
V
a
nd
t
v
≤ t
w
[
a
v
,
b
v
]
≤ [
a
w
,
b
w
]
is a ti
m
e w
in
dows
c
onditi
on
.
2.1.2.
Pseud
oc
od
e Th
e
algor
ithm di
jk
st
ra
fo
r
t
he
t
im
e w
indows
functi
on
Dij
s
ktra(Gr
a
ph, so
urce):
for
eac
h verte
x v i
n
gr
a
ph:
//
init
ia
l
iz
at
ion
s
dist[v
]
:=
infi
nity
:
//
un
kn
own dist
ance
functi
on
f
ro
m
so
urce t
o v
pr
e
vious
[v
]
:=
unde
fine
d;
//
pr
evi
ou
s
no
de
in op
ti
m
al
p
at
h
end f
or
//
fr
om
so
urce
dist[s
ource]
:=
0;
//
distance f
ro
m
sour
ce
to so
urce
Q
:=
the
set
of
al
l nodes i
n G
r
aph
;
//
al
l no
des
in
t
he gra
ph are
unopti
m
iz
ed
–
thu
s
are
in Q
Wh
il
e
Q
is
not
e
m
pty:
//t
he
m
ain
lo
op
u
:=
ve
rtex
in Q
w
it
h sm
allest
d
ist
ance i
n dist
[] ; //
sta
rt no
de
in first ca
se
rem
ov
e u f
ro
m
Q
if d
ist
[
u] = in
fi
nity
:
br
ea
k;
//
al
l rem
ai
nin
g ve
rtic
es
are
end if
//i
naccess
ible from
so
urc
e
for
eac
h neig
hbor
v of u:
//
wh
e
re
v has
not y
et
bee
n
//
re
m
ov
e f
r
om
Q
al
t :
=dist
[u
]
+
dist_
b
et
wee
n(u
,v);
if alt
<dist
[v
]
:
//rel
ax
(u,
v,
a
)
dist[v
]
:= al
t ;
pr
e
vious
[v
]
:=
u;
decr
ease
-
key v i
n Q;
//
re
order
v
in
the
Q
ueu
e
end if
end f
or
end whil
e
return
distance
;
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
Con
fl
ic
t
-
fre
e
d
ynamic
r
oute
m
ulti
-
A
GV
u
si
ng
dijkst
ra f
loy
d
-
w
ar
s
ha
ll
…
(
So
li
ch
udin
)
3601
2.1.3. D
ynami
c progr
ammin
g
w
ith
flo
yd
-
w
arshall
a
l
go
ri
t
hm
The
nex
t
ste
p
is
to
fin
d
the
s
hortest
distanc
e
with
dy
nam
i
c
pro
gr
am
m
ing
us
i
ng
t
he
fl
oy
d
-
w
ars
hall
al
gorithm
accor
di
ng to
t
hese t
wo stat
em
ents.
di
jo
= p
at
h
le
ngth b
et
ween p
oi
nts
i
a
nd
j
d
ijk
= m
in (
d
ijk
-
1
, d
ikk
-
1
+ d
kjk
-
1
)
Th
e
n
in
reso
l
vi
ng
the
A
G
V
confli
ct
-
free
rout
e
pro
blem
with
tim
e
wind
ows
in
a
dynam
ic
pr
ogram
us
in
g
forw
a
r
d
cal
culat
ion
s
by
fo
ll
ow
i
ng
t
he
ste
ps
dev
el
op
e
d
by
D
um
as
[28]
as
in
the
de
scriptio
n
a
bove
can
be
e
xp
la
in
ed
a
s foll
ow
s
.
Step
1.
I
niti
al
izati
on
Sta
te
f
or
m
F({
1,
j}
, j,
t)
=
c
ij
if
(i
, j
)
∈
A,
F(
{1
,
j},
j, t)
=
∞
if
(
i, j
)
∉
A,
wi
th
a
j
≤
t
≤ b
j
, t
=
m
a
x{
a
1
+ s
1
+
t
ij
, a
j
}.
Step
2.
