Indonesi
an
Journa
l
of El
ect
ri
cal Engineer
ing
an
d
Comp
ut
er
Scie
nce
Vo
l.
23
,
No.
1
,
Ju
ly
2021
, p
p.
549
~
557
IS
S
N: 25
02
-
4752, DO
I: 10
.11
591/ijeecs
.v
23
.i
1
.
pp
549
-
557
549
Journ
al h
om
e
page
:
http:
//
ij
eecs.i
aesc
or
e.
c
om
Appl
ication
pr
oject t
ask schedu
lin
g using d
olphi
n swarm
tec
hn
ology
Bara
a S.
M
os
t
afa, Fi
rdew
s
A.
Alsa
lm
an
Depa
rtment
o
f
s
oftwa
re Engi
n
eering
,
Co
ll
eg
e
of
Com
pute
r
Scie
n
ce
and
Ma
the
m
a
ti
cs
,
Univer
sit
i
o
f
Mos
ul
(UO
M)
,
Ira
q
Art
ic
le
In
f
o
ABSTR
A
CT
Art
ic
le
hi
story:
Re
cei
ved
Dec
3,
2020
Re
vised Ju
n 8
,
2021
Accepte
d
J
un
1
5
, 202
1
Projec
t
ta
sk
sch
edul
ing
,
which
i
s
par
t
of
softwa
re
project
m
an
a
gement,
h
as
gre
at
importan
c
e
in
the
succ
ess
or
fai
lure
of
pro
je
c
ts
as
40%
of
proje
c
ts
are
subjec
t
to
fa
il
ur
e
due
to
poor
pl
anni
ng
and
m
isu
se
of
hum
an
resourc
es,
and
thi
s
gre
atl
y
conf
irms
tha
t
the
wi
se
dec
isions
tha
t
are
m
ade
at
th
i
s
stage
le
ad
to
red
uce
d
Obs
ta
cles
fac
ing
work
and
le
adi
ng
to
succ
ess.
The
work
and
dec
ision
-
m
aki
ng
is
the
responsibil
ity
of
the
proj
e
ct
m
ana
ger
as
he
fac
es
all
the
complexit
y
a
nd
cri
tical
phase
s
of
thi
s
cri
ti
cal
phase
and
dea
ls
with
al
l
it
s
rel
a
te
d
pro
ce
ss
e
s
which
consist
of
def
ini
ng
the
proje
c
t
ta
sks
an
d
assigning
the
emplo
y
e
es
responsible
for
e
ac
h
ta
sk
and
th
e
n
pre
par
ing
proj
ec
t
pl
ans.
In
thi
s
res
e
arc
h
we
dea
l
t
with
pro
jec
t
sche
dul
ing
usin
g
t
he
sm
art
do
lp
hin
sw
arm
al
gorit
hm
,
th
e
proje
c
t
sche
dul
i
ng
proc
ess
is
essenti
a
l
in
the
software
engi
ne
eri
ng
indu
str
y
nowad
a
y
s
a
nd
the
poss
ibilit
y
of
findi
ng
goo
d
soluti
ons
which
save
s
the
software
compan
y
a lot
o
f
t
ime
a
nd
m
one
y
.
Ke
yw
or
d
s
:
D
ol
ph
i
n
intel
li
gen
t al
gorithm
Pr
oject
sc
he
du
l
ing
S
of
t
war
e
engin
eerin
g
Sw
arm
intel
li
gen
ce al
go
rithm
s
This
is an
open
acc
ess arti
cl
e
un
der
the
CC
B
Y
-
SA
l
ic
ens
e.
Corres
pond
in
g
Aut
h
or
:
Ba
raa S.
Mosta
fa
Dep
a
rtm
ent o
f S
of
t
war
e
E
ng
i
neer
i
ng
Un
i
ver
sit
i o
f
Mosu
l
(UOM)
Coll
ede
of
C
om
pu
te
r
Scie
nc
e an
d
Ma
them
at
ic
,
Ir
a
q
Em
a
il
:
bar
aa_s
a
m
i@uo
m
os
ul.
edu.iq
1.
INTROD
U
CTION
So
f
t
war
e
proj
e
ct
sched
uli
ng
pro
blem
(
SPSP
)
is
a
cri
ti
cal
process
of
assig
ni
ng
ta
sk
s
to
e
m
plo
ye
es
in
a
proj
ect
s
o
th
at
com
pletio
n
tim
e
and
cost
are
re
du
ce
d.
It
diff
e
rs
f
ro
m
t
he
well
-
known
resour
ce
-
c
on
s
trai
ne
d
pro
j
ect
sche
du
li
ng
pr
ob
le
m
(RCPSP)
i
n
tha
t
there
ar
e
t
wo
reducti
on
goa
ls
wh
il
e
the
R
CPSP
on
ly
ha
s
one
(ti
m
e
to
co
m
pleti
on
)
,
an
d
al
s
o
in
SP
SP
the
e
m
plo
ye
es
ar
e
the
on
ly
s
ource,
eac
h
with
a
set
of
sk
il
ls
and
a
sal
ary,
b
ut
in
RC
PSP
there
a
re
m
any
reso
urces
in
qu
a
ntit
at
ive
qu
a
ntit
ie
s
[1
]
.
SPSP
has
a
direct
i
m
p
act
on
the
su
ccess
of
the
pro
j
ect
,
it
is
th
e
pro
j
ect
m
anag
em
ent
act
ivity
in
w
hich
the
pro
j
ect
m
anager
is
re
spo
ns
ib
le
for
and
he
m
us
t
us
e
di
ff
e
ren
t
te
chn
i
qu
e
s
an
d
m
et
ho
dolo
gies
to
m
anag
e
the
pe
rs
onnel
and
t
h
e
ta
sk
s
of
t
he
so
ft
war
e
pro
j
e
ct
[2]
.
Th
us
,
t
he
pro
cess
of
sche
du
li
ng
s
of
tware
e
nginee
r
ing
pro
j
ect
s
is
an
im
po
rta
nt
process
that
work
s
to
a
ll
ocate
e
m
plo
ye
es
to
work
on
the
ta
sk
s
of
s
of
t
war
e
pr
oj
ec
ts
[3
]
,
as
there
are
m
any
sched
ule
s
for
a
sin
gle
proj
ect
[4
]
,
but
they
dif
fer
i
n
ti
m
e
and
c
os
t
an
d
the
best
sc
h
e
du
le
m
us
t
be
c
ho
s
en
a
nd
the
pro
j
ect
com
plete
d
in
t
he
le
ast
po
ssib
le
tim
e
and
cost,
wh
ic
h
inc
re
ases
.
The
com
plexity
of
the
issue
is
the
confli
ct
of
these
obje
ct
ive
s
(project
com
pleti
on
ti
m
e
an
d
pro
j
ect
co
st)
[5
]
,
a
nd
t
his
in
creases
the
diffi
culty
of
fin
din
g
the
ideal
so
l
ution.
Ther
e
f
or
e,
the
r
e
are
t
wo
diff
e
ren
t
te
c
hn
i
qu
e
s
to
s
olv
e
this
te
chn
ic
al
iss
ue,
t
he
first
is
t
he
use
of
the
f
ull
exam
i
nation
m
et
hod
that
exam
ines
al
l
po
ssible
di
stribu
ti
ons
or
al
locat
ion
.
T
he
te
chn
iq
ue
will
waste
tim
e,
t
he
seco
nd
te
ch
niq
ue
is
the
inc
om
plete
exam
inati
on
t
echn
i
qu
e
that
e
xam
ines
on
ly
a
pa
rt
of
the
pos
sible
distrib
utions
or
al
locat
ion
in
orde
r
to
fin
d
the
so
luti
on
cl
ose
st
to
the
ideal
so
luti
on
in
a
n
acce
ptable
ti
m
e
and
cost less
[6]
.
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
S
N
:
2502
-
4752
Ind
on
esi
a
n
J
E
le
c Eng &
Co
m
p
Sci,
Vo
l.
23
, N
o.
1
,
Ju
ly
2021
:
549
-
557
550
This
te
chn
i
que
includes
he
ur
ist
ic
s
an
d
m
et
aheu
risti
cs,
t
he
m
ai
n
dis
ti
nction
betwe
en
infe
renc
e
m
et
ho
ds
a
nd
m
et
aheu
risti
cs
is
that
infer
en
ce
are
m
e
tho
ds
desig
ne
d
to
so
lve
a
s
pecifi
c
prob
le
m
on
ly
wh
il
e
m
et
aheu
risti
cs
can
be
a
dap
ta
bl
e
to
so
lve
any
i
m
pr
ov
em
ent
pro
blem
,
and
their
resea
rc
h
does
not
de
pe
n
d
on
the p
r
operti
es o
f
the pr
ob
le
m
, th
eref
ore, it
b
ecom
es intuit
i
ve
to choo
s
e m
et
hods
abo
ve
m
et
aheu
risti
cs,
su
ch
as
al
gorithm
s
bas
ed
on
the
intel
li
gen
ce
of
the
swar
m
,
wh
ic
h
fin
d
the
cl
os
es
t
so
luti
on
to
th
e
op
ti
m
a
l
so
luti
on
in
record
ti
m
e
com
par
ed
to
the
tr
aditi
on
al
m
et
hods
[
7].
