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.
9
, No
.
5
,
Octo
ber
201
9
, pp.
3772~3
778
IS
S
N: 20
88
-
8708
,
DOI: 10
.11
591/
ijece
.
v9
i
5
.
pp3772
-
37
78
3772
Journ
al h
om
e
page
:
http:
//
ia
es
core
.c
om/
journa
ls
/i
ndex.
ph
p/IJECE
Projectil
e
-
t
arget
s
earch
a
lgorithm:
a
s
tocha
stic
m
eta
heuristi
c
o
ptim
iza
tion
t
ec
h
niq
ue
Ayong
Hiendr
o
Depa
rtment
o
f
E
le
c
tri
c
al E
ngin
eering,
Ta
n
jungpur
a
Univer
si
t
y
,
Ind
onesia
Art
ic
le
In
f
o
ABSTR
A
CT
Art
ic
le
history:
Re
cei
ved
Ja
n 1
5
, 2
01
9
Re
vised
Ma
r 3
0
,
201
9
Accepte
d
Apr 9
, 2
01
9
Thi
s
pape
r
prop
oses
a
new
stocha
stic
m
etahe
uri
stic
opti
m
izati
o
n
al
gorit
h
m
which
is
b
ase
d
o
n
kine
m
atics
of
proje
c
ti
l
e
m
oti
o
n
and
ca
l
le
d
pro
j
ec
t
il
e
-
ta
rg
et
sea
rch
(PTS)
al
gorit
hm
.
The
PTS
al
gorit
hm
emplo
y
s
the
enve
lop
e
of
proje
c
ti
l
e
tr
ajec
t
or
y
to
f
ind
th
e
t
arg
et
in
the
sea
r
ch
spac
e.
It
has
2
t
y
p
es
of
cont
rol
p
ara
m
eters.
Th
e
f
irst
t
y
pe
is
set
to
give
th
e
poss
ib
il
ity
of
th
e
al
gorit
hm
to
acc
el
er
at
e
conve
rge
nce
proc
ess,
wh
il
e
th
e
othe
r
t
y
p
e
is
set
to
enha
nc
e
th
e
po
ss
ibi
li
t
y
to
g
en
era
t
e
new
be
tt
e
r
proje
c
ti
l
es
for
sea
rch
ing
proc
ess.
How
ev
er,
both
are
r
es
ponsible
to
f
ind
bet
t
er
fit
n
ess
val
ues
in
the
sea
rch
spa
ce.
In
orde
r
to
p
erf
or
m
it
s
ca
pab
il
i
t
y
to
deal
with
g
lo
bal
opt
imum
proble
m
s,
th
e
P
TS
al
gor
it
hm
is e
val
u
at
ed
on
six
well
-
known be
nchm
ark
s a
nd
the
ir
shif
te
d
fu
nct
ions
with
1
00
dimensions.
Optimiza
t
ion
result
s
hav
e
demons
tra
te
d
th
at
th
e
PTS
al
gor
it
m
offe
rs
ver
y
good
per
form
an
ce
s
and
it
i
s
ver
y
compet
it
iv
e
compare
d
to oth
er
m
et
ah
eur
ist
ic
al
gorit
hm
s.
Ke
yw
or
d
s
:
Algorithm
Global
opti
m
um
Me
t
aheu
risti
c
Op
ti
m
iz
ation
t
echn
i
qu
e
Pr
oject
il
e m
oti
on
Copyright
©
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
:
Ayo
ng H
ie
ndr
o,
Dep
a
rtem
ent o
f
Ele
ct
rical
E
nginee
rin
g,
Tan
j
un
gpur
a
Unive
rsity
,
Jen
der
al
Ah
m
ad Yani
Street
, Po
ntianak
,
West
Kalim
antan
, I
ndonesi
a.
Em
a
il
:
ay
on
g.h
ie
ndro@ee.
un
t
an.
ac.i
d
1.
INTROD
U
CTION
Op
ti
m
iz
ation
te
chn
iq
ues
,
es
pecial
ly
stoch
ast
ic
natur
e
-
i
ns
pi
red
m
et
ah
eur
ist
ic
al
gori
thm
s
hav
e
beco
m
e
the
im
portant
a
nd
popu
la
r
to
ols
to
de
al
with
c
om
pl
ex
high
dim
ension
al
global
op
tim
iz
at
ion
pro
blem
s
in
m
any
real
-
lif
e
ap
plica
ti
on
s
.