D
is
place
m
ent o
f
each
k
-
1
sta
te
set.
F
or each
sta
te
(
S, i, t
)
of the
k
-
1
sta
te
set, r
e
pea
t st
ep 3.
Step
3.
B
uild s
ta
te
o
n st
at
e
k
that ca
n be
ac
hieved eve
ry stat
e
(
S,
i,
t)
on
the
set k
-
1.
F
(S,
i,
t
)
=
m
in
{F(S
=
{j},
i,
t
´
)
+
c + | t
≥
t
´
, a
1
≤
t
´
≤ b
1
},
f
or
S
⊆
V´
, j
∈
S,
an
d
a
j
≤
t ≤
b
j
.
Ob
ta
in
ed
st
ate (S
´
, j,
t
)
=
(S
∪
{j},
j,
ma
x
{a
j
, t
+
s
i
+ t
ij
})
as a f
eas
ible exte
ns
io
n of t
he
sta
te
(
S, i,
t)
.
Step
4.
Po
i
nt of
view
k
wi
th k
+ 1
, if
k
≤ n,
t
he
n procee
d
to
s
te
p
2
.
Step
5.
Ca
lc
ulate
the optim
al
s
olu
ti
on. T
he
op
tim
a
l z dista
nc
e so
l
ution i
s g
i
ven b
y:
z
= m
i
n(
i,
n
)
∈
A
m
i
n
{F
{V
´
´
, i,
t)
+
c
in
| t
≤ b
n
–
t
in
–
S
1
}.
a
1
≤ t
≤
b
1
3.
RESU
LT
T
he
f
ollo
wing
is
giv
e
n
T
a
ble
1
w
hich
co
ntains
the
distance
of
12
no
des
th
at
represent
t
he
cal
culat
e
d
workst
at
ion
an
d
T
a
ble
2
w
hich
c
on
ta
i
ns
t
he
A
GV
thir
d
t
i
m
e
window.
The
res
ults
of
the
cal
c
ulati
on
of
the
al
gorit
hm
app
li
ed
to
the
gri
d
to
po
l
og
y
pat
h
pro
du
ce
the
value
of
the
dista
nce
be
tween
the
no
des
i
n
T
able
1
as
the
shortest
distance
on
a
pred
et
erm
ined
w
or
ks
ta
ti
on.
T
he
sh
ort
est
dista
nc
e
betwee
n
no
des
on
a
wo
r
ks
ta
ti
on
will
def
ine
the
form
ation
of
a
sta
te
gr
ap
h
(
S,
i,
t
),
with
the
sta
te
fo
rm
ed
as
a
node
an
d
the
sta
te
transiti
on
as a
n arc.
The
pa
th
sel
ec
te
d
from
it
s
pr
edecess
or
on
the
th
ree
A
GVs
wh
e
n
sea
rch
i
ng
for
the
s
ho
rtest
path
is
sh
ow
n
i
n
T
a
bl
e
2,
s
o
t
hat
w
hen
the
sta
te
gr
a
ph
is
form
ed,
t
he
sta
rt
ti
m
e
of
t
he
P/D
proces
s
dep
e
nd
s
on
the
tim
e
wind
ows
f
or
eac
h
i
is
a
i
≤
t
≤
b
i
with
tim
e
windows
2
an
d
c
onditi
on
s
t
=
m
ax
{
a
1
+
s
1
+
t
ij
,
a
j
},
j
∈
S
for
sta
ge
1
a
nd
t
j
=
m
ax
{
aj,
t
i
,
+
s
i
+
t
ij
}
for
othe
r
sta
ge
s,
with
the
le
ngth
of
distri
bu
ti
on
s
i
an
d
the
travel
tim
e
t
ij
.
Table
1.