T
he
diff
ic
ulty
of
this
pr
oble
m
has
ge
ner
at
e
d
m
uch
interest
to
s
olve
sche
du
li
ng
pro
blem
sta
rting
with
im
ple
m
e
ntati
on
the
m
axim
u
m
and
m
i
nim
u
m
ant
al
g
or
it
hm
with
exc
essive
cub
ic
fr
am
e
t
o
s
olv
e
the
SP
SP
pr
ob
le
m
a
nd
c
om
par
ed
i
ts
resu
lt
s
w
it
h
the
res
ults
of
oth
e
r
al
gorithm
s
and
it
achieve
d
t
he
best
res
ults
wh
e
n
t
he
data
we
re
sim
ple
and
unc
om
plicated
[
8],
dif
fere
ntial
evo
l
ution
al
gorithm
and
com
par
e
d
it
s
res
ults
with
the
genet
ic
al
go
rithm
in
the
SPS
P
issue
[
9]
,
an
d
use
d
th
e
diff
e
re
ntial
evo
l
utio
n
m
et
ho
d
with
a
ne
w
br
ea
kthro
ugh
te
chnolo
gy
in
s
olv
in
g
t
he
SP
S
P
pro
blem
and
te
ste
d
the
al
gorithm
us
in
g
ra
nd
om
l
y
gen
e
rated
da
ta
and
c
om
pared
it
s
res
ults
with
ot
her
al
gorithm
s
and
it
reache
d
the
be
st
res
ults
in
a
shorte
r
ti
m
e
than
t
he
re
st
of
the
al
gori
t
hm
s
[1
0].
U
si
ng
an
ev
ol
utio
nar
y
hype
r
-
i
nference
al
gorithm
to
so
lve
t
he
S
PSP
pro
blem
[1
1],
lum
ino
us
bu
tt
e
rf
li
es
al
gorit
hm
to
so
l
ve
the
SPS
P
prob
le
m
.
They
com
par
ed
it
s
r
esults
with
t
he
resu
lt
s
of
t
he
ge
netic
al
gorithm
and
the
a
nt
col
on
y
al
go
rithm
The
lum
inou
s
bu
tt
er
flie
s
al
gorithm
achieved
t
he
best
resu
lt
s
fo
r
sm
al
l
dat
a
[12]
.
In
this
researc
h
the
s
m
art
do
lp
hin
s
war
m
al
gorithm
was
us
e
d
[
13
]
.
I
nsp
ired
by
the d
yn
a
m
ic
and
li
ving
cha
racteri
sti
cs
of
th
e
floc
k
of
dolp
hin
s
i
n
na
ture,
in
orde
r
to
so
l
ve
the
issue
of
s
c
hedulin
g
the
pro
j
ect
ta
sk
s
i
n
the
le
ast
tim
e
and
cost
[14]
.
The
rest
of
the
pap
e
r
is
or
ga
nize
d
as
fo
ll
ows:
the
se
cond
sect
io
n
de
scribes
the
s
oft
war
e
p
r
oject
sche
du
li
ng
pr
oble
m
SPSP
,
the
third
sect
ion
int
rod
uc
es
the
dolp
hi
n
s
war
m
al
go
r
it
h
m
(D
S
A)
,
t
he
f
or
t
h
s
ect
io
n
intr
oduces
t
he
pro
po
se
d
work,
the
fifth
sect
ion p
r
esents
resu
lt
, t
he
la
st sect
io
n i
s a conclusi
on
.
2.
SOFTW
AR
E
PRO
JEC
T
S
CHEDULI
N
G P
ROBLE
M
(
SP
SP
)
SPSP
is
def
i
ne
d
as
the
proce
ss
in
w
hic
h
a
pro
j
ect
m
anag
er
assig
ns
pro
je
ct
ta
sk
s
o
t
he
e
m
plo
ye
es
who
ca
n w
ork on i
t, an
d f
or
t
he
m
anag
er
to do t
hat,
he
m
ust
f
irst
ob
ta
in
th
e f
ollow
i
ng infor
m
at
ion
[
15
]
:
a.
Pr
oject
ta
sk
s
(
Tasks
):
Tasks
represe
nt
op
e
r
at
ion
s
that
m
u
st
be
execu
te
d
in
a
sequ
e
ntial
m
ann
er
[16
]
,
ta
kin
g
into
acc
ount
the
prece
den
ce
of
the
se
ta
sk
s
a
nd
th
e
dep
e
ndencies
betwee
n
them
in
orde
r
for
t
he
pro
j
ect
to
be
com
plete
d.
O
ne
ta
sk
is
kn
own
as
(t_
j
)
an
d
j
represe
nt
the
values
f
r
om
(1
)
to
(T
)
t
hat
represe
nt
the
t
otal
nu
m
ber
of
pr
oj
ect
ta
sks.
Each
ta
sk
po
s
s
esses
a
set
of
s
kill
s
kn
own
as
t
j
skil
ls
,
wh
ic
h
m
us
t
be
avail
able
in
the
e
m
plo
ye
e
or
gro
up
of
em
plo
ye
es
respon
si
ble
for
the
i
m
ple
m
e
ntati
on
of
this
ta
sk
, as
w
el
l as
the e
ffor
t
requ
ired
t
o
im
ple
m
ent it
(
E
ffor
t
),
wh
ic
h
is
know
n
as
t
j
e
ff
or
t
[11]
b.
Task
P
recede
nc
e
Gr
a
ph
:
T
he
ta
sk
em
erg
ence
dia
gr
am
,
known
as
(TPG),
is
us
ed
t
o
rep
r
esent
the
pr
ece
de
nce
of tasks a
nd the
d
e
pende
ncies
between t
hem
.
c.
Em
plo
ye
es:
E
m
plo
ye
es
are
t
he
m
os
t
i
m
po
rtant
so
urce
f
or
so
ftwa
re
com
pan
ie
s
a
nd
the
y
are
the
on
ly
so
urce
f
or
SP
S
P.
The
em
plo
ye
e
is
kn
ow
n
as
(ei)
an
d
i
rep
r
esents
the
val
u
es
ro
m
(1
)
to
(
E)
that
re
pr
ese
nt
the total
nu
m
ber
of
em
plo
ye
e
s workin
g o
n
t
he pr
oject
. E
ve
ry em
plo
ye
e possesses
a set
of s
kill
s
e
i
skil
l
s
.
3.
DOLPH
IN
S
WA
RM AL
G
ORI
TH
M (
DSA)
The
DSA
has
m
any
sta
ges,
s
uch
as
pr
e
dation,
sea
rch
i
ng,
receptio
n,
cal
li
ng,
a
nd
pre
dation,
an
d
it
include
s
the p
r
edato
ry
proces
s
of
dolp
hin
s [17
]
,
a
nd
these h
abits
a
nd
c
harac
te
risti
cs
help
the
dolp
hin
s
a
chiev
e
their
goal
duri
ng
the
pr
e
dation
proce
ss.
T
he
dolp
hin
s
warm
a
lgo
rithm
is
co
ns
ist
ent
wit
h
no
ti
o
ns
of
s
war
m
intel
li
gen
ce
[
18]
,
but
d
if
fers
from
tradit
i
on
al
s
war
m
i
ntell
igence
al
gorithm
s
[1
9].
Dep
e
ndin
g
on
the
intel
li
gen
ce
of
the
swa
rm
,
a
s
pecific
num
ber
of
dolp
hi
ns
ar
e
need
e
d
to
si
m
ula
te
the
li
fe
hab
it
s
an
d
bio
l
og
ic
al
char
act
e
risti
cs
ind
ic
at
ed
in
t
he
act
ual
pr
e
dat
or
y
p
r
ocess
of
the
dol
ph
i
ns
,
F
igure
1
s
how
n
the
fl
ow
c
ha
rt
of
t
he
do
l
ph
i
n
s
war
m
algorit
hm
[
11]
, and
DSA ca
n be
div
ide
d
i
nto five
stages
as
fo
ll
ows
[
20
]
:
In
it
ia
ti
on
sta
ge
:
initial
do
lph
i
n
swa
rm
sp
awn
eve
nly
and
r
andom
ly
,
Do
l
i
=
[x
1
,
x
2
,
.
..,
X
D
]
T
(i
=
1,
2,
.
.
.,
N)
Wh
e
re
N
is
th
e
nu
m
ber
of
do
lp
hin
s
.
X
j
r
epr
ese
nts
a
co
m
po
nen
t
relat
ed
to
each
of
th
e
di
m
ension
s
t
o
be
im
pr
oved
. For eve
ry dolp
hi
n,
t
her
e
are
t
w
o
ide
ntica
l va
riants:
a)
L
i
is t
he pe
rf
ec
t solutio
n
t
hat
Do
li
fou
nd at o
ne
ti
m
e.
b)
k
i
Neig
hborh
ood i
s the
b
est
s
olu
ti
on.
Wh
e
re
i
=
1
,2,
...,
N.
Af
te
r
c
onfig
ur
i
ng
the
value
of
t
he
do
l
ph
i
ns
,
cal
culat
e
the
fitnes
s
of
eac
h
of
the
do
l
ph
i
ns
a
nd
obta
in a
Fit
k
. F
it
k
= {Fit
k,1
, F
it
k, 2
, .
.
