Indonesi
an
Journa
l
of El
ect
ri
cal Engineer
ing
an
d
Comp
ut
er
Scie
nce
Vo
l.
13
,
No.
1
,
Jan
uar
y
201
9
,
pp.
162
~
169
IS
S
N: 25
02
-
4752, DO
I: 10
.11
591/ijeecs
.v1
3
.i
1
.pp
162
-
169
162
Journ
al h
om
e
page
:
http:
//
ia
es
core.c
om/j
ourn
als/i
ndex.
ph
p/ij
eecs
Lea
d
ers and
fo
ll
owe
rs al
gorithm
f
or const
ra
in
ed non
-
lin
ear
optimiz
ation
Hele
n Yuli
ana
A
n
gm
alisan
g
,
Syaiful
A
n
am
,
S
ob
ri
Ab
usi
ni
Brawij
a
y
a
Univ
e
rsit
y
,
Jl
.
Ve
te
r
an
,
Mal
ang
65145
,
Ea
st Ja
v
a
,
Indon
esia
Art
ic
le
In
f
o
ABSTR
A
CT
Art
ic
le
history:
Re
cei
ved
J
ul
11
, 2
01
8
Re
vised
N
ov
18
, 2
01
8
Accepte
d
Nov
30
, 201
8
Le
ad
ers
and
Followers
al
gorit
h
m
was
a
novel
m
et
ahe
urist
ic
s
proposed
b
y
Yass
er
Gonza
lez
-
Ferna
nde
z
an
d
Stephe
n
Che
n.
In
solving
u
nconstra
in
e
d
opti
m
iz
ation,
it
per
form
ed
be
tt
er
expl
or
at
ion
tha
n
o
the
r
well
-
know
n
m
et
ahe
urist
ic
s,
e.
g.
Gene
ti
c
Al
gorit
hm
,
Part
ic
l
e
Sw
arm
Optimiza
ti
on
an
d
Diffe
ren
t
ia
l
Ev
olut
ion.
The
r
ef
ore
,
i
t
per
for
m
ed
well
in
m
ult
i
-
m
odal
proble
m
s.
In
thi
s
pape
r,
L
ea
de
rs
and
Followers
was
m
odifi
ed
for
constra
in
ed
non
-
li
ne
ar
optim
iz
at
ion
.
Sever
al
w
ell
-
known
benc
hm
ark
proble
m
s
for
constra
in
ed
opt
i
m
iz
at
ion
were
used
to
eva
lu
ate
th
e
proposed
al
gori
thm.
The
r
esult
of
the
eva
lu
at
ion
show
ed
tha
t
the
pro
posed
al
gori
thm
consiste
n
t
l
y
and
succ
essfull
y
found
th
e
opt
i
m
al
soluti
on
of
low
dimensional
constraine
d
opti
m
iz
ation
pro
ble
m
s
and
high
dimensional
op
tim
iz
at
ion
with
h
i
gh
num
ber
of
li
ne
ar
ine
qu
a
li
t
y
constr
ai
n
t
o
nl
y
.
Moreove
r
,
t
he
proposed
al
g
orit
hm
had
diffi
cu
lty
in
sol
ving
high
d
imensional
opt
imiza
t
i
on
proble
m
with
non
-
li
ne
ar
constra
in
ts
and
an
y
proble
m
wh
i
ch
has
m
ore
th
an
one
equa
l
ity
constra
in
t.
In
the
compari
son
with
othe
r
m
et
ahe
urist
ic
s,
Le
ad
ers
and
Followers
had
bet
t
er
pe
rform
anc
e in
ov
era
l
l
b
en
chmark
probl
ems
.
Ke
yw
or
d
s
:
Con
st
raint
Leader
s a
nd fo
ll
ow
ers
Me
ta
heu
risti
cs
Non
-
li
near
opti
m
iz
at
ion
Copy
right
©
201
9
Instit
ut
e
o
f Ad
vanc
ed
Engi
n
ee
r
ing
and
S
cienc
e
.
Al
l
rights re
serv
ed
.
Corres
pond
in
g
Aut
h
or
:
Syai
fu
l
An
am
,
Brawijaya
U
niv
ersit
y,
Jl. V
et
era
n, M
al
a
ng
6514
5,
E
ast
Jav
a,
In
done
sia
.
Em
a
il
:
sya
iful
@ub.ac
.id
1.
INTROD
U
CTION
Nowa
days,
op
tim
iz
at
ion
play
s
an
i
m
po
rtant
ro
le
in
va
riou
s
fiel
ds
of
r
eal
-
w
or
ld
,
e.g.
eng
inee
rin
g,
fina
nce,
tra
nsp
or
ta
ti
on
an
d
operati
onal
rese
arch
[1
]
.
T
here
are
m
any
kin
ds
of
optim
iz
ati
on
pro
blem
s.
On
e
of
them
is
con
strai
ned
no
n
-
li
near
opti
m
iz
a
ti
on
.
A
n
opti
m
iz
at
ion
pr
ob
lem
is
c
la
ssif
ie
d
as
con
str
ai
ned
op
ti
m
iz
ation
if
the
ob
j
ect
ive
f
un
ct
io
n
is
m
ini
m
iz
ed
or
m
axi
m
ized
un
der
give
n
co
ns
trai
nts
[2
]
.
Con
st
raine
d
non
-
li
near
opti
m
iz
at
ion
is
de
fine
d
as
a
c
on
strai
ned
opti
m
iz
at
ion
pr
ob
le
m
wh
ere
it
s
obj
ect
iv
e
functi
on
or
at
le
ast
on
e
of
th
e
con
strai
nts
is
non
-
li
nea
r
functi
on
[
3].
I
n
real
li
fe,
con
st
raine
d
op
ti
m
izati
on
pro
blem
m
ay
b
e f
ound
ver
y
oft
en
beca
us
e
m
a
ny r
e
quired
r
es
ources a
re
not
al
ways unli
m
ited.
Me
ta
heu
risti
cs
hav
e
be
en
widely
i
m
ple
m
en
te
d
for
so
l
ving
m
any
kin
ds
of
optim
iz
ation
pro
blem
s,
includi
ng
co
nst
rained
no
n
-
l
inear
op
ti
m
iz
a
ti
on
.
In
so
l
vin
g
opti
m
iz
at
i
on
pro
blem
,
m
et
aheu
risti
cs
searc
h
so
luti
on
r
an
do
m
ly
and
by
tr
ia
l
and
er
r
or
.
They
are
not
li
ke
determ
inist
ic
m
et
ho
ds
wh
ic
h
re
quire
init
ial
gu
e
ss
[4
]
an
d
m
at
he
m
a
ti
cal
req
ui
rem
ents,
e
.g
.
gradie
nt
or
co
ntinuo
us
functi
on
s
[5]
.
They
on
ly
require
obj
ect
ive
f
unct
ion
an
d
the
sea
rch
i
ng
dom
ai
n
in
so
lvin
g
pr
oblem
s
[6
]
.
Moreo
ve
r,
they
relat
ively
need
che
ape
r
com
pu
ta
ti
on
c
os
t t
ha
n
t
he dete
rm
inistic on
es
.
Since 1960s
, me
ta
heu
risti
cs h
ave
bee
n
rap
i
dly dev
el
op
e
d [6
]
. Som
e o
f
the f
am
ou
s
m
et
a
heurist
ic
s ar
e
Gen
et
ic
Al
gori
thm
(G
A)
,
Par
ti
cl
e
Sw
arm
Op
tim
iz
at
ion
(PSO)
a
nd
Di
ff
e
ren
ti
al
Ev
olu
ti
on
(DE).
They
hav
e
been
widely
im
ple
m
ented
in
var
i
ous
op
ti
m
iz
at
ion
pro
bl
e
m
s.
H
ow
e
ve
r
,
they
ha
ve
a
sam
e
disadv
a
ntage
,
i.e.
easy
to
fa
ll
into
local
optim
a
[7
-
9]
or
te
nd
to
pr
em
at
ur
el
y
co
nv
e
r
ge
[
9,
10
]
.
A
s
the
co
ns
e
quence,
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
Leaders
and f
ol
lo
we
rs a
l
go
rit
hm for c
onstr
ai
ned
non
-
li
ne
ar o
ptimizati
on
(
Helen Yuli
ana An
gmalisa
ng
)
163
they
of
te
n
fail
to
appr
oach
th
e
op
ti
m
al
so
luti
on
.
T
heref
or
e
,
it
is
necessary
to
find
a
m
etah
eu
risti
cs
that
can
perform
b
et
te
r i
n
s
olv
in
g o
ptim
iz
at
ion
p
r
oblem
s.