D
ist
an
ce betwee
n n
odes
W
o
rks
tatio
n
/No
d
e
1
2
3
4
5
6
7
8
9
10
11
12
1
0
40
80
120
160
240
80
280
280
200
1200
80
2
40
0
40
80
120
200
240
240
200
160
160
120
3
80
40
0
40
80
160
200
200
160
120
200
160
4
120
80
40
0
40
120
160
160
120
160
240
200
5
160
120
80
40
0
80
120
120
160
200
280
240
7
240
200
160
120
80
40
0
120
160
200
280
240
13
80
160
120
200
240
240
200
200
160
120
40
0
18
280
240
200
160
120
40
0
80
120
120
240
200
20
280
240
200
160
120
120
80
0
40
80
120
200
21
240
200
160
120
160
160
120
40
0
40
120
160
22
200
160
120
160
200
200
160
800
40
0
80
120
24
120
160
200
240
280
280
240
160
120
80
0
40
Inf
o
r
m
atio
n
:
W
o
rks
ta
tio
n
: 1
2
3
4
5
6
7
8
9
10
1
1
1
2
node
: 1
2
3
4
5
7
1
8
2
0
21
22
24
13
Af
te
r
t
he
s
hor
te
st
path
val
ue
of
t
he
di
j
kst
r
a
and
floy
d
-
w
arsh
al
l
al
gorit
hm
cal
culat
ion
s
is
known
,
a
ne
w
sta
te
(
S,
i,
t
)
will
be
form
e
and
the
weig
htin
g
of
the
dista
nce
betwee
n
t
he
or
i
gin
node
an
d
the
destinat
io
n
node
al
read
y
known
us
i
ng
the
fo
r
ward
tracki
ng
syst
e
m
us
es
com
p
utati
on
al
cal
cu
la
ti
on
s,
as
in
Ta
ble
3.
I
n
Ta
ble
3,
dy
nam
ic
route
cal
culat
ion
with
a
ti
m
e
window
in
P/D
is
giv
e
n
acc
ordin
g
t
o
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.
10
, No
.
4
,
A
ugus
t
2020
:
3596
-
3604
3602
the
sche
duli
ng
ta
sk
f
or
eac
h
AGV.
Data
f
ro
m
Table
4
s
hows
th
at
thre
e
par
am
et
ers
(d
el
ivery,
route
an
d
sche
du
li
ng)
are
ab
le
to o
ver
c
om
e con
flic
t
-
fr
e
e AGV r
ou
te
s t
hat wor
k
a
uton
om
ou
sly
. Th
e
exp
e
rim
ental
r
esults
ob
ta
ine
d
data
that
the
pa
ram
e
te
rs
do
not
af
f
ect
the
nu
m
ber
an
d
s
peed
of
the
ve
hicle
,
but
aff
ect
the
a
rr
iva
l
interval
w
he
n
dem
and
incr
eases.
The
m
or
e
nu
m
ber
of
ve
hicle
s
us
e
d,
the
fa
ste
r
the
P/D
proc
ess
at
the
destinat
io
n
work
sta
ti
on.
So
the
wor
k
does
not
sp
e
nd
long
tim
e
in
the
qu
e
ue
wit
h
a
per
ce
ntage
of
1.6%
of
data
pr
oce
ssing
ti
m
e
and
63.
56
%
of
AGV
idle
po
sit
io
ns
as
the
fastest
tim
e
and
inclu
ded
in
the m
edium
ca
te
gory.
Table
2.
T
he
pat
h
sel
ect
ed fr
om
the predece
s
so
r
p
at
h
W
o
rks
tatio
n
No
d
es
Ru
te
Sh
o
rtest Path
1
1
12
-
1
40
2
2
12
-
1
-
2
-
11
-
12
-
1
40
3
3
1
-
2
-
3
-
10
-
11
-
12
-
1
80
4
4
1
-
2
-
3
-
4
--
9
-
10
-
11
-
12
-
1
120
5
5
1
-
2
-
3
-
4
-
5
-
8
-
9
-
10
-
11
-
12
-
1
160
6
7
1
-
2
-
3
-
4
-
5
-
6
-
7
-
8
-
9
-
10
-
11
-
12
-
1
240
7
18
1
-
2
-
3
-
4
-
5
-
6
-
7
-
18
-
17
-
16
-
15
-
14
-
3
-
12
-
1
280
8
20
1
-
2
-
3
-
4
-
5
-
8
-
17
-
20
-
21
-
22
-
23
-
24
-
13
-
12
-
1
280
9
21
1
-
2
-
3
-
4
-
9
-
16
-
21
-
22
-
23
-
24
-
13
-
12
-
1
240
10
22
1
-
2
-
3
-
10
-
15
-
22
-
23
-
24
-
13
-
12
-
1
200
11
24
1
-
2
-
11
-
14
-
23
-
24
-
13
-
12
-
1
120
12
13
12
-
1
-
12
-
13
40
Table
3.