.,
Fit
k, N
} [
21]
Searc
h
sta
ge
:
In
t
his
sta
ge,
e
ach
of
the
dolp
hin
s
perf
or
m
s
a
search
f
or
t
he
ir
surr
oundin
g
area
by
m
aking
so
un
ds
i
n
ra
nd
om
M
directi
on
s.
Like
wise,
t
he
s
ound
ca
n
be
represe
nted
in
the
f
ollo
wing
way:
V
i
=
[
v
1
,
v
2
,
.
..,
V
D
]
T
,
(i
=
1,
2,
.
..,
M)
in
the
c
urren
t
s
earch
,
M
sta
nd
s
for
nu
m
ber
of
sou
nd
s
an
d
v
j
,
(
V
i
=
[
v
1
,
v
2
,
...
,
V
D
]
T,
(i
=
1,
2,
..
.,
M)
in
the
curre
nt
search
,
M
sta
nd
s
f
or
nu
m
ber
of
s
ounds
an
d
V
j
,
(j
=
1,
2,
...,
D
)
Evaluation Warning : The document was created with Spire.PDF for Python.
Ind
on
esi
a
n
J
E
le
c Eng &
Co
m
p
Sci
IS
S
N:
25
02
-
4752
Ap
plicati
on pr
oject
task sc
he
du
li
ng
us
i
ng dolp
hin
swar
m
t
echnolo
gy
(
B
araa S
. M
os
t
afa)
551
represe
nt
the
c
om
po
ne
nt
of
e
ach
of
t
he
di
m
ension
s
that
are
the
s
ound
char
act
e
risti
c
of
t
he
directi
on.
More
ov
e
r,
the
sou
nd
s
sat
isfy
||
V
i
||
=
velo
ci
ty
,
(i
=
1,
2,
...,
M)
,
T
he
"
velocit
y"
re
pr
e
sents
a
c
onsta
nt
equ
i
valent
to
t
he
sou
nd
cha
ra
ct
erist
ic
of
the
velocit
y.
I
n
or
der
t
o
preve
nt
the
dolp
hin
s
from
getti
ng
stu
ck
at
this
sta
ge,
the
T
i
:
is
m
axim
u
m
search
ti
m
e
is
set
.
At
t
he
m
axi
m
u
m
search
tim
e
T
i
,
the
voic
e
of
V
j
wh
ic
h Do
l
i
,
(
i
= 1
,
2, ...,
N)
.
It r
es
ults in t
he
tim
e w
hen it
will
search f
or
a n
e
w
s
olu
ti
on
X
ijt
, whic
h
ca
n be e
xpresse
d
a
s:
=
+
∗
(1)
Fo
r
the
new
s
olu
ti
on
Xij
t
th
at
Do
l
i
is
getti
ng,
the
fitness
of
E
ijt
can
be
cal
culat
ed
usi
ng
t
he
f
ollo
wi
ng
form
ula:
=
(
)
(2)
If
(Eiα
b=m
in
j
=1,2,…
.M;
t=
1,2
,….,T1
E
ijt
)
.
In
this
case,
th
e
ind
i
vidual
opti
m
al
L
i
of
D
ol
i
so
luti
on
will
be
determ
ined
:
=
(3)
If
(F
it
nes
s
(L
i
)
<
Fit
nes
s
(
K
i
))
.
K
i
is
re
placed
by
L
i
;
Othe
r
wise,
i
n
order
no
t
t
o
c
ha
ng
e
.
Af
te
r
eac
h
Do
l
i
update
(i
=
1,
2,
...,
N)
thei
r
L
i
and
K
i
(i
n
case
they
can
be
up
dated
),
the
DSA
goes
into
the
c
al
l
sta
ge
[
22]
.
Ca
ll
sta
ge
In
t
his
sta
ge,
each
of
the
dolp
hins
produces
sou
nd
s
to
in
f
or
m
t
h
e
oth
e
r
dolp
hin
s
of
the
resu
lt
s
in
the
resea
rc
h
sta
ge
,
wh
ic
h
incl
ud
e
s
w
he
ther
a
m
or
e
adequate
s
olu
t
ion
has
bee
n
ob
ta
ine
d
a
nd
t
he
locat
ion
o
f
that
best
so
luti
on
[
23
]
. A
TS
m
at
ri
x
w
her
e
T
S
i,j
rep
rese
nts
(t
he
r
est
of
the
tim
e
for
the
au
dio
t
o
travel f
r
om
D
ol
j
to Dol
i
a
nd r
e
qu
i
res updat
in
g
in
the
f
ollow
i
ng w
ay
:
for K
i
,
K
j
a
nd TS
i,j
.
If
(
fit
ne
ss
(
k
i
)
)
<
fit
ne
ss
(
k
j
)
a
nd
TS
i
,
j
=
⌈
DDij
A
∗
spee
d
⌉
The
n
TS
i
,
j
=
⌈
DDij
A
∗
spee
d
⌉
Othe
rw
ise
,
TS
i,
j
rem
ai
ns
it
s v
a
lue.
Wh
e
re
(i =
1
,
2, ...,
N
;
j
=
1,
2, ...,
N) a
nd Dol
j
.
=
|
|
−
|
|
,
,
=
1
,
2
,
…
,
≠
j
(4)
Velocit
y
is
a
c
on
sta
nt
e
qu
i
valent
to
t
he
s
ound
property
of
ve
locit
y.
A
sym
bo
li
zes
a
co
ns
t
ant
re
pr
e
sentin
g
the
acce
le
rati
on
ca
pab
le
of
m
akin
g
s
ounds
t
rav
el
at
a
high
er
s
peed
in
cas
e
of
ve
ry
lo
w
sp
ee
d,
a
nd
the
n
TS
ij
un
derg
oes a
n update
base
d on
(
4) [2
3].
The
receivi
ng
sta
ge
i
n
t
he
DSA,
t
he
e
xc
hange
proce
ss
(
wh
ic
h
incl
udes
t
he
c
omm
un
ic
at
io
n
a
nd
the
receivin
g
phas
es)
will
be
m
a
i
ntained
with
T
S,
in
case
the
DSA
enters
t
he
receivin
g
phas
e,
each
TS
ij
te
rm
=
1,
2,.
...,
n;
j
=
1,
2,
...,
N
),
then
T
S
ij
decr
ea
ses
by
1
to
in
di
cat
e
that
the
s
ounds
pro
pa
ga
te
ov
er
on
e
unit
of tim
e. I
n
t
his
case,
DS
A
r
e
quires
ve
rificat
ion o
f
eac
h
T
S
ij
te
rm
in
an
a
rr
a
y:
(
=
0
)
(
5)
This
m
eans
tha
t
the
sou
nd
tha
t
is
sent
f
r
om
Do
l
j
t
o
D
ol
i
ca
n
be
ob
ta
ine
d
by
D
ol
i
,
i
n
w
hi
ch
case
t
her
e
is
a
need
to
re
place
TS
ij
with
a
ne
w
ti
m
e
te
r
m
r
efer
red
to
as
"
m
axi
m
u
m
transm
issi
on
ti
m
e"
(T
2),
to
in
dicat
e
that t
he
e
quiva
le
nt
sou
nd h
as
been recei
ved.
Ma
ke
a c
om
par
iso
n betwee
n K
i
a
nd K
j
:
(
(
)
>
(
)
)
(
6)
I
n
t
his
case,
th
en
K
i
will
be
r
eplace
d
by
K
j
;
Othe
rw
ise
,
K
i
rem
ai
ns
un
ch
ang
e
d.
A
fter
e
ach
te
rm
in
the
m
at
rix
a TS
ij
w
hich sat
isfie
s
in
(5)
is t
reate
d, an
d
the
DSA
init
ia
te
s the pre
dation p
hase
.
Pr
e
dation
sta
ge
At
this
sta
ge
,
each
of
the
do
l
ph
i
ns
is
re
qu
i
red
t
o
cal
c
ulate
the
ra
di
us
surr
oundin
g
the
do
l
ph
i
n
R2,
an
d
determ
ine
the
distance
bet
ween
the
op
ti
m
u
m
so
luti
on
for
the
dolp
hin
neigh
bor
hood
and
it
s
po
sit
io
n
aft
er
the p
re
datio
n
sta
ge
base
d
on
the
a
vaila
ble d
at
a,
a
nd
after
that,
he
obta
ins
a
ne
w
po
sit
io
n.
Fo
r
eac
h of
t
he
dolp
hin
s
, th
e
foll
ow
i
ng is cal
culat
ed
a)
Dista
nce
D
K
i
:
=
|
|
−
|
|
,
=
1
,
2
,
…
(7)
b)
DKl
i
Dista
nce:
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
S
N
:
2502
-
4752
Ind
on
esi
a
n
J
E
le
c Eng &
Co
m
p
Sci,
Vo
l.
23
, N
o.
1
,
Ju
ly
2021
:
549
-
557
552
=
|
|
−
|
|
,
=
1
,
2
,
…
(8)
R1
represe
nts
the
search
rad
i
us
,
a
nd
re
pres
ents
the
m
axi
m
u
m
search
phase
ra
ng
e
,
an
d
it
can
be
cal
culat
ed
base
d on the
fol
lowing e
qu
at
i
o
n:
1
=
1
∗
(9)
In
ge
ner
al
,
the
cal
culat
ion
of
the
s
urrou
nding
ra
diu
s
R2
a
nd
the
updatin
g
of
t
he
dolp
hin
po
sit
io
n
s
ho
uld
be
discusse
d i
n
t
hree case
s:
a)
(
(
≤
)
ℎ
2
=
(
1
−
2
)
)
(10)
=
+
−
2
(11)
b)
(
>
≥
ℎ
2
=
(
1
−
−
(
)
−
−
(
)
.