In
[9
]
,
it
is
sta
t
ed
that
the
m
ain
cause
of
pr
e
m
at
ur
e
co
nv
e
r
gen
ce
i
n
these
well
-
kn
own
al
gorithm
s
is
the
direct
com
par
ison
of
new
ly
discov
e
re
d
so
luti
on
s
with
th
e
cur
re
nt
be
st
-
know
n
sol
ution.
Ther
e
f
or
e,
G
onzal
es
-
Fe
r
nand
ez
an
d
C
hen
[
9]
propose
d
a
no
vel
m
et
aheu
ri
sti
cs
nam
ed
Leaders
an
d
F
ollow
e
r
s
(LaF
)
w
hich
avo
i
ds
this
ki
nd
of
c
om
par
ison.
I
n
[
9],
L
aF
is
bette
r
in
so
l
ving
unc
on
st
raine
d
no
n
-
li
nea
r
op
ti
m
iz
ation
than
P
SO
a
nd
DE.
It
is
able
to
exp
lo
re
be
tt
er
so
that
it
can
pe
rfor
m
bette
r
in
m
uti
-
m
od
al
op
ti
m
iz
ation
pro
blem
s.
Mor
eov
e
r,
LaF
is
sim
ple
and
do
e
s
no
t
nee
d
a
ny
pa
ram
e
te
r,
s
o
it
m
a
y
save
com
pu
ta
ti
on
ti
m
e
becau
se
th
ere
is
no
need
to
est
i
m
a
te
a
ny
par
am
et
er.
Howev
e
r,
in
[9
]
there
is
no
an
y
discuss
i
on
a
bout
bounda
ry
c
on
st
raint
ha
nd
l
ing
,
e
ven
th
ough
the
re
is
a
possibil
it
y
that
so
m
e
new
so
l
utions
create
d by the
op
e
rato
r
in
La
F is outsi
de
the
searc
hing s
pac
e.
Ther
e
a
re
s
ome
m
et
ho
ds
to
deal
with
the
bounda
ry
co
nst
raint
vio
la
ti
on
s.
So
m
e
of
th
e
m
are
re
-
init
ia
li
zation
a
nd
cl
am
pin
g
(
br
i
ng
back
t
he
so
luti
on
to
th
e
peak
value
).
In
[
11]
,
it
is
pro
ven
that
cl
a
m
pin
g
m
et
ho
d
is
m
ore
eff
ect
ive
tha
n
re
-
init
ia
li
zat
i
on
m
et
ho
d.
It
can
im
pr
ove
the
s
olu
ti
on
m
uch
bette
r
tha
n
the
re
-
init
ia
li
zation
m
et
hod.
The
refo
re,
it
ca
n be
use
d
in
LaF
to ha
nd
le
t
he bou
ndary co
ns
trai
nt
vio
la
ti
o
n.
Fo
r
s
olv
i
ng
c
on
st
raine
d
optim
iz
ation
pro
blem
,
a
m
et
a
heurist
ic
sho
ul
d
be
m
od
ifie
d
us
in
g
a
const
raint
-
ha
ndli
ng
te
c
hn
i
que.
The
re
are
var
i
ou
s
c
onstr
ai
nt
-
ha
ndli
ng
t
echn
i
qu
e
s.
T
he
m
os
t
widely
us
ed
te
chn
iq
ue
is
pen
al
ty
functi
on
[12].
It
m
od
i
fies
the
ob
j
ect
ive
f
un
ct
io
n
by
ad
ding
a
pen
al
ty
fun
ct
ion
.
This
te
ch
nique
has
bee
n
being
us
e
d
with
var
i
ou
s
m
et
ah
eur
ist
ic
s,
both
the
old
a
nd
the
new
ones.
In
[
13
]
,
Har
m
on
y
Sea
r
ch
(
HS)
al
go
rithm
was
m
od
ifie
d
us
in
g
death
pen
al
ty
,
sta
ti
c
pen
al
ty
an
d
a
ne
w
pen
al
ty
f
unct
ion
te
chn
i
que,
na
m
ed
two
sta
ge
pe
nalty
functi
on.
In
[
14]
,
sta
ti
c
pen
al
ty
an
d
feasibil
it
y
r
ules
m
et
ho
d
were
us
e
d
with
Fire
fly
Algorithm
(FA)
f
or
c
onstr
ai
ned
optim
izati
on
.
In
[
15]
,
sta
ti
c
pen
al
t
y
te
chn
iq
ue
was
al
s
o
com
bin
ed
with
a
no
vel
m
etah
eu
risti
c,
na
m
ed
Ba
ct
erial
-
insp
ire
d
Al
gor
it
h
m
,
for
c
ons
trai
ned
opti
m
i
zat
ion
.
Stat
ic
pen
al
ty
and
dy
nam
ic
pen
al
ty
functi
on
we
re
al
so
use
d
with
a
n
em
erg
i
ng
m
et
aheu
risti
c,
nam
ed
Cohort
In
te
ll
igence
(C
I)
f
or
co
ns
trai
ne
d
opti
m
iz
at
io
n
[
16
]
.
In
[17],
Diff
e
ren
ti
al
S
earch
(DS)
al
gorithm
is
de
velop
e
d
for
c
onstrai
ne
d o
pti
m
iz
at
ion
w
it
h
sta
ti
c an
d dynam
ic
p
enalt
y functi
on.
In
t
his
stu
dy,
LaF
is
im
ple
mented
for
s
olv
i
ng
c
onstrai
ne
d
op
ti
m
iz
ation
pro
blem
us
ing
sta
ti
c
pen
al
ty
functi
on
f
or
ha
nd
li
ng
the
c
on
st
raints
a
nd
cl
a
m
pin
g
m
e
thod
[
11
]
f
or
handlin
g
t
he
bounda
ry
co
ns
trai
nt
vio
la
ti
on.
Af
te
r
bein
g
m
od
ifie
d,
the
pro
pos
ed
al
gorithm
was
evaluate
d
us
in
g
seve
ral
well
-
kn
own
be
nch
m
ark
pro
blem
s.
Then
,
t
he
e
valuati
on
res
ults
of
t
he
propose
d
a
lgorit
hm
are
com
par
ed
with
oth
e
r
m
et
aheu
risti
cs
[13,
14
]
,
[
16
-
1
8].
Sect
io
n
2
int
rod
uces
the
propose
d
al
gorithm
.
S
ect
ion
3
is
the
researc
h
m
et
ho
d.
Sect
ion
4 p
rese
nts a
nd d
isc
us
s
es the
res
ults.
The
n,
t
he
c
on
c
lusio
ns
a
re
give
n
in
Secti
on
5.
2.
THE
PROPO
SED
ALGO
R
ITHM
Leader
s
a
nd
F
ollow
e
rs
(LaF
)
al
gorithm
us
e
s
two
di
ff
e
ren
t
popula
ti
ons,
i.
e.
Le
ad
e
rs
a
nd
Followers
.
Followers
is
assigne
d
t
o
e
xp
l
or
e
so
m
e
new
s
ub
-
re
gions
of
the
sear
chin
g
s
pace
that
hav
e
local
opti
m
a
(which
cal
le
d
at
tract
ion
ba
sin),
w
her
ea
s
Le
ad
e
rs
is
assig
ned
t
o
sto
re
prom
isi
ng
so
l
ution
s
w
hich
m
a
y
be
a
global
optim
um
.
In
this
al
go
rithm
,
there
is
no
c
om
par
ison
of
ne
w
dis
cov
e
re
d
so
l
ution
s
a
nd
a
best
curren
t
so
luti
on.
T
his
kind
of
c
omparis
on
is
a
voide
d
to
pr
e
ve
nt
prem
at
ur
e
conve
rg
e
nce.
Algorithm
1
is
the
ps
e
udoc
od
e
of
Leader
s a
nd F
ollow
e
rs
al
gori
thm
.
Ther
e
is
possi
bili
ty
that
Trial
is
form
ed
ou
tsi
de
the
se
arch
i
ng
s
pace.
To
handle
th
e
bo
undary
const
raint
vio
l
at
ion
,
this
st
udy
us
es
cl
am
pi
ng
m
et
ho
d
by
br
i
ng
i
ng
the
s
olu
ti
on
t
o
th
e
bounda
ry
valu
e
[
11
]
.
The
al
gorithm
is
m
od
ifie
d
by
add
in
g
the
c
onditi
on
al
s
on
li
ne
15
-
16.
If
the
p
os
it
ion
of
Trial
is
not
in
the
searchi
ng
s
pac
e,
the
posit
ion
is
m
ov
e
d
to
t
he
boun
dar
y.
To
ha
nd
le
t
he
c
onstrai
nts,
t
his
al
gorithm
us
es
pe
nalty
te
chn
iq
ue.
T
hi
s
te
ch
nique
is
the
m
os
t
wi
dely
co
ns
trai
nt
-
ha
nd
li
ng
te
c
hn
i
qu
e
.