Forwa
rd tracki
ng stat
e
AGV
Label
W
o
rks
tatio
n
No
d
es
Ro
u
te State
(s,
i,
t
)
Ti
m
e
wind
o
ws
F(s,
i,
t)
Load
Un
lo
ad
AGV1
W
ar
eh
o
u
se
1
1
12
-
1
-
-
80
DMD
2
2
1
-
2
-
11
-
12
-
1
4
1
.80
7
4
.74
40
Die ca
stin
g
3
3
1
-
2
-
3
-
10
-
11
-
12
-
1
9
7
.47
2
4
3
.82
80
Machin
in
g
4
4
1
-
2
-
3
-
4
-
9
-
10
-
11
-
12
-
1
2
7
6
.59
4
2
2
.85
120
Pain
tin
g
5
5
1
-
2
-
3
-
4
-
5
-
8
-
9
-
10
-
11
-
12
-
1
4
6
5
.66
5
9
1
.91
160
AGV2
Plastic
Injectio
n
6
7
12
-
11
-
10
-
9
-
8
-
7
-
8
-
9
-
10
-
11
-
12
-
1
7
2
.99
2
0
0
.94
200
W
eld
in
g
7
18
1
-
2
-
11
-
10
-
9
-
8
-
17
-
18
-
17
-
16
-
15
-
14
-
13
-
12
-
1
2
7
3
.52
3
6
9
.93
280
Ass
e
m
b
ly
8
20
1
-
2
-
11
-
10
-
9
-
8
-
17
-
20
-
17
-
8
-
9
-
10
-
11
-
12
-
1
4
4
2
.50
5
2
8
.89
280
AGV3
Rep
air
9
21
13
-
24
-
23
-
22
-
21
-
22
-
10
-
15
-
14
-
13
-
12
-
1
5
7
.90
1
3
7
.86
200
Fin
al
Ins
p
ectio
n
10
22
1
-
2
-
11
-
10
-
15
-
22
-
15
-
14
-
13
-
12
-
1
1
9
0
.60
2
9
6
.91
200
PPIC
11
24
1
-
2
-
11
-
14
-
23
-
24
-
13
-
12
-
1
3
4
9
.56
4
5
5
.92
200
Parkin
g
12
13
1
-
12
-
13
-
1
-
-
80
Table
4.
T
he
a
ver
a
ge n
um
ber
of AG
Vs
i
n
th
e queue
Co
n
d
itio
n
Low
Mediu
m
Hig
h
No
.
o
f
AGVs
1
2
3
AGV sp
eeds
4
10
30
De
m
an
d
a
rr
iv
al int
erval
4
2
.7
3
2
.58
3
6
.56
Idle AGV
Pos
itio
n
in
g
8
3
.62
%
6
3
.56
%
7
1
.42
%
Proces
sin
g
1
.8 %
1
.6 %
1
.7%
4.
CONCL
US
I
O
N
This
pa
per
pre
sent
a
bid
irect
ion
al
path
la
yout
with
a
r
outi
ng
al
gorithm
that
al
lows
AG
V
ta
sk
t
o
com
plete
pick
-
up
/
dro
p
-
off
(P
/
D)
.
The
path
l
ay
ou
t
co
ns
ist
s
of
tw
o
par
al
le
l
li
near
li
nes
as
su
m
ing
a
le
ng
t
h
of
40
cm
that
is
connecte
d
by
ver
ti
ces
as
m
a
ny
as
24
po
i
nts
co
nn
ect
e
d
to
a
pr
e
determ
ined
P/D
w
ork
s
ta
ti
on
.