The
global
optim
iz
at
ion
prob
le
m
s
can
be
m
ultim
od
al
with
a
huge
num
ber
of
local
op
ti
m
a
a
nd
non
-
dif
fere
ntiable
w
hich
cannot
be
s
olv
ed
by
us
in
g
tradit
ion
al
nu
m
erical
op
tim
i
z
at
ion
m
et
ho
ds
[1
]
.
Ther
e
f
or
e,
m
a
ny
m
et
aheu
risti
c
op
ti
m
iz
at
io
n
al
gorithm
s
hav
e
bee
n
propose
d
to
al
l
eviat
e
the pr
ob
le
m
s
.
The
nat
ur
e
of
a
stochastic
m
et
aheu
risti
c
op
ti
m
iz
ation
al
gorithm
is
e
m
plo
yi
ng
the
rand
om
-
searc
h
m
echan
ism
to
visit
diff
e
re
nt
par
ts
of
t
he
se
arch
s
pace
a
nd
the
n
ap
proa
ches
as
cl
ose
as
p
os
sible
to
global
op
ti
m
u
m
po
int
[
2,
3].
Im
po
rta
nt
pr
ob
le
m
of
t
he
m
et
aheu
rist
ic
al
gorithm
s
i
s
ho
w
to
inc
rea
se
t
he
pro
ba
bili
ty
to
ov
e
rc
om
e the local o
ptim
a and
fin
d
the
b
et
te
r glo
bal opti
m
um
v
al
ue
.
In
t
his
pa
pe
r
,
a
new
m
et
aheu
risthic
al
go
rithm
is
intro
du
c
ed.
T
he
propo
sed
al
gorithm
is
based
on
kin
em
at
ic
s
of
pro
j
ect
il
e
m
oti
on.
T
he
pro
j
ec
ti
le
is
la
un
che
d
f
r
om
a
po
int
at
the
gro
und
le
vel
with
a
giv
e
n
velocit
y,
m
ov
ed
in
va
rio
us
directi
on
s
unde
r
a
un
if
orm
gr
av
it
y,
and
la
nded
on
a
ta
r
get
at
a
su
r
face.
How
ever,
the
pro
po
se
d
pro
j
ect
il
e
-
ta
rg
e
t
search
(P
T
S)
al
go
rithm
do
e
s
no
t
em
ph
asi
ze
on
the
pr
oject
il
e
trajectory
that
la
un
c
hed
from
an
an
gle.
The
al
go
rithm
te
nd
s
to
util
iz
e
the
trackin
g
of
the
en
velo
pe
of
proj
ect
il
e
traj
ect
o
ry
wh
ic
h
enclo
se
s
al
l
po
ssible
po
i
nts
in
the
search
s
pace.
The
en
velo
pe
of
pr
oj
ect
il
e
trajecto
ry
cou
l
d
reach
po
i
nts
in
the
s
earch
s
pace
w
hich
are
ou
t
of
reach
from
a
ny
pro
j
ect
il
es
m
ov
ing
with
a
ny
init
ia
l
po
ints
and
velocit
ie
s [
4
].
T
he
in
form
ation
of the
t
raj
ect
or
y co
uld
be
use
fu
ll
for
a pr
oject
il
e to h
it
a t
arg
et
fro
m
it
s s
ta
rting
po
i
nt.
I
n
t
he
pro
po
se
d
PTS
al
gorithm
,
the
be
nef
ic
ia
l
pr
op
e
rty
of
the
env
el
op
e
of
pro
j
ect
il
e
trajec
tory
is
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
Project
il
e
-
tar
ge
t searc
h alg
ori
thm: a stoc
has
ti
c m
et
ah
e
ur
ist
ic
o
ptimiz
atio
n t
echn
i
qu
e
(
Ay
ong Hien
dro
)
3773
app
li
ed
t
o
fi
nd
a
ta
r
get
on
a
li
near
s
hap
e
su
r
face
.
Fu
t
he
rm
or
e,
the
P
TS
al
gorithm
has
tw
o
m
ai
n
con
t
ro
l
par
am
et
ers
in
order
to
acce
le
rate
conve
rg
e
nc
e
and
sea
rch
i
ng
p
r
ocesses
a
nd
h
ence
,
it
can
fin
d
the
bette
r
global
op
ti
m
u
m
value
faster
than
a
ny
oth
er
opti
m
izati
on
al
gorith
m
s.