It
tra
ns
f
or
m
s
co
ns
t
raine
d
pro
blem
into
un
c
onstrai
ne
d
prob
le
m
.
The
obj
ect
ive
f
unct
ion
is
m
odifie
d
by
ad
di
ng
pe
nalty
functi
on.
The ge
ner
al
form
o
f
pe
nalty
fun
ct
io
n
is
as
fol
lows
.
(
⃗
)
=
(
⃗
)
+
∑
×
ma
x
[
0
,
(
⃗
)
]
×
+
∑
×
(
|
ℎ
(
⃗
)
|
−
)
×
(
⃗
)
is
m
od
ifie
d
ob
j
ect
ive
f
un
ct
io
n,
(
⃗
)
is
ori
gin
al
obj
ect
iv
e
functi
on
of
co
ns
trai
ne
d
optim
iz
ation
pro
blem
,
is
pen
al
ty
fac
tor
w
hich
sho
uld
be
la
rg
e
e
nough
for
m
ini
m
iz
ation
pro
bl
e
m
s,
(
⃗
)
is
or
i
gin
al
ineq
ualit
y
const
raints,
ℎ
(
⃗
)
is or
i
gin
al
e
qu
al
it
y const
raints,
and
are
bo
t
h
c
on
sta
nts a
nd
is err
or tolera
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.
13
, N
o.
1
,
Ja
nu
a
ry 20
19
:
162
–
169
164
3.
RE
SEA
R
CH MET
HO
D
The
pro
posed
al
gorithm
was
eval
uated
usi
ng
well
-
know
n
ben
c
hm
ark
pro
blem
s
fo
r
const
raine
d
op
ti
m
iz
ation
(
1
-
13
).
Table
1
is
t
he
su
m
m
ary
of
the
ben
c
hm
ark
pro
blem
wh
e
r
e
ρ
is
t
he
rati
o
of
t
he
f
easi
ble
search
s
pace
s
iz
e
an
d
the
e
nt
ire
searc
h
s
pa
ce,
L
I
is
t
he
num
ber
of
li
ne
ar
ine
qual
it
y
con
st
raint,
N
I
i
s
the
nu
m
ber
of
no
n
-
li
near
i
nequal
it
y
con
strai
nt,
NE
is
t
he
nu
m
ber
of
nonline
ar
e
qu
al
it
y
co
ns
trai
nt
a
nd
a
is
the
nu
m
ber
of
act
ive
co
ns
trai
nt.
Table
2
pr
es
ents
the
detai
l
s
of
pro
blem
.
Each
opti
m
izati
on
pr
ob
le
m
was
evaluate
d
in
t
wen
ty
-
five
i
ndepende
nt
r
uns
with
va
rio
us
popula
ti
on
siz
e,
=
10
,
25
,
50
,
100
.
The
al
gorithm
was
st
oppe
d
if
there
was
no
be
tt
er
so
l
ution
f
ound
in
5000
it
erati
on
s
in
a
row
or
the
al
gori
thm
had
been
r
un
i
n
600
sec
onds.
T
he
propose
d
al
gorithm
us
es
s
ta
ti
c
pen
al
ty
fa
ct
or
a
nd
pa
ram
et
ers,
i.e.
=
50
,
000
,
=
1
,
=
1
an
d
=
0
.
0001
.
If
the
pr
opos
e
d
al
gorithm
m
eet
s
diff
ic
ulty
to
reach
t
he
optim
al
so
luti
on
of
a
te
st
functi
on,
the alg
or
it
hm
w
il
l be e
valuat
ed wit
h
a
bigge
r pop
ulati
on
si
ze an
d
lo
nger
c
om
pu
ta
ti
on
ti
m
e lim
it
.
Algorithm
1.
P
seu
do
c
ode
of
Leader
s a
nd F
ollow
e
rs Alg
or
it
h
m
Table
1.
Su
m
m
ary o
f
the
Benc
hm
ark
Pro
blem
s
Test Fun
ctio
n
Op
ti
m
al Solu
tio
n
Di
m
en
sio
n
Ty
p
e of
(%)
LI
NI
N
E
a
-
15
13
Qu
ad
ratic
0
.00
0
3
9
0
0
6
-
0
.80
3
6
1
9
1
20
No
n
lin
ear
9
9
.99
6
2
0
2
0
1
-
1
.00
0
5
0
0
1
10
Po
ly
n
o
m
ial
0
.00
0
2
0
0
1
1
-
3
0
6
6
5
.5
3
9
5
Qu
ad
ratic
2
6
.90
8
9
0
6
0
2
5
1
2
6
.4
9
6
7
1
4
4
Cu
b
ic
0
.00
0
0
2
0
3
3
-
6
9
6
1
.8138
7
6
2
Cu
b
ic
0
.00
6
5
0
2
0
2
2
4
.30
6
2
0
9
0
7
10
Qu
ad
ratic
0
.00
0
1
3
5
0
6
-
0
.09
5
8
2
5
0
4
2
No
n
lin
ear
0
.84
8
4
0
2
0
0
6
8
0
.6300
5
7
4
7
Po
ly
n
o
m
ial
0
.53
1
9
0
4
0
2
7
0
4
9
.2
4
8
0
2
0
5
8
Linear
0
.00
0
5
3
3
0
6
0
.74
9
9
2
Qu
ad
ratic
0
.00
9
9
0
0
1
1
-
1
3
Qu
ad
ratic
4
.74
5
2
0
1
0
0
0
.05
3
9
4
1
5
1
4
5
Exp
o
n
en
tial
0
.00
0
0
0
0
3
3
1:
= nu
m
b
er
o
f
decis
io
n
variabl
es
2:
= po
p
u
latio
n
size
3:
(
)
= lower
bo
u
n
d
o
f
-
th
decis
io
n
variabl
es
4:
(
)
= up
p
er
b
o
u
n
d
o
f
-
th
decis
io
n
variabl
es
5:
= initialize
L
eade
r
s with
u
n
ifor
m
r
an
d
o
m
vecto
rs
6:
= initialize
Follo
wers with
u
n
if
o
r
m
r
an
d
o
m
vecto
rs
7:
repeat
8:
for
= 1:
do
9:
= r
o
u
n
d
(
rand
*
)
10:
= r
o
u
n
d
(
rand
*
)
11:
=
(
,:)
12:
=
(
,:)
13:
for
= 1:
do
14:
(
)
=
(
)
+
rand
*
2
*
(
(
)
–
(
))
15:
if
(
)
<
(
)
then
(
)
=
(
)
16:
if
(
)
>
(
)
then
(
)
=
(
)
13:
end f
o
r
14:
if
(
)
<
(
)
then
(
,:)
=
(:)
15:
end f
o
r
1
6
:
if
m
ed
i
an
(
(
))
<
m
e
d
ian
(
(
))
then
17:
(
1
)
= an ele
m
en
t of
o
r
wh
ich
has
the b
est f
itn
ess
18:
for
= 2:
do
19:
= pick
an ele
m
en
t
o
f
rand
o
m
l
y
20:
= pick
an ele
m
en
t
o
f
rand
o
m
l
y
21:
if
(
)
<
(
)
then
(
)
=
22:
else
(
)
=
23:
end
24:
= r
ein
itial
ize Foll
o
wers un
if
o
r
m
l
y
25:
end if
26:
until
th
e t
er
m
in
ati
o
n
cr
iterion
is satis
f
ied
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
Leaders
and f
ol
lo
we
rs a
l
go
rit
hm for c
onstr
ai
ned
non
-
li
ne
ar o
ptimizati
on
(
Helen Yuli
ana An
gmalisa
ng
)
165
T
able
2.
Detai
ls o
f
the
Benc
hm
ark
P
roblem
s
Ob
jectiv
e Fun
ctio
n
Co
n
strain
ts
Bo
u
n
d
s
(
⃗
)
=
5
∑
−
4
=
1
5
∑
2
−
4
=
1
∑
13
=
5
1
(
⃗
)
=
2
1
+
2
2
+
10
+
11
−
10
≤
0
2
(
⃗
)
=
2
1
+
2
3
+
10
+
12
−
10
≤
0
3
(
⃗
)
=
2
2
+
2
2
+
11
+
12
−
10
≤
0
4
(
⃗
)
=
−
8
1
+
10
≤
0
5
(
⃗
)
=
−
8
2
+
11
≤
0
6
(
⃗
)
=
−
8
3
+
12
≤
0
7
(
⃗
)
=
−
2
4
−
5
+
10
≤
0
8
(
⃗
)
=
−
2
6
−
7
+
11
≤
0
9
(
⃗
)
=
−
2
8
−
9
+
12
≤
0
= (
0
,
0, …
,
0)
= (
1
,
1, 1,
1
,
1
,
1
,
1
,
1
,
1
,
1
0
0
,
1
0
0
,
1
0
0
,
1
)
(
⃗
)
=
−
|
∑
cos
(
)
4
−
2
∏
cos
(
)
2
=
1
=
1
√
∑
2
=
1
|
= 20
1
(
⃗
)
=
0
.