1
(
)
)
)
(12)
=
+
|
|
|
|
2
(13)
c)
(
(
<
)
ℎ
2
=
(
1
−
(
)
−
−
(
)
.
1
(
)
)
)
(14)
Figure
1. Flo
w
char
t
dolp
hin
s
war
m
algo
rit
hm
Evaluation Warning : The document was created with Spire.PDF for Python.
Ind
on
esi
a
n
J
E
le
c Eng &
Co
m
p
Sci
IS
S
N:
25
02
-
4752
Ap
plicati
on pr
oject
task sc
he
du
li
ng
us
i
ng dolp
hin
swar
m
t
echnolo
gy
(
B
araa S
. M
os
t
afa)
553
Ca
lc
ulate
newDo
l
i
as
(
13
)
,
wh
e
re
e
is
a
c
on
sta
nt
great
er
than
2.
Af
te
r
Do
l
i
m
ov
es
to
the
ne
wDol
i
posit
ion,
newD
ol
i
is
com
par
ed
to
Ki
in
te
rm
s
of
fitness.
Fit
ness
(
newD
ol
i
)
<
Fit
ness
(
K
i
)
.
T
he
n
Ki
is
place
d
by
newD
ol
i
;
Othe
rw
ise
,
i
n
order
no
t
t
o
cha
nge.
Af
te
r
al
l
Do
l
i
(i
=
1,
2,
...
,
N)
updates
t
he
ir
locat
ions
a
nd
K
i
(i
n
the
case
it
m
i
gh
t
up
date),
if
sp
eci
fied
D
ol
i
m
eet
s
the
endi
ng
co
ndit
ion
.
If
the
en
ding
conditi
on
is
m
et
,
the
DSA
beg
i
ns
t
he
term
inati
on
phase.
Else, t
he DS
A beg
i
ns
t
he
searc
h ph
a
se
again
.
4.
PROP
OSE
D
WORK
Ther
e
are
m
any
chall
en
ges
f
aci
ng
resea
rchers
in
s
of
twa
r
e
pro
j
ect
m
anag
em
ent
issue
s,
es
pecial
ly
wh
e
n
co
nducti
ng
researc
h
on
plann
i
ng
a
nd
sche
du
li
ng,
one
of
them
is
t
he
la
ck
of
real
data
for
the
pl
ann
i
ng
and
sc
hedulin
g ph
ase o
f
the project
as the
softw
are c
om
pan
ie
s r
eserv
e in
form
ation
r
el
at
ed
to
e
m
plo
ye
es, their
sal
aries
and
t
he
ir
sk
il
ls
and
do
not
pro
vid
e
any
inf
or
m
at
ion
on
ho
w
to
di
vid
e
a
nd
distribu
te
the
wor
k
or
a
ny
inf
or
m
at
ion
re
la
te
d
to
m
anag
em
ent
and
le
ader
s
hip
withi
n
the
w
ork
te
a
m
.
Re
searche
rs
ha
ve
trie
d
[
24
]
to
pro
vid
e
data
F
or
real
project
s
in
cooper
at
i
on
with
a
Jo
r
da
nian
softwa
re
co
m
pan
y,
bu
t
tha
t
info
rm
at
ion
is
sti
ll
few
a
nd
lim
it
e
d,
s
o
re
searc
he
rs
res
ort
to
usi
ng
rand
om
ly
gen
e
rated
dat
a
that
si
m
ulate
real
pro
j
ect
s
wh
e
n
cond
ucting
res
earch
,
a
n
d
des
pite
the
good
r
esults
achie
ve
d
by
the
te
c
hn
iqu
es
an
d
al
gorithm
s
us
ed,
th
ey
are
sti
ll
n
ot r
ea
dy for use
in
t
oo
ls
a
nd
c
omm
ercia
l app
li
cat
io
ns
[
25
]
.
The
data
us
e
d
in
this
stud
y
a
r
e
rando
m
ly
gen
erate
d
data
th
at
si
m
ulate
rea
l
so
ftwa
re
pr
oject
s.
It
is
the
o
nly
s
ource
av
ai
la
ble
to
resea
rch
e
rs
f
or
t
he
pur
po
se
of
stu
dying
the
m
ult
i
-
pur
pose
S
PS
P
issue
,
w
hich
can
be
ob
ta
ine
d
from
[26]
,
This
data
con
sist
s
of
36
gro
up
s
,
each
gro
up
re
pr
ese
nting
one
pro
j
ect
.
These
groups
sh
ar
e
so
m
e
gen
eral
char
act
e
risti
cs
and
dif
fer
in
oth
er
cha
ract
erist
ic
s,
as
al
l
gr
ou
ps
posse
ss
10
di
ff
e
rent
sk
il
ls
avail
able
in
th
e
pro
j
ect
sta
ff
,
an
d
eac
h
em
plo
ye
e
possesse
s
6
-
7
ra
ndom
sk
il
ls
out
of
a
total
of
te
n
sk
i
ll
s,
a
s
well
as
each
ta
sk
po
s
sesses
6
-
7
rand
om
sk
ills
ou
t
of
t
he
te
n
s
kill
s
that
m
us
t
be
pro
vid
e
d.
I
ha
ve
a
gro
up
of
e
m
plo
ye
es
res
pons
i
ble
fo
r
c
arr
yi
ng
out
this
ta
sk
[6
]
.
Th
ese
groups
di
f
fer
in
the
num
ber
of
ta
sk
s
a
nd
the
nu
m
ber
of
em
plo
ye
es
in
ord
er
to
stu
dy
th
e
eff
ect
of
the
increase
a
nd
decr
ease
of
th
ese
nu
m
ber
s
on
the
resu
lt
in
g
so
l
utio
ns
a
nd
their
a
ccur
acy
.
T
her
e
fore,
these
gro
up
s
ha
ve
6
di
fferent
num
ber
s
for
the
pro
j
ect
ta
sk
s
wh
ic
h
a
re (1
6,
32, 64,
128, 25
6,
512) a
nd
6 d
iffer
e
nt
nu
m
ber
s
of
em
plo
ye
es (8, 1
6,
32, 6
4, 12
8,
256)
[26
]
.
Each
gro
up
of
data
represent
s
an
ind
e
pe
ndent
pro
j
ect
f
r
om
the
so
ftware
proj
ect
s
.
Th
eref
or
e
,
eac
h
gro
up
has
di
fferent
de
pe
nde
ncies
an
d
prec
eden
ts
betwee
n
it
s
ta
sk
s,
m
e
anin
g
that
eac
h
gro
up
has
a
diff
e
re
nt
schem
e
fr
om
t
he
ta
sk
em
erg
ence
sc
hem
es
(TPG).
Ta
ble
1
s
hows
t
he
si
m
il
ar
and
dif
fe
ren
t
c
har
act
e
risti
cs
of
t
he
gr
oups
w
hile t
he
T
a
ble 2 s
hows Ge
ner
al
descr
i
ption o
f e
ach
gro
up of
data.
Table
1.
T
he
sim
il
ar an
d dif
fe
ren
t c
har
a
ct
eris
ti
cs o
f
t
he data
set
s
Tr
aits
Descripti
o
n
The to
tal nu
m
b
er
of
gro
u
p
s
3
6
gro
u
p
s
The to
tal nu
m
b
er
of
sk
ills
1
0
diff
erent sk
ills
Nu
m
b
e
r
o
f
sk
ills f
o
r
each task
6
-
7
dif
f
erent sk
ills
The n
u
m
b
e
r
o
f
sk
ills p
er
e
m
p
lo
y
ee
6
-
7
dif
f
erent sk
ills
The to
tal nu
m
b
er
of
task
s
Equ
als o
n
e of
the follo
win
g
valu
es (1
6
,
3
2
,
6
4
,
1
2
8
,
2
5
6
,
5
1
2
)
The a
m
o
u
n
t of
ef
fort
Each task
has
its o
wn
r
an
d
o
m
vo
ltag
e va
lu
e
The to
tal nu
m
b
er
of
e
m
p
lo
y
ees
Equ
als o
n
e of
the follo
win
g
valu
es (8
,
1
6
,
3
2
,
6
4
,
1
2
8
,
2
5
6
)
The a
m
o
u
n
t of
the salary
Each e
m
p
lo
y
ee ha
s th
eir
o
wn
r
an
d
o
m
sala
ry v
alu
e
Task
app
eara
n
ce d
iag
ra
m
Dif
f
erent r
an
d
o
m
sch
e
m
e f
o
r
ea
ch
pr
o
ject
Table
2.