A
com
bin
at
io
n
of
t
he
di
j
ks
tra
an
d
fl
oyd
-
war
s
hall
al
gor
it
h
m
s
prov
e
d
to
be
e
ff
ic
ie
nt
giv
e
n
for
the
AGV
route
on
a
pr
e
determ
ined
pa
th.
T
he
res
ults
of
t
he
de
velo
pm
ent
of
thes
e
two
al
go
rith
m
s
can
be
use
d
as
an
auto
nom
ous
m
ulti
-
AG
V
con
t
ro
ll
er
in
ov
e
rc
om
ing
confli
ct
-
fr
e
e
routes
in
the
m
anufactu
rin
g
industry
war
e
hous
e
.
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
Con
fl
ic
t
-
fre
e
d
ynamic
r
oute
m
ulti
-
A
GV
u
si
ng
dijkst
ra f
loy
d
-
w
ar
s
ha
ll
…
(
So
li
ch
udin
)
3603
REFERE
NCE
S
[1]
W
.
Malopol
ski,
“
A
Sus
ta
ina
ble
an
d
Confli
ct
-
Free
Opera
ti
on
of
AG
V
s
in
a
Sq
uar
e
Topol
og
y
,
”
Computers
and
Industrial
Eng
in
ee
ring,
vol
.
126
,
pp.
472
-
481,
20
18.
[2]
I.
F.
A.
Vis,
“
Surve
y
of
Res
ea
rch
in
Th
e
Design
and
Control
of
Aautomate
d
Gui
ded
Vehi
cl
e
S
y
s
te
m
s,”
Europea
n
Journal
Operati
onal
R
ese
ar
ch,
v
ol.
170
,
no
.
3
,
pp
.
677
-
709
,
2006
.
[3]
W.
J.
Hs
u,
et
al.
,
“
Schedul
ing
and
Routi
ng
Algorit
hm
s
for
AGVs:
a
Survey
,
”
Inte
rnational
Jo
urnal
Product
io
n
Re
search,
vo
l
.
40
,
no
.
3
,
pp.
745
-
760,
20
02
.
[4]
A.
La
ng
evi
n
,
D
.
L
auz
on
and
D.
Riopel,
“
Dispatc
h
ing,
Rou
ting,
and
Schedu
li
ng
of
Two
Autom
at
ed
Guid
e
d
Vehic
l
es
in
a
Fl
e
xibl
e
Manufa
ct
u
ring
S
y
st
em,”
In
t
.
J
.
Flex
ibl
e
Ma
nufac
turing
Syst
em,
vol
.
8
,
pp
.
2
47
-
262,
1996
.
[5]
E.
Gawri
low,
M.
Klimm
,
R.
H.
Mo,
and
B.
Stenz
e
l,
“
Confl
i
ct
-
Free
Vehi
cl
e
Routi
ng,
”
Euro
Journal
Tr
anspot
Logis
ti
c,
vol. 1,
pp.
87
-
111
,
201
2.
[6]
X.
Sun
,
et
al.
,
“
Schedul
ing
Multi
ple
AG
Vs
with
D
y
n
amic
Ti
m
e
-
windows
for
Sm
art
Indoor
Parking
Lo
t,”
in
2018
I
EEE
22
nd
Int
.
Conf
ere
n
ce
on
Computer
Supported
Coop
erat
ive
Work
in
Design
,
pp
.
864
-
868
,
2018
.
[7]
V.
Jaiga
nesh,
D.
K.
J
a
y
asha
nk
ar
,
and
J.
Girij
adevi,
“
Autom
at
ed
Guided
Vehic
l
e
with
Roboti
c
Lo
gisti
cs
S
y
stem
,
”
in
Proc
edi
a
Eng
ine
ering
,
vol
.
97
,
pp
.
2011
-
2021
,
2014.
[8]
D.
Antakly
,
J
-
J.
Loi
sea
u
,
and
R
.
Abbou
,
“
A
Te
m
poris
ed
Confli
ct
-
Free
Rout
ing
Policy
for
AG
Vs
,
”
IFA
C
-
Pa
p.