75
−
∏
≤
0
=
1
2
(
⃗
)
=
∑
=
1
−
7
.
5
≤
0
= 0;
= 10
;
(
⃗
)
=
−
(
√
)
∏
=
1
=
10
ℎ
1
(
⃗
)
=
∑
2
−
1
=
0
=
1
= 0;
= 1;
(
⃗
)
=
5
.
3578547
3
2
+
0
.
8356891
1
5
+
37
.
293239
1
−
40792
.
141
1
(
⃗
)
=
85
.
334407
+
0
.
0056858
2
5
+
0
.
0006262
1
4
−
0
.
0022053
3
5
−
92
≤
0
2
(
⃗
)
=
−
85
.
334407
−
0
.
0056858
2
5
−
0
.
0006262
1
4
+
0
.
0022053
3
5
≤
0
3
(
⃗
)
=
80
.
51249
+
0
.
0071317
2
5
+
0
.
0029955
1
2
+
0
.
0021813
3
2
−
110
≤
0
4
(
⃗
)
=
−
80
.
51249
−
0
.
0071317
2
5
−
0
.
0029955
1
2
−
0
.
0021813
3
2
+
90
≤
0
5
(
⃗
)
=
9
.
300961
+
0
.
0047026
3
5
+
0
.
0012547
1
3
+
0
.
0019085
3
4
−
25
≤
0
6
(
⃗
)
=
−
9
.
300961
−
0
.
0047026
3
5
−
0
.
0012547
1
3
−
0
.
0019085
3
4
+
20
≤
0
= (
7
8
,
3
3
,
2
7
,
2
7
,
2
7
)
= (
1
0
2
,
4
5
,
4
5
,
4
5
,
4
5
)
(
⃗
)
=
3
1
+
0
.
000001
1
3
+
2
2
+
(
0
.
000002
3
)
2
3
1
(
⃗
)
=
−
4
+
3
−
0
.
55
≤
0
2
(
⃗
)
=
−
3
+
4
−
0
.
55
≤
0
ℎ
3
(
⃗
)
=
1000
sin
(
−
3
−
0
.
25
)
+
1000
sin
(
−
4
−
0
.
25
)
+
894
.
8
−
1
=
0
ℎ
4
(
⃗
)
=
1000
sin
(
3
−
0
.
25
)
+
1000
sin
(
3
−
4
−
0
.
25
)
+
894
.
8
−
2
=
0
ℎ
5
(
⃗
)
=
1000
sin
(
4
−
0
.
25
)
+
1000
sin
(
4
−
3
−
0
.
25
)
+
1294
.
8
=
0
= (
0
,
0,
-
0
.55
,
-
0
.55
)
= (
1
2
0
0
,
1
2
0
0
,
0
.55
,
0
.55
)
(
⃗
)
=
(
1
−
10
)
3
+
(
2
−
20
)
3
1
(
⃗
)
=
−
(
1
−
5
)
2
−
(
2
−
5
)
2
+
100
≤
0
2
(
⃗
)
=
(
1
−
6
)
2
+
(
2
−
5
)
2
−
82
.
81
≤
0
= (
1
3
,
0
)
= (
1
0
0
,
100)
(
⃗
)
=
1
2
+
2
3
+
1
2
−
14
1
−
16
2
+
(
3
−
10
)
2
+
4
(
4
−
5
)
2
+
(
5
−
3
)
2
+
2
(
6
−
1
)
2
+
5
7
2
+
7
(
8
−
11
)
2
+
2
(
9
−
10
)
2
+
(
10
−
7
)
2
+
45
1
(
⃗
)
=
−
105
+
4
1
+
5
2
−
3
7
+
9
8
≤
0
2
(
⃗
)
=
10
1
−
8
2
−
17
7
+
2
8
≤
0
3
(
⃗
)
=
−
8
1
+
2
2
+
5
9
−
2
10
−
12
≤
0
4
(
⃗
)
=
3
(
1
−
2
)
2
+
4
(
2
−
3
)
2
+
2
3
2
−
7
4
−
120
≤
0
5
(
⃗
)
=
5
1
2
+
8
2
+
(
3
−
6
)
2
−
2
4
−
40
≤
0
6
(
⃗
)
=
1
2
+
2
(
2
−
2
)
2
−
2
1
2
+
14
5
−
6
6
≤
0
7
(
⃗
)
0
.
5
(
1
−
8
)
2
+
2
(
2
−
4
)
2
+
3
5
2
−
6
−
30
≤
0
8
(
⃗
)
=
−
3
1
+
6
2
+
12
(
9
−
8
)
2
−
7
10
≤
0
=
-
10
= 10
(
⃗
)
=
−
(
sin
(
2
1
)
)
3
sin
(
2
2
)
1
3
(
1
+
2
)
1
(
⃗
)
=
−
1
2
−
2
+
1
≤
0
2
(
⃗
)
=
1
−
1
+
(
2
−
4
)
2
≤
0
= 0
= 10
(
⃗
)
=
(
1
−
10
)
2
+
5
(
2
−
12
)
2
+
3
4
+
3
(
4
−
11
)
2
+
10
5
6
+
7
6
2
+
7
4
−
4
6
7
−
10
6
−
8
7
1
(
⃗
)
=
−
127
+
2
1
2
+
3
2
4
+
3
+
4
4
2
+
5
5
≤
0
2
(
⃗
)
=
−
282
+
7
1
+
3
2
+
10
3
2
+
4
−
5
≤
0
3
(
⃗
)
=
−
196
+
23
1
+
2
2
+
6
6
2
−
8
7
≤
0
4
(
⃗
)
=
4
1
2
+
2
2
−
3
1
2
+
2
3
2
+
5
6
−
11
7
≤
0
=
-
10
= 10
(
⃗
)
=
1
+
2
+
3
1
(
⃗
)
=
−
1
+
0
.
0025
(
4
+
6
)
≤
0
2
(
⃗
)
=
−
1
+
0
.
0025
(
5
+
7
−
4
)
≤
0
3
(
⃗
)
=
−
1
+
0
.
01
(
8
−
5
)
≤
0
4
(
⃗
)
=
−
1
6
+
833
.
33252
4
+
100
1
−
83333
.
333
≤
0
5
(
⃗
)
=
−
2
7
+
1250
5
+
2
4
−
1250
4
≤
0
= (
1
0
0
,
1
0
0
0
,
1
0
0
0
,
1
0
,
1
0
,
1
0
,
1
0
,
1
0
)
= (
1
0
0
0
0
,
1
0
0
0
0
,
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
S
N
:
2502
-
4752
Ind
on
esi
a
n
J
E
le
c Eng &
Co
m
p
Sci,
Vo
l.
13
, N
o.
1
,
Ja
nu
a
ry 20
19
:
162
–
169
166
Ob
jectiv
e Fun
ctio
n
Co
n
strain
ts
Bo
u
n
d
s
6
(
⃗
)
=
−
3
8
+
1250000
+
3
5
−
2500
5
≤
0
1
0
0
0
0
,
1
0
0
0
,
1
0
0
0
,
1
0
0
0
,
1
0
0
0
,
1
0
0
0
)
(
⃗
)
=
1
2
+
(
2
−
1
)
2
ℎ
(
⃗
)
=
2
−
1
2
=
0
=
-
1
= 1
(
⃗
)
=
−
(
100
−
(
1
−
5
)
2
−
(
2
−
5
)
2
−
(
3
−
5
)
2
)
/
100
(
⃗
)
=
(
1
−
)
2
+
(
2
−
)
2
+
(
3
−
)
2
−
0
.
0625
≤
0
= 0
= 10
(
⃗
)
=
1
2
3
4
5
ℎ
1
(
⃗
)
=
1
2
+
2
2
+
3
2
+
4
2
+
5
2
−
10
=
0
ℎ
2
(
⃗
)
=
2
3
−
5
4
5
=
0
ℎ
3
(
⃗
)
=
1
3
+
2
3
+
1
=
0
=
-
2
.3
= 2.3
4.
RESU
LT
S
A
ND D
I
SCUS
S
ION
Table
3
a
nd
4
pr
ese
nts
t
he
evalu
at
ion
re
su
lt
s
with
popula
ti
on
siz
e
(
)
=
10
,
25,
50
an
d
100.
The
al
gorithm
obta
ins
the
optim
al
so
luti
on
f
or
1
,
2
,
4
,
6
,
8
,
9
,
11
and
12
.