Ge
neral
descr
i
ptio
n of eac
h g
rou
p of data
Dataset
Gen
eral Des
criptio
n
16*8
co
n
sis
ts o
f
1
6
task
and
8 e
m
p
lo
y
ees
1
6
*
1
6
co
n
sis
ts o
f
1
6
task
and
16
e
m
p
lo
y
ees
1
6
*
3
2
co
n
sis
ts o
f
1
6
task
and
32
e
m
p
lo
y
ees
1
6
*
6
4
co
n
sis
ts o
f
1
6
task
and
64
e
m
p
lo
y
ees
1
6
*
1
2
8
co
n
sis
ts o
f
1
6
task
and
12
8
e
m
p
lo
y
ee
s
1
6
*
2
5
6
co
n
sis
ts o
f
1
6
task
and
25
6
e
m
p
lo
y
ee
s
32*8
co
n
sis
ts o
f
3
2
task
and
8 e
m
p
lo
y
ees
3
2
*
1
6
co
n
sis
ts o
f
3
2
task
and
16
e
m
p
lo
y
ees
3
2
*
3
2
co
n
sis
ts o
f
3
2
task
and
32
e
m
p
lo
y
ees
2
5
6
*
2
5
6
co
n
sis
ts o
f
2
5
6
tas
k
and
25
6
e
m
p
lo
y
e
es
Th
e
data c
ollec
ti
on
process
takes
place acc
ordin
g
to
the
pro
je
ct
infor
m
at
ion
and its sc
hedu
li
ng
m
echan
is
m
:
Pr
oject
ta
sks
(
Tasks
):
The
ta
sk
s
re
present
t
he
oper
at
ion
s
that
m
us
t
be
execu
te
d
in
a
s
equ
e
ntial
m
ann
er
,
ta
kin
g
into
ac
count
the
prec
eden
ce
of
thes
e
ta
s
ks
a
nd
t
he
de
pende
ncies
betwee
n
t
he
m
in
order
f
or
the
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
S
N
:
2502
-
4752
Ind
on
esi
a
n
J
E
le
c Eng &
Co
m
p
Sci,
Vo
l.
23
, N
o.
1
,
Ju
ly
2021
:
549
-
557
554
pro
j
ect
to
be
c
om
plete
d.
O
ne
ta
sk
is
known
as
(
t
j
)
an
d
j
repr
esents
the
val
ue
s
from
(1
)
to
(
T)
that
represe
nt
the
t
otal
num
ber
of
pro
j
ect
t
asks.
Eac
h
ta
s
k
possesse
s
a
set
of
sk
il
ls
know
n
a
s
(
t
j
ski
lls
)
w
hi
ch
m
us
t
be
avail
able
in
the
e
m
plo
ye
e
or
gro
up
of
em
pl
oyees
respo
ns
i
ble
for
carryin
g
out
this
ta
sk
,
and
al
so
poss
esses
the ef
fort re
qui
red to im
ple
m
e
nt it
(
E
ffor
t)
, w
hich
is
kn
own as (
t
j
e
ffort
)[19
]
.
Task
prece
dence
gr
a
ph
:
T
he
ta
sk
prece
de
nce
gr
a
ph,
know
n
as
(TPG)
is
use
d
to
re
pr
ese
nt
the
pr
ece
den
c
e
of
ta
sk
s
an
d
the
dep
e
ndencies
betwee
n
them
.
It
is
a
non
-
dir
ect
ed
pie
cha
rt
(G
(
V,
A))
c
onsist
ing
of
t
he
top
gro
up
(
(V
=
{t
1
,
t
2
,
..,
t
T
})
vert
ex
set
)
wh
ic
h
expresses
the
pro
j
ect
ta
sk
s
and
the
arc
set
(
(
A)
arc
set
)
w
hich
expresses
the
dep
e
ndency
be
tween
the
ta
sks
as
((
ti
,
tj)
ϵ
A)
m
eans
that
the
ta
sk
(t
i
,
t
j
)
It
cann
ot
be
gi
n
to
execu
te
unti
l afte
r
the
task
(
t
i
)
is com
plete
d
f
irst [
13
]
.
Em
plo
ye
es:
E
mp
loye
es
are
the
m
os
t
i
m
po
rt
ant
s
ource
for
so
ft
war
e
com
pa
nies
a
nd
they
are
the
only
s
ource
for
S
PSP
.
T
he
e
m
plo
ye
e
is
known
a
s
(ei)
a
nd
i
represe
nts
t
he
values
f
ro
m
(
1)
to
(E
)
that
represe
nt
the
total
nu
m
ber
of
em
plo
ye
es
w
orki
ng o
n
the
pr
oj
ec
t.
Each
em
plo
ye
e
possesse
s
a
set
of
s
kill
s
(
e
i
ski
lls
),
each
s
kill
is
known
as
(
s
i
)
an
d
i
re
pr
ese
nts
t
he
values
f
ro
m
(
1)
to
(S)
wh
i
ch
represe
nt
th
e
total
num
ber
of
s
kill
s
of
th
e
e
m
plo
ye
e,
w
hile
(
S
k
)
re
pr
ese
nt
s
the
num
ber
.
The
va
rio
us
s
kill
s
avail
able
to
the
em
plo
ye
es,
on
e
of
the
e
m
plo
ye
es
can
possess
tw
o
diff
e
ren
t
s
kil
ls
(
web
de
sig
n,
program
m
ing
in
j
ava
)
out
of
6
diff
e
re
nt
sk
il
ls
that
the
em
plo
ye
es
po
s
sess
(
desi
gnin
g
we
bs
it
es,
pro
gr
am
m
ing
in
j
a
va,
te
am
lead
ers
hip,
data
ba
se
m
anag
em
e
nt
[26],
pro
gram
m
ing
in
Ma
tl
ab
la
ngua
ge,
de
sign
us
in
g
unif
ie
d
m
od
el
ing
la
ngua
ge
(
UM
L
)
)
,
an
d
the
s
ki
ll
s
of
a
sin
gle
em
ploy
ee
are
co
ns
id
ered
a
subset
of
th
e
overall
s
ki
ll
set
(
∈
S
k
e
i
skil
l
s
).
T
he
e
m
plo
ye
e
ha
s
his
own
sal
ary,
al
so
known
as
(
e
i
sa
la
ry
)
a
nd
is
re
pr
e
sente
d
by
a
unit
of
m
on
ey
,
as
well
as
the
value
of
t
he
m
axi
m
u
m
tim
e
sp
ent
the
e
m
plo
ye
e
to
w
ork
in
the
pro
je
ct
(
m
axi
m
u
m
de
dicat
ion
),
w
hich
is
know
n
as
(
e
i
ma
x
d
e
d
)
an
d
is represe
nted by a
un
it
of ti
m
e, an
d t
his
va
lue
is ass
um
ed
to b
e
equal t
o on
e
[
24]
.
The
so
l
ution
to
t
he
s
of
t
war
e
pro
j
ect
sc
he
duli
ng
prob
le
m
is
a
m
at
rix
X
of
siz
e
(
×
)
w
her
e
(
X
=
x
ij
)
the
c
olu
m
n
r
epr
ese
nts
th
e
pro
j
ect
ta
sks
ar
r
ang
e
d
acc
ordi
ng
to
prece
de
nc
e
of
im
ple
m
entat
ion
a
nd
the
r
ow
represe
nts
the
e
m
plo
ye
es
wor
king
on
t
he
pro
j
ect
a
nd
t
he
(
x
ij
)
values
re
prese
nt
a
pe
rce
ntage
of
th
e
up
per
li
m
it
of
the
ti
m
e
tha
t
t
he
e
m
plo
ye
e
sp
en
ds
it
in
c
arr
yi
ng
out
the
ta
sk
assig
ned
to
him
,
if
(
x
ij
=
0
)
t
he
n
this
m
eans
that
the
em
ploy
ee
(
ei
)
is
not
re
sp
onsi
ble
f
or
the
ta
s
k
(
tj
),
an
d
x
ij
if
(
=1
)
t
hen
this
m
eans
that
the
em
plo
ye
e
(
ei
)
devotes
al
l
his
tim
e
to
the
ta
sk
(
tj
),
bu
t
if
it
is
(
0
<
x
ij
<
1
)
the
n
this
m
eans
that
the
em
plo
ye
e
(
ei
)
al
locat
es
(
x
ij
)
of
his tim
e to car
r
y ou
t t
he
ta
s
k
(
tj
)
[
27]
. T
he
a
bove
ca
n be s
how
n
in
Ta
ble 3.
Table
3.
A
s
um
m
ary of
the
issue of
sc
hedul
ing
t
he
s
of
t
ware p
roject
Descripti
o
n
Sy
m
b
o
l
The to
tal nu
m
b
er
of
pro
ject task
s
T
Task
nu
m
b
e
r
j
in
t
h
e pro
ject
t
j
The sk
ill
set f
o
r
tas
k
t
j
skills
The ef
f
o
rt
requ
ired to
i
m
p
le
m
en
t the t
ask
t
j
ef
f
or
t
Task
app
eara
n
ce d
iag
ra
m
TGP
Su
m
m
it
gro
u
p
,
=
{
,
,
.
.
.
,
}
V
The b
racket set
,
(
,
)
∈
w
h
ere
task
p
rece
d
es
task
b
y
execu
tio
n
A
The to
tal nu
m
b
er
of
e
m
p
lo
y
ees in
the
p
roject
E
E
m
p
lo
y
e
e nu
m
b
er
in
the p
roject
e
i
E
m
p
lo
y
e
e sala
ry
e
i
sala
r
y
E
m
p
lo
y
e
e Skills
et
e
i
skills
The to
tal nu
m
b
er
of
diff
erent sk
ills av
ailab
le in th
e pro
je
ct
SK
The
m
ax
i
m
u
m
ti
m
e that an e
m
p
lo
y
ee
ded
icates (
)
to
wo
rk o
n
the p
roject
e
i
max
d
ed
Matr
ix
so
lu
tio
n
by
vo
lu
m
e
(
)
X
Prop
o
rtion
of
work
do
n
e by
e
m
p
lo
y
ee
to
perf
o
r
m
task
x
ij
5.