,
vol.
50
,
no
.
1
,
pp
.
1
-
7
,
2018
.
[9]
Q.I.
U.
L
ing
and
H.S.
U.
W
en
-
ji
n
g,
“
Confli
c
t
-
Fre
e
AG
V
Routi
ng
in
a
Bi
-
Dir
ectio
nal
Path
Lay
ou
t,”
In
Proc
ee
ding
s
of
th
e
5th
Int
ernati
onal
Con
fe
ren
ce
on
Computer
Inte
grated
Manu
fac
turing
,
v
ol
.
1
,
pp.
392
-
403
,
20
00.
[10]
S.
Pe
y
er
,
D.
Raut
enbach,
and
J.
V
y
gen
,
“
A
Gen
e
ral
i
za
t
ion
of
Dijkstra
’s
Shortest
Path
Algorit
hm
with
Applic
ation
s
to
VLSI Rout
ing
,
”
Journal
o
f Di
screte
Al
gorithm
s,
vol. 7, no. 4, p
p.
377
-
390
,
200
9.
[11]
M.
Ham
ze
ei,
R.
Z
an
ji
ran
i,
and
H.
Rashidi
-
B
e
jga
n,
“
An
Exa
c
t
and
A
Sim
ula
te
d
Anne
al
ing
Algorit
hm
for
Sim
ult
ane
ousl
y
Dete
rm
ini
ng
Fl
ow
Path
and
T
he
Locat
ion
of
P/D
Stat
ions
in
Bidi
recti
on
al
Path,”
Journal
of
Manufac
turing
S
yste
m,
vo
l. 32, n
o.
4
,
pp
.
648
-
65
4,
2013
.
[12]
T.
Nishi
an
d
Y
.
Ta
n
aka,
“
Petri
Net
Dec
om
positi
on
Approac
h
for
Dispatc
hing
and
Confli
c
t
-
Free
Routi
ng
of
Bidi
re
ct
ion
al
Autom
at
ed
Guide
d
Vehic
le
S
y
s
tem
s,”
in
The
4th
IEE
E
Inte
rnat
i
onal
Confe
renc
e
on
Aut
omatio
n
Sci
en
ce and
Eng
ine
ering
,
vol
.
42
,
no
.
5
,
pp
.
1230
-
1243
,
2012
.
[13]
B.
Rahna
m
a,
S.
Mem
ber
,
K.
Ebedi,
and
H.
M.
Sa
deghi
,
“
Self
-
Cor
rec
t
ive
Casc
ade
Control
Obs
ta
cle
Avoidance
and
Devia
t
ion
Corr
e
ct
ion
S
y
st
em for
Roboti
cs
S
y
st
e
m
s,”
In
IE
EE R
O
-
MAN.
I
EE
E
,
pp.
133
-
136
,
20
13.
[14]
A.
Baha
ri
,
“
Autom
at
ed
Guided
Vehic
l
es
Routi
ng,
”
Te
chn
ic
al
Journal
of
Enginee
ring
and
Ap
pli
ed
Sc
ie
n
ce
s
,
vol.
4
,
no
.
2
,
pp
.
60
-
66,
2014
.
[15]
N.
V.
Kum
ar
an
d
C.
S.
Kum
ar,
“
Deve
lopment
o
f
Coll
ision
Fre
e
Path
Planni
ng
Algorit
hm
for
W
are
house
Mobile
Robot,
”
in
Procedia
Computer
S
c
ie
nc
e
,
vol. 133,
pp.
456
-
463
,
20
18.
[16]
J.
P.
Ko
,
J.W
.
Jung,
and
J.
W
.
Je
on,
“
Anti
-
Coll
isi
on
Method
for
AG
V
Us
ing
RF
I
D
and
Z
igBe
e
Network,
”
in
13
th
Inte
rntional
Con
fe
renc
e
on
Contr
ol,
Aut
omation
a
nd
Syste
ms
,
pp.
599
-
604
,
2013
.
[17]
M.
Polanc
z
y
k
,
M.