T
he
sta
nd
a
r
d
devi
at
ion
of
al
l
ob
ta
ine
d
s
olu
t
ion
s
f
or
4
,
6
,
8
a
nd
12
ap
proac
hes
zer
o.
T
his
m
eans
that
in
al
l
r
uns,
the
al
gorithm
consi
ste
ntly
obta
ins
optim
al
so
l
ution
s
f
or
these
pro
ble
m
s.
Table
1
sh
ows
that
al
l
of
these
pr
ob
le
m
s
(
4
,
6
,
8
and
12
)
a
re
lo
w dim
ension
al
(
≤
5
) a
nd h
a
ve n
o
e
qual
it
y con
st
raints.
Table
3.
T
he
Result
s Obtai
ne
d by the
Pro
posed Alg
or
it
hm
f
or
1
−
7
Prob
le
m
1
2
3
4
5
6
7
Op
ti
m
al Solu
tio
n
-
15
-
0
.80
3
6
-
1
.00
0
5
-
30666
5
1
2
6
.5
-
6
9
6
1
.8
2
4
.3
0
6
2
Ob
tain
ed
Solu
tio
n
n
=
1
0
Bes
t
-
15
-
0
.80
3
6
-
1
.00
0
2
-
30666
5
1
2
6
.8
-
6
9
6
1
.8
2
4
.35
5
8
Mean
-
1
2
.98
4
3
-
0
.68
4
9
-
1
.00
0
2
-
30666
5
2
6
4
.9
-
6
9
6
1
.8
2
5
.41
9
9
W
o
rst
-
1
0
.10
9
4
-
0
.57
0
2
-
1
.00
0
2
-
30666
5
6
9
3
.3
-
6
9
6
1
.8
2
7
.43
5
9
Std
1
.49
2
3
0
.06
8
5
2E
-
06
4E
-
06
2
0
9
.63
4
.5E
-
11
0
.74
8
5
n
=
2
5
Bes
t
-
15
-
0
.78
8
1
-
1
.00
0
2
-
30666
5
1
3
4
.0
-
6
9
6
1
.8
2
4
.31
8
8
Mean
-
1
4
.08
4
4
-
0
.73
1
9
-
0
.98
3
5
-
30666
5
2
3
9
.4
-
6
9
6
1
.8
2
4
.72
0
2
W
o
rst
-
1
1
.82
8
1
-
0
.60
5
7
-
0
.83
2
6
-
30666
5
7
9
0
.5
-
6
9
6
1
.8
2
6
.04
0
5
Std
0
.94
8
3
0
.04
5
2
0
.05
3
0
7E
-
10
1
9
8
.89
2
.3E
-
12
0
.39
3
3
n
=
5
0
Bes
t
-
15
-
0
.80
3
6
-
1
.00
0
3
-
30666
5
1
2
8
.6
-
6
9
6
1
.8
2
4
.31
5
6
Mean
-
1
4
.47
3
7
-
0
.76
4
5
-
0
.95
3
8
-
30666
5
2
5
3
.1
-
6
9
6
1
.8
2
4
.65
0
4
W
o
rst
-
1
1
.28
1
2
-
0
.57
4
4
-
0
.71
9
1
-
30666
5
6
7
3
.6
-
6
9
6
1
.8
2
5
.07
7
Std
1
.05
3
3
0
.04
9
7
0
.10
0
6
1E
-
11
1
6
4
.81
0
0
.25
4
5
n
=
1
0
0
Bes
t
-
15
-
0
.80
3
6
-
1
.00
0
2
-
30666
5
1
2
6
.7
-
6
9
6
1
.8
2
4
.30
8
4
Mean
-
15
-
0
.78
9
7
-
0
.95
2
1
-
30666
5
3
2
7
.2
-
6
9
6
1
.8
2
4
.45
4
9
W
o
rst
-
15
-
0
.76
9
2
-
0
.80
7
4
-
30666
5
7
1
4
.9
-
6
9
6
1
.8
2
4
.82
4
2
Std
7
.25
E
-
16
0
.00
9
0
.06
1
1
6E
-
12
2
2
4
.09
0
0
.13
0
3
Std
=
Stan
d
ard De
v
iatio
n
Table
4.
T
he
Result
s Obtai
ne
d by the
Pro
posed Alg
or
it
hm
f
or
8
−
13
Prob
le
m
8
9
10
11
12
13
Op
ti
m
al Solu
tio
n
-
0
.09
5
8
6
8
0
.6301
7
0
4
9
.2
0
.74
9
9
-
1
0
.05
3
9
Ob
tain
ed
Solu
tio
n
n
=
1
0
Bes
t
-
0
.09
5
8
6
8
0
.6333
7
0
9
5
.9
0
.74
9
9
-
1
0
.17
8
9
M
ean
-
0
.09
5
8
6
8
0
.6518
7
8
9
7
.8
0
.74
9
9
-
1
1
.07
8
5
W
o
rst
-
0
.09
5
8
6
8
0
.6736
1
0
9
1
1
0
.74
9
9
-
1
4
.95
1
1
Std
4
.01
E
-
18
1
.03
E
-
02
8
6
8
.6
9
.3E
-
8
0
1
.51
9
4
n
=
2
5
Bes
t
-
0
.09
5
8
6
8
0
.6324
7
0
4
9
.5
0
.74
9
9
-
1
0
.08
6
5
Mean
-
0
.09
5
8
6
8
0
.6360
7
4
6
8
.2
0
.74
9
9
-
1
1
.03
2
3
W
o
rst
-
0
.09
5
8
6
8
0
.6423
8
0
7
6
.3
0
.74
9
9
-
1
5
.05
5
1
Std
4
.91
E
-
18
2
.80
E
-
03
2
9
8
.6
6
.5E
-
8
0
1
.47
7
1
n
=
5
0
Bes
t
-
0
.09
5
8
6
8
0
.6305
7
1
1
4
.1
0
.74
9
9
-
1
0
.07
3
0
Mean
-
0
.09
5
8
6
8
0
.6319
7
2
9
8
.3
0
.75
0
2
-
1
1
.64
2
6
W
o
rst
-
0
.09
5
8
6
8
0
.6339
7
5
4
7
.2
0
.75
2
9
-
1
1
1
.09
9
6
Std
8
.96
E
-
18
1
.00
E
-
03
1
1
8
.2
9
.6E
-
4
0
2
.31
4
1
n
=
1
0
0
Bes
t
-
0
.09
5
8
6
8
0
.6301
7
0
8
1
.5
0
.74
9
9
-
1
0
.61
4
8
Mean
-
0
.09
5
8
6
8
0
.6307
7
2
3
9
.8
0
.75
0
0
-
1
0
.96
2
5
W
o
rst
-
0
.09
5
8
6
8
0
.6321
7469
0
.75
0
5
-
1
3
.18
4
3
Std
8
.96
E
-
18
4
.59
E
-
04
1
1
1
.7
2
.0E
-
4
0
0
.78
7
7
Fo
r
1
wh
ic
h
is
a
high
dim
ensi
on
al
pro
blem
(
=
13
),
the
obta
ine
d
error
s
is
quit
e
high
wh
e
n
th
e
popula
ti
on
siz
e
(
)
=
10,
25
and
50,
but
w
hen
the
po
pu
l
at
ion
siz
e
(
)
=
100,
the
er
r
ors
ap
proac
hes
z
ero.
The
al
gorit
hm
su
ccess
fu
ll
y
ob
ta
in
s
the
optim
a
l
resu
lt
in
each
r
un
w
he
n
the
popula
ti
on
siz
e
(
)
=
100,
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
Leaders
and f
ol
lo
we
rs a
l
go
rit
hm for c
onstr
ai
ned
non
-
li
ne
ar o
ptimizati
on
(
Helen Yuli
ana An
gmalisa
ng
)
167
al
tho
ug
h
t
he
num
ber
of
dim
e
ns
io
n
a
nd
i
nequali
ty
const
raint
in
1
is
higher
tha
n
7
,
9
a
nd
10
.
Ta
ble
1
s
ho
w
s
that
the
diff
e
re
nce
of
1
an
d
7
,
9
,
10
excep
t
t
he
dim
ensio
nalit
y
is
the
ty
pe
o
f
ine
qu
al
it
y
co
ns
tra
ints
in
t
he
pro
blem
.
1
has
on
ly
li
near
c
on
strai
nts,
unli
ke
7
,
9
and
10
wh
ic
h
ha
ve
no
nlinear
const
raints.
T
he
al
gorithm
te
nd
s
to
hav
e
diff
ic
ulty
in
s
olv
in
g
the
op
t
i
m
iz
ation
pro
bl
e
m
s
with
e
qu
al
it
y
con
strai
nt
(s)
(
3
,
5
and
13
),
excep
t
11
.