RESU
LT
S
AND DI
SCUS
S
ION
The
two
al
gorit
hm
s
us
ed
i
n
th
is
m
essage
we
r
e
i
m
ple
m
ented
us
in
g
(Mat
la
b
R201
7a)
la
ng
uag
e
an
d
on
a
la
ptop
t
hat
i
s
r
unning
on
(
W
i
ndows
10
P
ro)
syst
em
with
(Core
(TM)
i5
-
4200U
CPU
160G
Hz
-
2.3
0GHz
)
and
has
an
e
xt
ern
al
m
e
m
or
y
siz
e
(4
GB)
.
A
s
f
or
t
he
va
riab
le
s
us
e
d
by
both
al
gorithm
s,
they
can
be
s
ho
wn
in
Table
4.
Evaluation Warning : The document was created with Spire.PDF for Python.
Ind
on
esi
a
n
J
E
le
c Eng &
Co
m
p
Sci
IS
S
N:
25
02
-
4752
Ap
plicati
on pr
oject
task sc
he
du
li
ng
us
i
ng dolp
hin
swar
m
t
echnolo
gy
(
B
araa S
. M
os
t
afa)
555
Table
4.
T
he
val
ues of
t
he var
ia
bles of the
DSA al
gorithm
s
Variables
DSA
Beh
av
io
rs
The action
s o
f
pred
atio
n
,
swar
m
in
g
,
catchin
g
,
an
d
m
o
v
e
m
en
t
Distan
ce of
vis
io
n
1
.5
Step
0
.01
The n
u
m
b
e
r
o
f
do
lp
h
in
es
in
the swarm
20
The size of
th
e log
of
un
co
n
trolled
s
o
lu
tio
n
s
20
Extern
al log
size
100
The n
u
m
b
e
r
o
f
ses
sio
n
s
1000
Nu
m
b
e
r
o
f
ti
m
es
of
i
m
p
le
m
en
tatio
n
s
10
Af
te
r
usi
ng the
D
S
A
al
gorith
m
in so
l
ving th
e SPSP
prob
le
m
, th
e fo
ll
owin
g resu
lt
s
wer
e
ob
ta
ine
d:
Exec
ution
ti
m
e
r
esults:
Tab
le
5
show
s the average
e
xecu
ti
on tim
e p
er s
econ
d of
the alg
ori
th
m
fo
r
al
l data
set
s.
The
exe
cution
ti
m
e
r
esults
ob
ta
i
ne
d
us
i
ng
the
do
l
ph
i
n
al
gori
thm
wer
e
co
m
par
ed
with
the
so
urce
[
28
]
t
ha
t
us
e
d
the
A
r
ti
fici
al
fish
s
w
arm
al
go
rithm
-
one
be
hav
i
or
(A
F
SA
-
1b)
a
nd
a
rt
ific
ia
l
fish
swar
m
al
go
rith
m
based
on
glo
bal
(
G
AF
S
A
)
in
the
sche
duli
ng
process.
T
he
F
ig
ur
e
2
s
hows
t
he
A
vera
ge
execu
ti
on ti
m
e
in
sec
ond bet
ween
the
DSA
al
gorithm
an
d
t
he GA
FS
A
,
A
FSA
-
1b al
gorithm
s.
Table
5.
A
ver
a
ge
e
xecu
ti
on ti
m
e in secon
d
Data
DSA
AFSA
-
1b
GAFSA
Data
DSA
AFSA
-
1b
GAFSA
16*8
1
.53
7
1
.74
3
6
.13
3
1
2
8
*
8
6
.77
0
8
.47
9
3
7
.40
7
1
6
*
1
6
1
.62
1
1
.79
1
6
.81
2
1
2
8
*
1
6
6
.00
5
7
.15
0
2
6
.01
8
1
6
*
3
2
1
.72
8
2
.42
7
1014
1
2
8
*
3
2
9
.87
3
1
1
.53
5
4
0
.68
3
1
6
*
6
4
2
.00
4
3
.17
9
1
3
.36
4
1
2
8
*
6
4
1
9
.89
4
2
2
.23
4
8
1
.68
2
1
6
*
1
2
8
3
.50
2
5
.21
1
8
.80
5
1
2
8
*
1
2
8
2
8
.68
9
4
3
.80
9
1
7
3
.293
1
6
*
2
5
6
5
.88
6
9
.67
8
3
2
.38
3
1
2
8
*
2
5
6
5
5
.98
4
8
8
.98
6
3
0
6
.882
32*8
1
.68
0
2
.61
6
1
0
.14
8
2
5
6
*
8
9
.50
9
1
5
.91
6
6
7
.59
5
6
3
2
*
1
6
1
.85
3
2
.96
3
1
1
.66
0
2
5
6
*
1
6
1
1
.89
4
2
2
.94
3
1
0
0
.193
3
2
*
3
2
2
.62
7
4
.19
6
1
7
.51
4
2
5
6
*
3
2
1
8
.55
2
3
2
.5
29
1
3
6
.628
3
2
*
6
4
3
.55
6
5
.69
8
2
4
.13
3
2
5
6
*
6
4
4
3
.47
2
5
6
.92
2
2
0
9
.057
3
2
*
1
2
8
5
.56
4
1
0
.69
4
4
3
.89
6
2
5
6
*
1
2
8
6
8
.37
0
9
7
.27
1
3
8
0
.603
3
2
*
2
5
6
1
0
.48
3
1
9
.68
0
7
0
.60
8
2
5
6
*
2
5
6
1
0
5
.122
1
9
4
.133
7
5
2
.299
64*8
2
.10
8
4
.29
3
1
8
.47
0
5
1
2
*
8
2
5
.53
4
4
.39
9
2
1
1
.408
6
4
*
1
6
4
.56
1
5.
041
2
0
.56
3
5
1
2
*
1
6
4
1
.62
1
6
5
.93
8
3
1
3
.128
6
4
*
3
2
5
.59
6
7
.97
6
3
6
.10
8
5
1
2
*
3
2
5
6
.72
8
7
1
.86
1
3
1
5
.462
6
4
*
6
4
9
.76
7
1
1
.77
7
3
8
.48
4
5
1
2
*
6
4
1
0
2
.004
1
2
7
.558
5
4
3
.767
6
4
*
1
2
8
1
8
.77
3
2
1
.33
7
6
1
.43
5
5
1
2
*
1
2
8
1
6
9
.502
2
0
5
.791
7
5
1
.681
6
4
*
2
5
6
3
6
.82
3
4
0
.78
3
1
7
5
.746
5
1
2
*
2
5
6
2
4
4
.886
4
1
3
.649
1
5
0
3
.5
7
Dive
rsity
of
r
esults:
The
di
ver
sit
y
of
res
ults
is
sti
l
l
th
e
m
os
t
i
m
po
rtant
m
easur
e
of
pe
rfo
rm
ance
evaluati
on.
T
he
refor
e
,
the
D
SA
al
gorithm
was
able
to
s
olv
e
the
SPSP
issue
in
rec
ord
tim
e.
Table
6
sh
ows
the
m
ean
H
V
valu
e
of
DS
A
a
lg
ori
th
m
,
so
luti
on
s
w
it
h
an
H
V
inde
x
val
ue
wer
e
sh
ow
n
f
or
m
os
t
dataset
s.
H
V
value
was
c
om
par
ed
with
Re
so
urce
[
13]
wh
ic
h
co
ntains
8
al
gorith
m
s
us
ed
in
t
he
sche
du
li
ng
pro
cess, as
w
el
l as
w
it
h AF
SA
-
1b,
GAFS
A
[
28
]
shown i
n
T
a
bl
e 7
.
Table
6.
T
he
m
ean
value o
f
th
e oversize i
nde
x
Data
DSA algo
rith
m
Data
DSA algo
rith
m
Data
DSA algo
rith
m
Data
DSA algo
rith
m
16*8
0
.26
6
3
2
*
6
4
0
.16
9
1
2
8
*
8
0
.13
7
2
5
6
*
6
4
0
.07
5
1
6
*
1
6
0
.27
0
3
2
*
1
2
8
0
.29
2
1
2
8
*
1
6
0
.16
3
2
5
6
*
1
2
8
0
.06
5
1
6
*
3
2
0
.36
4
3
2
*
2
5
6
0
.17
5
1
2
8
*
3
2
0
.07
1
2
5
6
*
2
5
6
0
.12
4
1
6
*
6
4
0
.19
5
64*8
0
.14
2
1
2
8
*
6
4
0
.17
1
5
1
2
*
8
0
.02
5
1
6
*
1
2
8
0
.22
7
6
4
*
1
6
0
.16
2
1
2
8
*
1
2
8
0
.09
3
5
1
2
*
1
6
0
.04
8
1
6
*
2
5
6
0
.19
8
6
4
*
3
2
0
.14
8
1
2
8
*
2
5
6
0
.13
9
5
1
2
*
3
2
0
.03
9
32*8
0
.13
1
6
4
*
6
4
0
.18
1
2
5
6
*
8
0
.09
8
5
1
2
*
6
4
0
.04
8
3
2
*
1
6
0
.28
7
6
4
*
1
2
8
0
.10
6
2
5
6
*
1
6
0
.13
2
5
1
2
*
1
2
8
0
.05
7
32*3
2
0
.85
5
6
4
*
2
5
6
0
.15
8
2
5
6
*
3
2
0
.06
7
5
1
2
*
2
5
6
0
.04
5
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
S
N
:
2502
-
4752
Ind
on
esi
a
n
J
E
le
c Eng &
Co
m
p
Sci,
Vo
l.