Strze
l
ec
k
i
a
nd
K.
Slot,
“
Obs
ta
cle
Avoidan
ce
Proce
du
re
a
nd
Le
e
Algorithm
Based
Path
Repl
ann
er
for
Autonom
ous Mob
il
e
Plat
form
s,”
I
nt
.
J
.
of
Elec
tron
ic
s and
Te
lecomm
,
vol
.
59
,
no
.
1
,
pp.
85
-
91,
2013
.
[18]
R.
Ta
i
,
et
al
.
,
“
A
Ti
m
e
-
Eff
icien
t
Approac
h
to
Solve
Confli
c
ts
and
Dea
dloc
ks
for
Scheduling
AG
Vs
in
W
are
housing
Applic
a
ti
ons,”
i
n
Inte
rnat
ional
Confe
renc
e
on
R
eal
-
ti
m
e
Comput
ing
and
Robot
i
cs
,
pp
.
166
-
171
,
2
018.
[19]
K.
Arutselva
n
,
and
A.
Vijay
ak
um
ari
,
“
As
si
stive
Autonom
ous
Gro
und
Vehic
les
in
S
m
art
Gri
d,
”
in
Proce
d
ia
Technol
ogy
,
vo
l. 21, pp. 232
-
239
,
2015
.
[20]
Y.
W
an,
e
t
al
.
,
“
Control
le
r
Desi
gn
for
Avoiding
Coll
isions
in
A
utomate
d
Guide
d
Vehic
l
e
S
y
st
e
m
s
Via
La
beled
Petri
Ne
ts,”
IF
A
C
-
Pape
rs
OnL
ine
Conf
ere
nc
e
Pa
per
Archive
,
vo
l. 51, no
.
7,
pp.
13
9
-
144,
2018
.
[21]
C.
M.
Kle
i
and
J.
Kim
,
“
AG
V
Di
spatc
hing
,
”
In
te
r
nati
onal
Journal
Product
ion
Re
s
earc
h,
vol
.
34
,
n
o.
1,
pp
.
95
-
110
,
2007.
[22]
U.A.
Um
ar,
M.K.A.
Ariffi
n,
N.
Ism
ai
l,
and
S.
H
.
Ta
ng
,
“
Confli
c
t
-
Free
Autom
at
e
d
Guided
Vehicl
es
Routi
ng
Us
ing
Multi
-
Obje
ct
iv
e
Gene
tic
Algorithm
,
”
Re
search
Journal
of
Appl
i
ed
scie
nc
e,
Enginee
ring
and
Tec
hnology
,
vol
.
6
,
no.
14
,
pp
.
2681
-
2684,
2013
.
[23]
A.
I.
Corré
a,
A.
La
ngev
in,
and
L
.
Rouss
ea
u,
“
Sc
hedul
ing
and
R
outi
ng
of
Auto
m
at
ed
Guided
Vehic
l
es:
A
H
y
brid
Approac
h,
”
in
C
omputers
&
Ope
rations R
ese
arch
,
vol
.
34
,
no
.
6
,
p
p.
1688
-
1707
,
2
007.
[24]
C.
Li
u
,
et
al
.
,
“
Path
Plann
ing
an
d
Inte
l
li
gen
t
Sch
edul
ing
of
Multi
-
AG
V
Sy
stems
i
n
W
orkshop,”
in
Proceedi
ngs
of
The
36th
Ch
ine
s
e
Control
Confer
enc
e,
pp.
2735
-
2739
,
2017
.
[25]
Z.
Zh
ang,
Q.
G
uo,
J.
Chen
,
an
d
P.
Yuan,
“
Col
li
sion
-
Free
Rout
e
Planni
ng
for
Multi
ple
AG
Vs
in
an
Autom
at
e
d
W
are
house
Base
d
on
Col
li
sion C
la
ss
ifi
c
at
ion
,
”
in
IEE
E
Acce
ss
,
vo
l.
6
,
pp
.
26022
-
2
6035
,
2018
.
[26]
D.