I
n
11
,
th
e
pro
po
se
d
al
gorithm
con
sist
ently
appro
ac
he
s
the
opti
m
al
so
luti
on
wh
e
n
=
10
an
d
25.
Table
1
s
hows
that
the
di
ff
e
r
ence
of
11
a
nd
t
he
ot
her
s
is
it
is
lo
w
dim
ension
al
a
nd
has
only
one
e
qu
al
i
ty
const
raint.
Mo
reover
,
the
al
gorithm
te
nd
s
to
fin
d
di
ff
ic
ul
ty
in
so
lving
t
he
high
dim
en
sion
al
opti
m
izati
on
pro
blem
s w
it
h
nonlinea
r
ine
qual
it
y con
st
raints
on
ly
(
2
,
7
,
9
an
d
10
).
Wh
e
n
t
he
pro
pose
d
al
gorithm
is
evaluate
d
on
2
,
7
,
9
and
10
with
a
big
popula
ti
on
siz
e,
e.g.
=
2000,
a
nd
the
sam
e
te
r
m
inatio
n
crit
erio
n,
th
e
obta
ined
s
olu
ti
ons
are
m
uch
bette
r
a
nd
th
e
sta
ndar
d
devi
at
ion
s
are
m
uch
sm
aller
eve
n
th
ough
the
c
om
pu
ta
ti
on
al
ti
m
e
l
i
m
i
t
is
sa
m
e,
e.g
.
600
sec
onds
(
Table
5).
Wh
e
n
th
e
tim
e
lim
it
is
lon
ge
r,
i.e.
1200
seco
nd
s
,
Table
5
sho
ws
that
LaF
does
not
obta
in
bette
r
s
ol
utions,
exc
ept
on
2
.
Th
us
,
in
s
olv
i
ng
t
he
opti
m
i
zat
ion
pro
ble
m
s
with
high
dim
ension
al
optim
iz
at
ion
prob
le
m
s
with
nonlinea
r
ineq
ualit
y con
s
trai
nts only
,
La
F r
e
quires a
b
i
g pop
ulati
on
siz
e (
≥
2000)
.
Table
5.
T
he
Result
s Obtai
ne
d by the
Pro
posed Alg
or
it
hm
for Hig
h Dim
e
ns
io
nal
Op
ti
m
i
zat
ion
P
r
ob
le
m
s
with
ineq
ualit
y con
s
trai
nts only
(
2
,
7
,
9
an
d
10
)
w
he
n
= 2,0
00
Ti
m
e
L
i
m
it
600s
1200s
Prob
le
m
2
7
9
10
2
7
9
10
Bes
t
-
0
.80
3
5
2
4
.31
1
4
6
8
0
.6303
7
.14
E+03
-
0
.80
3
6
2
4
.30
8
5
6
8
0
.6303
7
.09
E+03
Mean
-
0
.80
3
5
2
4
.31
6
8
6
8
0
.6305
7
.19
E+03
-
0
.80
3
6
2
4
.31
9
9
6
8
0
.6305
7
.17
E+03
W
o
rst
-
0
.80
3
4
2
4
.33
4
9
6
8
0
.6308
7
.24
E+03
-
0
.80
3
6
2
4
.35
9
7
6
8
0
.6311
7
.29
E+03
Std
2
.8E
-
05
0
.00
6
7
1
.88
E
-
04
3
3
.79
3
9
5
.5E
-
06
0
.01
5
4
2
.34
E
-
04
6
9
.33
7
8
Table
6
prese
nts
the
com
par
iso
n
of
s
olut
ion
s
obta
ine
d
by
the
pr
op
os
e
d
al
gorith
m
and
oth
er
m
et
aheu
risti
cs,
i.e.
Har
m
on
y
Searc
h
with
two
sta
ge
pe
na
lt
y
f
un
ct
io
n
(
HS
)
[
13]
,
Fire
fly
Algorithm
with
com
bin
at
ion
of
sta
ti
c
pen
al
ty
and
feasi
bili
ty
ru
le
s
(F
A
)
[14],
Co
hort
I
ntell
igence
(C
I
)
with
sta
ti
c
pe
nalty
(S
CI
)
an
d
dynam
ic
pen
al
ty
(D
CI)
[
16
]
,
Dif
fer
e
ntial
Searc
h
with
sta
ti
c
pen
al
ty
(S
D
S)
a
nd
dynam
ic
pen
al
ty
(D
DS)
[
17]
an
d
Musica
l
Co
m
po
sit
ion
Me
thod
(MCM
)
[
18
]
.
T
he
pro
pose
d
al
gorithm
ob
ta
i
ns
the
s
m
al
le
st
values
of
best,
m
ean,
w
or
st
and
sta
nda
rd
dev
ia
ti
on
valu
es
in
t
his
c
omparis
on
on
1
,
3
,
4
,
6
,
9
and
12
.
It
m
eans
that
LaF
is
bette
r
and
m
or
e
co
nsi
ste
nt
or
sta
bl
e
than
the
ot
her
m
et
aheu
risti
cs
in
s
olv
in
g
these
pro
blem
s.
In
t
he
oth
e
r
pro
bl
e
m
s
(
2
,
5
,
7
,
10
and
13
),
e
xcep
t
8
an
d
11
,
LaF
is
sti
ll
not
c
om
petitive
com
par
ed
to
the
oth
e
r
m
e
tah
eu
risti
cs,
sin
ce
it
has
diff
ic
ulti
es
in
so
lv
ing
hi
gh
dim
ensio
nal
op
ti
m
i
zat
ion
pro
blem
with
non
-
li
near
c
onstrai
nts
a
nd
a
ny
pro
blem
wh
i
ch
has
m
or
e
t
ha
n
one
e
qual
it
y
co
ns
trai
nt.
In
8
an
d
11
,
LaF
obta
ins
t
he
kn
own
op
ti
m
al
so
luti
on
s,
bu
t
S
DS
on
8
,
FA
a
nd
MC
M
on
11
ap
par
e
ntly
ob
ta
in
bette
r
so
luti
ons
t
han
the
op
ti
m
al
so
luti
ons
t
hat
ha
ve
been
know
n
s
o
far
.
Howe
ver,
in
ov
e
rall
,
LaF
is
m
ore
com
petit
ive than
the
o
t
her m
et
aheurist
ic
s.
5.
CONCL
US
I
O
N
Ba
sed
on
t
he
r
esult
an
d
analy
sis,
it
is
con
cl
uded
t
hat
Lea
de
rs
an
d
F
ollo
we
r
s
(La
F)
al
gori
thm
can
be
i
m
ple
m
ented
to
so
l
ve
co
ns
tra
ined
non
-
li
nea
r
opti
m
iz
at
ion
pro
blem
s.
W
it
h
sm
al
l
popu
la
ti
on
siz
e,
i.e.
≤
10,
LaF
co
ns
ist
en
tl
y
and
su
cce
ssfu
ll
y
fin
d
t
he
opti
m
a
l
so
luti
on
of
lo
w
dim
ension
al
(
≤
5
)
op
ti
m
iz
a
tio
n
pro
blem
s
with
ine
qu
al
it
y
co
nst
raints
on
ly
a
nd
the
lo
w
di
m
ension
al
(
≤
2
)
pro
blem
with
on
ly
one
e
qu
al
it
y
const
raint.
W
it
h
popula
ti
on
siz
e,
i.e.
≥
100,
La
F
can
opti
m
all
y
so
lve
any h
ig
h
dim
ension
al
co
nst
raine
d
no
n
-
li
near
op
ti
m
izati
on
pro
blem
that
ha
s
high
nu
m
ber
of
li
near
i
nequa
li
ty
con
strai
nt
s
an
d
no
non
-
li
near
const
raint.
La
F
has
di
ff
ic
ult
y
in
so
l
ving
hig
h
dim
ension
a
l
op
ti
m
iz
a
ti
on
pro
blem
with
non
-
li
near
c
on
strai
nts
and any
pro
ble
m
w
hich
ha
s m
or
e
tha
n on
e
equali
ty
co
ns
t
raint.
In
the
c
om
par
ison
with
oth
e
r
m
e
ta
heu
risti
cs,
LaF
has
be
tt
er
per
f
orm
a
nce
in
overall
ben
c
hm
ark
pro
blem
s.
It
shou
l
d
al
so
be
note
d
t
hat
the
c
on
st
raint
-
ha
nd
l
ing
m
et
ho
d
use
d
in
the
pro
posed
al
go
rithm
i
s
only
the
cl
assic
al
sta
ti
c
pen
al
ty
f
unct
ion
a
nd
the
LaF
al
gorit
hm
us
e
d
in
this
st
udy
is
the
ori
gi
na
l
on
e
.