23
, N
o.
1
,
Ju
ly
2021
:
549
-
557
556
Table
7.
T
he
Com
par
ison wit
h value
H
V
Ins
tan
ce
NSGA
-
||
SPEA2
PAES
DEPT
MO
-
FA
MOAB
C
MOCell
GDE3
AFS
A
-
1b
GAFSA
DSA
i6
4
-
8
0
.46
9
0
.3
45
0
.8
11
0
.0
86
0
.1
45
0
.0
25
0
.5
26
0
.5
28
0
.07
85
0
.0
42
0
.1
42
i6
4
-
16
0
.21
9
0.
1
54
0
.9
61
0
.0
24
0
.3
41
0
.1
73
0
.2
64
0
.3
22
0
.07
25
0
.0
92
0
.1
62
i6
4
-
32
0
.06
5
0
.0
31
0
.7
98
0
0
.3
12
0
.0
11
0
.0
59
0
.2
39
0
.09
2
0
.0
485
0
.1
48
i6
4
-
64
0
.07
4
0
.0
46
0
.8
72
0
0
.0
29
0
0
.0
81
0
.2
58
0
.10
35
0
.0
81
0
.1
81
i6
4
-
128
0
.02
6
0
.0
12
0
.7
37
0
.0
05
0
0
0
.0
23
0
.1
57
0
.09
75
0
.0
66
0
.1
06
i6
4
-
256
0
0
0
.6
19
0
0
0
0
.0
07
0
.1
5
0
.12
15
0
.1
185
0
.1
58
6.
CONCL
US
I
O
N
Wh
e
n
previe
w
ing
the
previ
ous
res
ults
,
the
fo
ll
owin
g
ca
n
be
obser
ve
d;
t
he
dolp
hin
al
gorithm
was
able
to
fi
nd
res
ults
in
t
he
da
ta
set
(32
*
25
6)
duri
ng
t
he
e
xec
ution
tim
e
(10.483
)
se
c
onds
,
wh
ic
h
is
a
rela
ti
vely
sh
ort
ti
m
e.
Its
new
f
or
m
at
is
m
or
e
su
it
able
for
us
e
i
n
c
omm
ercial
app
li
c
at
ion
s
a
nd
s
of
t
war
e
to
ols
co
m
par
ed
with
the
two
a
lgorit
hm
s
(A
FSA
-
1b
G
AF
S
A)
that
took
(
19.
680,
70.60
8)
resp
e
ct
ively
with
the
sam
e
data
set
.
The
do
l
ph
i
n
al
go
rithm
pr
ove
d
that
it
is
able
to
fi
nd
va
rio
us
so
luti
on
s
w
he
re
the
m
ean
va
lues
of
the
H
V
in
de
x
in
the
dataset
(64
*64)
is
0.1
81.
It
ga
ve
go
od
di
ver
sit
y
of
so
l
ution
s
com
par
e
d
t
o
the
al
gorithm
s
us
ed
in
the
sche
du
li
ng
pro
cess.
REFERE
NCE
S
[1]
Press
m
an,
R.
S.,
“
Software
engi
n
ee
ring
:
a
pra
ctitioner'
s
appr
oa
ch,”
Mc
Gr
aw
Hi
ll
,
2010
.
[2]
Am
iri
,
M.
and
Barbi
n,
J.P
.
,
“
New
appr
oa
ch
fo
r
solving
softwa
re
project
sch
ed
uli
ng
probl
em
using
diffe
ren
ti
a
l
evol
uti
on
a
lgori
t
hm
,
”
Int
er
nati
on
al
Journal
in
Fo
undati
ons
of
Computer
Scienc
e
&
Technol
ogy
(
IJF
CST
)
,
vol.
5,
no.
1
,
pp
.
1
-
9,
20
15,
doi
:
10
.
5121
/i
jfc
st
.
2015.
5101
.
[3]
Aziz
i
,
R.
,
“
Empiric
a
l
stud
y
of
ar
t
ifi
cial
f
ish
sw
arm
al
gorit
hm
,
”
ar
Xiv
pre
print
arXiv
:1405.
4138
,
v
ol.
3,
n
o
.
1,
pp.
1
-
7,
2014
.
[4]
Sebt,
M.
H.,
Af
shar,
M.
R
.
,
an
d
Alipouri
,
Y.
J
.
E
.
O.
,
“
H
y
bri
diz
a
ti
on
of
g
enetic
al
gori
thm
an
d
fully
informed
par
ticle
sw
arm
for
solving
th
e
m
ult
i
-
m
ode
r
esourc
e
-
constr
ained
project
s
c
hedul
ing
prob
lem
,
”
Engi
n
ee
rin
g
Optimizati
on
,
vo
l.
49
,
no
.
3
,
pp
.
5
13
-
530
,
2017
,
d
oi:
10
.
1080/030
5215X.2016.
119
7610
.
[5]
Xiao,
J.,
Gao
,
M.
L.,
and
Huang,
M.
M,
“
Empiric
al
Stud
y
of
Mult
i
-
objecti
v
e
Ant
Colon
y
Optimiz
at
ion
to
Softwar
e
Projec
t
Schedu
l
ing
Problem
s,
”
In
Proce
edi
n
gs
of
the
Ann
ual
Confe
renc
e
on
Gene
ti
c
a
nd
Ev
olutionar
y
Computati
on
,
20
15
,
pp
.
759
-
766
,
doi:
10.
1145
/27
39480.
2754702
.
[6]
Koulina
s,
G.
,
Kotsikas,
L
.
,
a
nd
Anagnostopoulos,
K.,
“
A
par
ticle
sw
arm
opti
m
iz
ation
ba
sed
h
y
p
er
-
heur
i
stic
al
gorit
hm
for
th
e
cl
assic
r
esourc
e constra
in
ed
proj
ec
t
sch
edul
ing
p
ro
bl
em,”
Inform
ati
on
Scienc
es
, v
ol
.
277,
pp
.
680
-
693,
2014
,
doi
:
1
0.
1016/j.ins.
201
4.
02.
155
.
[7]
Neshat
,
M.
,
Sepi
dnam,
G.,
Sargo
lz
a
ei
,
M.,
and
T
oosi,
A.
N,
“
Arti
fic
i
al
fish
sw
arm
al
gori
thm:
a
sur
ve
y
of
th
e
state
-
of
-
the
-
ar
t,
h
y
bri
diz
a
ti
on,
combi
nat
ori
al
and
indi
ca
t
ive
applicatio
ns
,”
Arti
ficia
l
Inte
lligen
ce
R
e
vie
w
,
vol.
42,
n
o.
4
,
pp.
965
-
997
,
20
14
,
doi
:
10
.
1007
/s10462
-
012
-
9342
-
2
.
[8]
Crawford,
B.
,
S
oto,
R.
,
Johns
on,
F.,
Monfro
y
,
E
.
and
Pare
des
,
F.,
“
A
m
ax
–
m
in
ant
s
y
st
em
al
gor
it
hm
to
solve
the
software
project
sche
duli
ng
prob
l
em
,
”
Ex
p
ert
Syst
ems
wit
h
Appl
i
c
ati
ons
,
v
ol
.
41,
n
o.
15,
pp.
6634
-
6645,
2014
,
doi
:
10.
1016/j.e
sw
a.
2
014.
05.
003
.
[9]
Alla
n
Vinic
ius
Rez
end
e,
L
ei
l
a
Silva
,
André
Brit
to,
Rodrig
oAm
ara
l,
"S
oftwa
re
Project
Sche
duli
ng
Problem
in
th
e
Conte
xt
of
Sear
c
h
-
B
ase
d
Software
Engi
ne
eri
ng
:
A
Sy
stema
ti
c
R
e
vie
w
",
Journal
o
f
Syste
ms
and
Soft
ware,
Vol
155
,
Pages
43
-
56,
20
19.
[10]
Bij
u,
A.C
.
,
Vi
ctoire
,
T.
and
Mohana
sundara
m
,
K.
,
“
An i
m
prove
d
diffe
ren
ti
a
l
evolution
soluti
on
fo
r
software
projec
t
sche
duli
ng
probl
em
,”
The
Scient
i
fi
c
World
Journal
,
vo
l, pp.
1
-
9
,
2
015
,
doi
:
10
.
115
5/2015/
232193
.
[11]
W
u,
X.,
Consoli,
P.,
Minku,
L.,
O
choa
,
G.
and
Yao,
X
.
,
“
An
evo
lut
iona
r
y
h
y
per
-
heur
isti
c
for
th
e
software
proj
ect
sche
duli
ng
p
rob
le
m
,
”
In
te
rnatio
nal
Confe
ren
ce
on
Parall
e
l
Pro
ble
m
So
l
vi
ng
from
Nature
,
pp
.
3
7
-
47,
2016
,
doi:
10.