Gigli
o
,
“
Ta
s
k
Scheduling
fo
r
Multi
pl
e
Forkl
ift
AG
Vs
in
Dis
tri
buti
on
W
are
h
ouses,”
in
Eme
r
ging
Techno
log
y
and
Factory Aut
omation
,
pp
.
1
-
6
,
2014
.
[27]
N.
A.
El
-
sh
erb
en
y
,
“
The
Algor
ithm
of
The
Ti
m
e
-
Depe
ndent
Shortest
Path
Proble
m
with
Ti
m
e
W
indows,”
Applied
Mathe
mati
cs
,
vo
l.
5
,
no
.
17
.
pp
.
2
764
-
2770,
2014
.
[28]
N.
A.
El
-
Sherb
e
n
y
,
“
Vehic
l
e
Ro
uti
ng
with
Ti
m
e
W
indows:
An
Overvi
ew
of
Ex
ac
t
,
Heuri
st
ic
a
nd
Meta
heu
ristic
Methods,
”
Journ
al
of
King
Saud
Unive
rs
it
y
-
S
ci
e
nce
,
vol
.
22
,
no
.
3,
pp
.
123
-
131
,
2010.
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.
10
, No
.
4
,
A
ugus
t
2020
:
3596
-
3604
3604
BIOGR
AP
H
I
ES
OF
A
UTH
ORS
Solic
hu
din
,
obta
ine
d
h
is
Bac
h
el
or
of
Educ
a
tion
(S.Pd),
fro
m
the
Depa
rtm
ent
of
El
e
ct
ri
cal
Engi
ne
eri
ng
Edu
ca
t
ion,
Sem
ara
n
g
Stat
e
Univer
si
t
y
(UN
NES).
Curre
ntly
pursuing
educ
a
ti
on
in
th
e
Master
of
El
e
ct
ri
ca
l
Engi
n
eering
Pos
tgra
dua
te
Program
,
D
ipone
goro
Univ
ersity
(UN
DIP
)
Sem
ara
ng.
Aris
Tri
w
i
y
a
tno
,
Rec
e
ive
ST.
,
MT.
,
and
Dr.
degr
ee
s
in
Depa
rtment
of
El
ectr
ic
a
l
Engi
ne
eri
ng
,
Facul
t
y
of
Eng
ine
er
ing,
Sepu
l
uh
Novem
ber
Instit
ute
of
T
echnolog
y
(IT
S)
Suraba
y
a.
Now
a
per
m
ane
nt
l
e
ct
ure
r
in
th
e
Depa
rtment
of
El
e
ct
ri
ca
l
Engi
n
ee
ring
,
Facult
y
of
Engi
nee
rin
g,
Dipon
egor
o
Univer
sit
y
wi
th
rese
arc
h
in
te
r
est:
Artifi
c
ia
l
Int
el
l
i
gent
,
Software
Engi
ne
eri
ng
an
d
Control
S
y
s
te
m
.
Munaw
a
r
A.
Riy
adi
,
Recei
ved
degr
ee
ST
.
and
MT.
at
th
e
School
of
El
ectri
ca
l
and
Inform
at
io
n
Engi
ne
eri
ng,
B
andung
Instit
ute
of
Te
chnol
og
y
(IT
B).
P
hD
d
egr
ee
obt
ai
ned
from
Univer
siti
Te
knik
al
Mal
a
y
s
ia
Mel
aka
(UT
e
M).
His
rese
ar
ch
int
er
ests:
ga
te
a
rra
y
s
tha
t
ca
n
be
progra
m
m
ed
in
the
f
ie
ld
,
b
i
om
edi
ca
l
m
eas
ure
m
ent
s,
s
y
nthe
sis
con
trol
s
y
stems
,
cr
y
ptogr
aph
y
,
a
nd
m
ic
roc
ontrollers
.
Now
a
p
erman
ent
l
ecture
r
in
t
he
Depa
r
tment
of
Elec
tr
ical
En
gine
er
ing,
Fa
cult
y
of
Engi
n
ee
rin
g,
Diponegor
o
Uni
ver
sit
y
.
Evaluation Warning : The document was created with Spire.PDF for Python.