It
m
ean
s
that
there
is
a
big
possibil
it
y
to
us
e
s
om
e
bette
r
co
ns
trai
nt
-
handlin
g
m
et
ho
d
or
to
m
od
i
fy
the
ori
gi
nal
LaF
al
gorithm
in
order t
o o
btain m
uch b
et
te
r
p
e
rfor
m
ance in
t
he
fur
the
r
st
ud
ie
s
.
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.
13
, N
o.
1
,
Ja
nu
a
ry 20
19
:
162
–
169
168
Table
6.
C
om
par
iso
n of S
olu
t
ion
s
Obtai
ne
d by the
Pro
po
se
d Alg
or
it
hm
an
d othe
r
Me
ta
he
ur
ist
ic
s
LaF
HS [
1
3
]
FA [
1
4
]
SCI [
1
6
]
DCI [
1
6
]
SDS [
1
7
]
DDS [
1
7
]
MCM
[
1
8
]
1
Bes
t
-
15
-
1
4
.99
9
NA
-
1
4
.99
7
-
15
-
15
-
15
-
15
Mean
-
15
-
1
4
.95
9
NA
NA
-
1
4
.9
-
1
4
.8
-
1
2
.3
NA
W
o
rst
-
15
-
1
4
.89
3
NA
NA
-
13
-
6
-
13
NA
Std
7
.25
E
-
16
0
.02
2
9
NA
0
.19
8
2
4
.5E
-
01
2
.55
6
7
0
.01
8
1
0
.14
7
3
2
Bes
t
-
0
.80
3
6
-
0
.72
5
5
NA
-
0
.80
3
6
-
0
.80
3
6
-
0
.80
3
6
-
0
.80
3
5
-
0
.80
3
6
Mean
-
0
.78
9
7
-
0
.70
0
9
NA
NA
-
0
.78
6
4
-
0
.79
2
0
-
0
.78
8
0
NA
W
o
rst
-
0
.76
9
2
-
0
.65
4
3
NA
NA
-
0
.73
9
5
-
0
.77
2
9
-
0
.77
4
3
NA
Std
0
.00
9
0
.03
9
7
NA
0
.03
6
1
0
.02
0.
0009
0
.00
0
7
0
.02
5
3
3
Bes
t
-
1
.00
0
2
-
1
.00
0
0
NA
-
1
.00
1
3
-
0
.99
9
9
NA
NA
-
0
.99
9
7
Mean
-
1
.00
0
2
-
0
.98
8
NA
NA
-
0
.98
3
9
NA
NA
NA
W
o
rst
-
1
.00
0
2
-
0
.95
1
NA
NA
-
0
.73
9
5
NA
NA
NA
Std
2
.00
E
-
06
0
.01
3
7
NA
0
.00
1
1
5
.0E
-
02
NA
NA
0
.00
0
8
4
Bes
t
-
30666
-
30665
-
306
65
-
30666
-
30665
-
30666
-
30666
-
30666
Mean
-
30666
-
30582
-
30665
NA
-
30665
-
30662
-
30666
NA
W
o
rst
-
30666
-
30405
-
30664
NA
-
30665
-
30599
-
30666
NA
Std
6
.00
E
-
12
2
4
.25
6
7
0
.47
5
5
0
.04
5
4
.9E
-
03
1
.49
6
8
0
.12
0
4
1
6
.17
5
5
Bes
t
5
1
2
8
.6
5
1
1
2
.3
NA
5
1
1
9
.1
4232.
6
5
1
3
1
.3
5
1
3
1
.3
5
1
2
1
.2
Mean
5
2
5
3
.1
5
1
1
5
.2
NA
NA
4
8
9
6
.6
5
5
5
7
.3
5
7
4
5
.1
NA
W
o
rst
5
6
7
3
.6
5
1
2
5
.3
NA
NA
5
6
1
2
.5
6
1
1
2
.2
6
1
1
2
.2
NA
Std
1
6
4
.81
1
.25
NA
4
0
.42
3
.9E+0
2
4
3
.56
4
0
.64
4
2
.19
6
Bes
t
-
6
9
6
1
.8
-
6
9
6
1
.6
-
6
9
6
0
.5
-
6
9
6
1
.8
-
6
9
6
1
.8
-
6
9
6
1
.8
-
6
9
6
1
.8
-
6
9
6
1
.8
Mean
-
6
9
6
1
.8
-
6
9
6
1
.3
-
6
9
5
6
.6
NA
-
6
9
6
1
.8
5
.9E+1
3
1
.8E+6
NA
W
o
rst
-
6
9
6
1
.8
-
6
9
6
0
.9
-
6
9
5
3
.5
NA
-
6
9
6
1
.8
2
.4E+1
5
3
.6E+7
NA
Std
0
0
.24
4
3
2
.19
2
8
1
.5E
-
05
0
1
.1E+1
4
8
.1E+6
3
.8E
-
07
7
Bes
t
2
4
.30
8
4
2
4
.55
2
2
4
.38
0
5
2
4
.30
4
4
2
4
.32
8
1
2
4
.33
0
2
2
4
.31
5
2
4
.35
0
6
Mea
n
2
4
.45
4
9
2
7
.61
2
2
4
.47
0
5
NA
2
4
.46
7
7
2
4
.34
1
2
4
.71
5
3
NA
W
o
rst
2
4
.82
4
2
3
1
.23
1
2
4
.60
2
4
NA
2
4
.98
7
2
5
.51
6
9
2
5
.53
3
6
NA
Std
0
.13
0
3
1
.65
4
5
0
.05
9
7
0
.22
1
6
0
.18
0
.03
8
2
0
.03
0
6
0
.21
3
5
8
Bes
t
-
0
.09
5
8
-
0
.09
5
8
-
0
.09
5
8
-
0
.09
5
8
-
0
.09
5
8
-
0
.09
5
9
-
0
.09
5
8
-
0
.09
5
8
Mea
n
-
0
.09
5
8
-
0
.08
0
7
-
0
.09
5
8
NA
-
0
.09
5
8
-
0
.09
5
9
-
0
.09
5
8
NA
W
o
rst
-
0
.09
5
8
-
0
.07
6
1
-
0
.09
5
8
NA
-
0
.09
5
8
-
0
.09
5
9
-
0
.09
5
8
NA
Std
4
.01
E
-
18
0
.01
3
6
2
.88
E
-
06
-
1
.1E
-
12
2
.1E−1
2
0
0
6
.2E
-
08
9
Bes
t
6
8
0
.6301
6
8
0
.656
6
8
0
.8463
6
8
0
.6726
6
8
4
.1806
6
8
0
.63
6
8
0
.63
6
8
0
.673
8
Mean
6
8
0
.6307
6
8
0
.742
6
8
1
.0415
NA
6
8
4
.1996
6
8
0
.7093
6
8
0
.7132
NA
W
o
rst
6
8
0
.6321
6
8
0
.779
6
8
1
.2603
NA
6
8
4
.2519
6
8
0
.9682
6
8
1
.1324
NA
Std
4
.59
E
-
04
0
.07
2
5
0
.15
3
3
6
0
.25
9
8
1
.66
E
-
02
0
.00
8
2
0
.00
1
1
0
.28
8
2
10
Bes
t
7
0
8
1
.5
7
0
8
2
.6
NA
7
0
5
1
.8
8
6
4
8
.2
7
0
5
8
.1
9
7
0
5
6
.7
6
7
0
5
1
.9
Mean
7
2
3
9
.8
7
1
1
0
.2
NA
NA
9
2
8
6
.5
7
2
9
7
.5
9
5
7
3
5
0
.3
5
NA
W
o
rst
7469
7
1
1
0
.3
NA
NA
1
1
7
4
5
.2
7
6
2
1
.0
0
5
7
8
4
6
.7
9
NA
Std
1
1
1
.7
2
.08
5
4
NA
1
1
.55
8
6
8
.6E+0
2
1
6
.58
1
2
0
.04
2
1
5
.38
8
1
11
Bes
t
0
.74
9
9
0
.74
9
0
.74
9
0
0
.74
9
7
0
.75
0
1
-
0
.74
9
9
-
0
.74
9
9
0
.74
8
9
M
ean
0
.74
9
9
0
.74
9
0
.74
9
0
NA
0
.77
9
8
-
0
.84
5
7
-
0
.77
3
1
NA
W
o
rst
0
.74
9
9
0
.74
9
0
.74
9
0
NA
0
.88
0
1
-
1
-
1
NA
Std
6
.50
E
-
08
3
.0E
-
06
3
.42
E
-
06
0
.00
1
3
3
.5E
-
02
0
.01
1
6
0
.00
6
0
.00
1
1
12
Bes
t
-
1
-
0
.99
0
9
-
0
.99
9
9
5
-
1
-
1
-
1
-
1
-
1
Mean
-
1
-
0
.95
2
5
-
0
.99
9
9
5
NA
-
1
-
1
-
1
N
A
W
o
rst
-
1
-
0
.89
1
3
-
0
.99
9
9
5
NA
-
1
-
1
-
1
NA
Std
0
0
.08
8
8
7
.78
E
-
07
1
.6E
-
12
1
.4E−0
9
0
0
0
.00
2
5
13
Bes
t
0
.17
8
9
0
.05
7
1
NA
NA
NA
NA
NA
NA
Mean
1
.07
8
5
0
.05
9
5
NA
NA
NA
NA
NA
NA
W
o
rst
4
.95
1
1
0
.07
0
3
NA
NA
NA
NA
NA
NA
Std
1
.51
9
4
0
.07
2
6
NA
NA
NA
NA
NA
NA
NA =
No
t Availab
le
Std
=
Stan
d
ard De
v
iatio
n
REFERE
NCE
S
[
1
]
Bashat
h
S,
Ismai
l
AR.