1007/978
-
3
-
3
19
-
45823
-
6_4
.
[12]
Crawford,
B.
,
S
oto,
R.
,
Johns
on,
F.,
Val
encia
,
C.
an
d
Pare
d
es,
F.
,
“
Firefly
Algorithm
to
Solve
a
Pr
oje
c
t
Scheduling
Problem
,
”
Art
if
i
ci
al
I
ntelli
g
ence
Pe
rs
pectives
in
Inte
l
li
g
ent
S
yst
ems
,
pp.
449
-
45
8
,
2016
,
doi:
0.
1007/978
-
3
-
319
-
33625
-
1_40
.
[13]
Luna
,
F.,
Gonz
ál
e
z
-
Álvar
ez
,
D.
L.
,
Chi
ca
no,
F.
a
nd
Veg
a
-
Rod
rígue
z
,
M.A.
,
“
The
softwar
e
pr
oje
c
t
sche
du
li
ng
proble
m
:
A
sca
l
abi
lit
y
anal
y
s
is
of
m
ult
i
-
objective
m
et
ahe
ur
isti
c
s
,
”
Applied
Soft
Computing
,
vol
.
15,
pp.
136
-
14
8
,
2014
,
doi
:
10
.
10
16/j
.
asoc
.
2013
.
1
0.
015
.
[14]
W
u,
T.
Q.,
Ya
o,
M.,
and
Yan
g,
J.
H.
,
“
Dolphin
sw
arm
al
gor
it
hm
,
”
Front
ie
r
s
of
Information
Technol
ogy
&
El
e
ct
ronic
Engi
n
ee
ring,
vol
.
17
,
no.
8,
pp.
717
-
72
9,
Aug.
2016
,
do
i:
10
.
1631/FITEE.
1500287
.
Evaluation Warning : The document was created with Spire.PDF for Python.
Ind
on
esi
a
n
J
E
le
c Eng &
Co
m
p
Sci
IS
S
N:
25
02
-
4752
Ap
plicati
on pr
oject
task sc
he
du
li
ng
us
i
ng dolp
hin
swar
m
t
echnolo
gy
(
B
araa S
. M
os
t
afa)
557
[15]
F.
Habibi,
F.
B
arz
inpour
,
S.
J.
Sadja
di
.,
“
Resourc
e
-
constr
ai
n
ed
project
s
che
dul
ing
probl
em:
re
vie
w
of
p
ast
an
d
rec
e
n
t
dev
el
op
m
ent
s
,
”
Journ
al
of
Proj
ec
t
Manage
ment
,
vol.
3
,
no
.
2,
pp
.
55
–
8
8,
2018
,
do
i:
10.
5267/j.jpm.2
018.
1.
005
.
[16]
Crawford,
B.
,
Soto,
R.
,
Johns
on,
F.,
Misra,
S.
and
Pare
des,
F.,
“
The
use
of
me
ta
h
eur
isti
cs
to
software
projec
t
sche
duli
ng
probl
em
,
”
In
Inte
r
nat
ional
Confe
ren
c
e
on
Computati
onal
Sci
en
ce
and
Its
Appl
ic
a
ti
ons
,
2014
,
pp.
215
-
226
,
doi
:
10
.
100
7/978
-
3
-
319
-
091
56
-
3_16
.
[17]
Am
it
Ku
m
ar,
Sanjay
Sing
la,
Me
ena
kshi
Sharm
a,
and
Shakti
Kundu.
,
“
Moth
-
Dolphin
Optimiza
tio
n
Algorit
hm
:
A
Natur
e
Inspire
d
Te
chn
ique
,
”
Inter
nati
onal
Journal
of
Innov
ative
Technol
ogy
and
Ex
ploring
Eng
in
ee
ring
(
IJI
TEE)
,
v
ol
.
8
,
no
.
12
,
20
19
,
doi
:
10
.
3594
0/i
jitee.L
3014
.
1
081219
.
[18]
Bini
tha,
S.
and
Sath
y
a,
S.S.,
“
A
surve
y
o
f
bio
inspire
d
op
ti
m
i
za
t
ion
a
lgori
thm
s
,
”
Int
ernati
onal
Journal
of
S
of
t
Computing
and
Engi
ne
ering
,
v
ol
.
2
,
no
.
2
,
pp
.
137
-
151,
2012
.
[19]
Minku,
L.
L
.
,
Su
dholt
,
D.
and
Ya
o,
X.,
“
Im
prove
d
evol
uti
ona
r
y
algorithm
design
for
the
proje
c
t
sche
duli
ng
probl
em
base
d
on
runti
m
e
ana
l
y
sis
,
”
IEEE
Tr
ansacti
ons
on
Soft
ware
Engi
nee
ring
,
vol
.
40,
no.
1,
pp.
83
-
102,
2014
doi:
10.
1109/T
SE
.
20
13.
52
.
[20]
F
adhi
l,
A.
A
.
,
Alsarra
j,
R
.
G.
,
and
Altaie,
A
.
M
,
“
Software
c
ost
esti
m
at
ion
b
ase
d
on
dolphi
n
al
gorit
hm
,
”
IE
EE
Ac
c
ess
,
vol
.
8,
pp
75279
-
75287
,
2
020
,
doi
:
10
.
110
9/ACCESS
.
2020.
2988867
.
[21]
W
.
Qiao
a
nd
Z.
Yang,
“
Modifie
d
dolphi
n
sw
arm
al
gorit
hm
bas
ed
on
cha
otic
m
aps
for
solving
high
-
dimensiona
l
func
ti
on
opti
m
izati
on
proble
m
s,”
IEEE Access
,
v
ol.
7
,
2019
,
do
i:
10.
1109/ACCESS
.
2019.
2931910
.
[22]
K.
Le
nin
,
B.
R
.
Redd
y
,
and
M.
S.
Kala
va
thi
,
“
Dolphin
ec
ho
loc
ation
al
gori
thm
for
solving
opti
m
a
l
re
a
ct
iv
e
power
dispat
ch
probl
e
m
,
”
Inte
rnation
al
Journal
of
Computer
(
IJ
C)
,
vol.
12,
no.
1,
pp.
1
–
15,
2014
,
doi
:
10.
11648/j.ijepe
.
20140301.
11
.
[23]
Qiao,
W
.
,
and
Yang,
Z
.
,
“
An
Im
prove
d
Dolphin
Sw
arm
Algorit
hm
Based
o
n
Ker
nel
Fuzz
y
C
-
Mea
ns
in
t
h
e
Applic
a
ti
on
of
Solving
the
Opt
i
m
al
Problems
of
La
rge
-
Sc
al
e
Fu
nct
ion
,
”
IE
EE
A
cc
ess
,
vol
.
8,
pp
.
2073
-
2089
,
202
0
,
doi:
10
.
1109/AC
CESS
.
2019.
295
8456
.
[24]
W
.
Yong,
W
.
Ta
o,
Z
.
Cheng
-
Z
hi,
and
H.
Hua
-
Juan,
“
A
new
s
toc
hastic
opti
m
i
za
t
ion
a
pproa
ch
-
Dolphin
sw
arm
opti
m
iz
ation
a
lg
orit
hm
,
”
Int
erna
ti
onal
Journal
of
Computati
on
al
Intelli
g
ence
and
Appl
i
cat
ion
s
,
vol.
15
,
no.
2,
2016
,
doi
:
10
.
11
42/S1469026816500115
.
[25]
Al
Khati
b,
S.M.
and
Noppen,
J.,
“
Benc
hm
ark
ing
and
Com
par
ison
of
Software
Pro
je
c
t
Hum
an
Res
ourc
e
Allocati
on
Optimiza
ti
o
n
A
pproa
che
s,
”
AC
M
SIGSO
FT
S
oft
ware
Engi
n
e
ering
Note
s
,
vo
l.
41,
n
o.
6
,
2
017
,
pp.
1
-
6.
do
i:
10.
1145/301128
6.
3011305
.
[26
]
Ferruc
ci
F.
,
Harm
an
M.
and
Sarro
F.
,
“
Sear
ch
-
Based
Sof
twar
e
Proje
ct
Mana
gement
,”
Soft
ware
Project
Manage
ment in a Changi
ng
Wor
ld
,
pp
.
373
-
399
,
2014
,
doi
:
10
.
10
07%2F978
-
3
-
642
-
55035
-
5_15
.
[2
7
]
Alba,
E.
and
C
hic
ano
,
J.F.
,
“
Software
pro
ject
m
ana
gement
wit
h
Gas
,”
In
formation
Sc
ie
n
ce
s
,
vol.
177
,
n
o
.
11
,
pp.
2380
-
2401,
2
007
,
doi
:
10
.
101
6/j
.
ins
.
2006.
12
.
0
20
.
[2
8
]
Sara
h
E.
Alm
shhada
n
y
,
La
h
eeb
Ibra
him
,
“
U
sing
m
ult
i
-
objective
artifi
ci
a
l
fis
h
sw
arm
al
gorithm
to
solve
t
h
e
software
projec
t
sche
duli
ng
pr
oble
m
,
”
Master
's
the
sis
,
in
Software
Eng
ine
e
r
ing
,
Col
l
age
of
computer
of
m
at
hemati
cs
,
Un
ive
rsit
y
Of
Mos
ul,
Ir
aq,
2019
.
Evaluation Warning : The document was created with Spire.PDF for Python.