Com
par
ison
of
Sw
ar
m
Inte
ll
ig
ence
Algorit
hm
s
for
High
Dim
ensio
nal
Optimizati
o
n
Problems
.
Indone
sian J
ournal
of
El
e
ct
rica
l
Eng
in
ee
ring a
nd
Computer
Sc
ie
nc
e
.
2
018;
11(1):
300
-
307.
[
2
]
Abd
-
El
-
W
ahe
da
W
F,
Mous
a
AA
,
El
-
Shorba
g
y
MA
.
Int
egr
at
ing
Part
icle
Sw
arm
Optimiza
ti
on
with
Gene
tic
Algorit
hm
s
for
Solving
Nonlinear
Optimization
Problems
.
Journal
of
Computat
ional
and
Appli
ed
Mathe
mat
ic
s
.
2011;
235:
1446
-
1453.
[
3
]
Griva
I,
Nash
SG
,
Ariel
a
S
.
Li
ne
ar
and
No
nli
ne
ar
Optimiz
at
ion
.
Second
Edi
ti
on
.
Phil
adelphia
:
Societ
y
f
or
Industria
l
Applied Mat
hemat
ic
s.
2009:
14.
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
Leaders
and f
ol
lo
we
rs a
l
go
rit
hm for c
onstr
ai
ned
non
-
li
ne
ar o
ptimizati
on
(
Helen Yuli
ana An
gmalisa
ng
)
169
[
4
]
Na
y
ak
V,
Suth
a
r
H,
Gadi
t
J.
Im
ple
m
ent
a
ti
on
of
Artifi
c
ia
l
Bee
C
olon
y
Algorit
hm
.
IA
ES
Int
ernational
Journal
o
f
Arti
ficial Int
el
l
ig
enc
e
. 2
012
;
1(3)
:
112
-
120.
[
5
]
Garg
H.
A
Hy
b
rid
PS
O
-
G
A
Al
gorit
hm
for
Constrai
ned
Optim
iz
a
ti
on.
Appl
i
ed
Mathe
matic
s
a
nd
Computati
on
.
2016;
274:
292
-
305.
[
6
]
Raj
purohit
J,
Sharm
a
TK,
Abra
ham
A,
Vaisha
li.
Gloss
ar
y
of
Meta
h
eur
isti
c
Alg
orit
hm
.
Int
ernational
Journal
o
f
Compute
r Inf
or
mation
Syst
ems and
Industrial
M
anageme
nt Appl
i
cat
ions.
2017;
9:
181
-
205.
[
7
]
W
ei
dong
J,
Jun
Z.
PS
O
Algorit
hm
Based
on
Acc
um
ula
ti
on
E
ffe
ct
and
Muta
t
ion.
TEL
KOMNIKA
Indone
sian
Journal
of
Elec
t
rical
Engi
ne
erin
g
.
2013;
11(12):
7344
-
7350.
[
8
]
Anam
K,
Al
-
Ju
m
ai
l
y
A
.
Opti
m
iz
ed
Kerne
l
E
xtre
m
e
Learni
n
g
Mac
hine
for
M
y
oelec
tri
c
Pa
t
te
rn
Re
cogni
t
io
n.
Inte
rnational
Jo
urnal
of El
e
ct
ri
c
al
and
Comput
er
Engi
n
ee
ring
.
20
18;
8(1):
483
-
49
6.
[
9
]
Gonza
lez
-
Ferna
ndez
YG
,
Ch
en
S.
Leaders
and
Fol
lowe
rs
–
A
N
ew
Me
tah
euristic
to
Av
oid
the
Bia
s
of
Accumulat
ed
Information.
I
E
EE
Congr
ess on Evolut
ion
ar
y
Co
m
puta
ti
on.
Send
ai
.
2015:
776
-
78
3.
[
1
0
]
Li
u
J,
Cai
Z,
L
iu
J.
Premature
Conve
rgenc
e
i
n
Gene
tic
Al
gor
it
hm:
Anal
ysis
and
Prev
en
ti
on
Based
on
Chaos
Operator.
Proce
edi
ngs of
3r
d
W
orld
Congress o
n
Intelli
g
e
nt
Con
trol
and
Autom
a
ti
on.
Hefe
i. 2000
:
495
–
499.
[
1
1
]
Kaur
A,
Kaur
M
.
Dea
li
ng
with
B
oundar
y
Constr
a
int
Viola
t
ions
in
Parti
cle
Sw
arm
Optimiza
ti
o
n
with
Aging
Le
ader
and
Cha
ll
eng
ers (A
LC
-
PS
O).
Inte
rnational
Journ
al
of
Computer
Appl
ic
a
ti
ons.
20
15;
121(11):
13
-
19.
[
1
2
]
M
ez
ura
-
Montes
E,
Coello
C
oel
lo
CA.
Co
nstrai
nt
-
h
andl
in
g
in
Na
ture
-
In
spired
Num
eri
c
al
Opt
imiza
t
ion
:
Past,
Present
and
Future.
Swarm
and
Ev
o
lut
ionar
y
Computat
ion.
2011;
1(4):
173
-
194.
[
1
3
]
Zha
ng
B,
Duan
J,
Sang
H,
Li
J,
Yan
H.
A
Ne
w
Pe
nalty
Fun
c
ti
on
Me
thod
for
Con
strained
Optimizati
on
Us
in
g
Har
mony
Searc
h
Al
gorithm
.
I
EEE
Congress on
E
volut
iona
r
y
Com
puta
ti
on
.
B
ei
j
ing
.
2014:
853
-
859.
[
1
4
]
Deshpande
AM
,
Phatna
n
i
GM
.
Constr
aint
Ha
ndli
ng
in
F
iref
l
y
A
lgorit
hm
.
IE
EE
In
te
rna
ti
on
a
l
Confer
ence
o
n
C
y
ber
n
etics
(CY
BCO).
La
us
anne.
20
13;
186
-
190.
[
1
5
]
Niu
B,
W
ang
J,
W
ang
H.
Bac
t
eri
a
l
-
inspir
ed
Algorit
hm
s
for
Solving
Constrai
n
ed
Optimiz
at
ion
Probl
ems
.
Neurocomputi
ng
.
2015;
148:
54
-
62.
[
1
6
]
Kulkar
ni
O,
Kul
kar
ni
N,
Kulkar
ni
AJ
,
Kaka
ndik
ar
G.
Constrai
ne
d
Cohort
Inte
lli
genc
e
using
Sta
t
ic
and
D
y
n
amic
P
ena
lty
Func
ti
o
n
Approac
h
for
Mec
hanica
l
Co
m
ponent
s
Design.
Inte
rnat
ional
Journal
of
Paral
le
l
,
Eme
rgen
t
an
d
Distribute
d
S
yst
ems.
2016.
DO
I:
htt
ps:/
/doi.org/
1
0.
1080/1744576
0.
2016.
1242728
[
1
7
]
Li
u
J,
Te
o
KL
,
W
ang
X,
W
u
C.
An
Exa
ct
Penalt
y
Funct
ion
-
base
d
Diffe
ren
ti
al
S
ea
rch
Algor
it
hm
for
Constrai
ne
d
Global
Opt
imization.
Soft Computing
.
2016;
20(4)
:
1305
-
1313.
[
1
8
]
Mora
-
Gutié
rre
z
RA,
Ramírez
-
Rodrígue
z
J,
R
inc
ón
-
Garc
ía
E
A,
Pons
ic
h
A,
Herre
ra
O,
L
ara
-
Vel
áz
qu
ez
P.
Adapta
t
ion
of
the
Mus
ic
al
Com
positi
on
Method
for
Solving
Cons
tra
in
ed
Optimization
Problems
.
S
oft
Com
puti
ng
.
2014;
18(10):
19
31
-
1948.
Evaluation Warning : The document was created with Spire.PDF for Python.