In
or
de
r
to
exam
ine
the
gen
eral
pe
rfo
rm
ance
of
the
pr
op
os
ed
al
gorithm
,
it
i
s
te
ste
d
on
six
be
nc
hm
ark
s
and
t
heir
s
hifted
f
unct
ions.
The
pe
rfo
rm
ance
i
s
com
par
ed
t
o other al
gorithm
s’
res
ults w
hich
hav
e
b
ee
n re
ported
i
n
[
5
-
10
].
2.
PRO
JEC
TIL
E MO
DEL
The
pr
oj
ect
il
e
is
def
i
ned
to b
e
la
un
c
hed
f
r
om
a
groun
d
le
vel
(h
=
0), w
it
h
a
n
init
ia
l
velocit
y
v,
an
d
at
an
an
gle
of
inc
li
nation
θ
m
ea
su
re
d
wit
h
res
pect
to
the
hor
iz
on
ta
l
as
sho
wn
in
Fig
ure
1.
The
path
f
un
ct
ion
of
the pr
oj
ect
il
e (
y) as a
f
un
ct
io
n of h
or
iz
on
ta
l
distance
(x)
is
sp
eci
fied
as
fo
l
lows
:
(
)
=
.
(
)
−
.
2
2
2
2
(
)
(1)
wh
e
re
:
g
=
9.8
1
m
/s
2
The
siz
e
an
d
s
hap
e
of
t
he
proj
ect
il
e
tra
j
ect
or
ie
s
va
ry
acc
ordin
g
t
o
the
la
un
c
h
a
ngle
s
at
an
i
niti
al
velocit
y
v
a
bove
t
he
horizo
ntal,
as
seen
in
Fi
gure
2.
T
hese
t
raj
ect
or
i
es
ha
ve
a
n
en
velo
pe
of
project
il
e
trajecto
ry. T
he e
nv
el
op
e
of pr
oj
ect
il
e traject
or
y (
φ
)
is a path
that encl
os
es
an
d
i
ntersects
al
l po
ssible
pro
j
ect
il
e
paths
to fin
d
it
s targ
et
onto
a
hill
. Th
e
shape
of the
h
il
l su
rfac
e is de
fine
d as t
he
im
pact f
un
ct
io
n
ψ
.
Figure
1.
The
pro
j
ect
il
e
m
ov
ing pat
h
Figure
2
.
The
e
nv
el
op
i
ng p
a
ra
bo
la
path
The
e
quat
ion f
or the e
nvel
op
e of
proj
ect
il
e t
raj
ect
or
y
[
4
,
11
]
is d
et
erm
ined
by
(
)
=
2
2
−
2
2
2
(2)
Fo
r
a conti
nuous im
pact fu
nc
ti
on
ψ(x)
on
0
≤
x
<
with
ψ
(0)
= 0,
the s
urface
of
ψ
a
nd
the p
at
h
of
φ
hav
e
ex
act
ly
one
po
i
nt of i
ntersecti
on. It
is noted t
hat the
re
exists a
un
i
qu
e
targ
et
for w
hich
ψ(x)
=
φ(
x
)
.
The
m
ai
n
go
al
of
the
pro
po
s
ed
PT
S
al
gorit
hm
is
to
m
ini
m
iz
e
the
diff
e
ren
ce
betw
een
ψ
a
nd
φ
i
n
order
to
ens
ure
a
proj
e
ct
il
e
re
ach
t
he
ta
r
get
pr
eci
sel
y.
A
f
unct
ion
γ
(
x
)
is
t
hen
im
ple
m
ent
ed
t
o
sea
rch
th
e
ta
rg
et
on the im
pact su
r
face a
nd it
is g
ive
n by
(
)
=
(
)
−
(
)
(3)
3.
PRO
JEC
TIL
E
-
TAR
GET
S
EAR
CH ALG
ORI
TH
M
The
init
ia
l
para
m
et
ers
of
t
he
PTS
al
go
rith
m
are
the
po
pula
ti
on
siz
e
N,
the
num
ber
of
var
ia
bles
D
,
the
lo
wer
boun
d
xm
in,
the
up
per
boun
d
xm
a
x,
a
nd
the
m
axi
m
u
m
it
erati
on
Im
ax.
The
l
ower
an
d
uppe
r
boun
ds
of v
a
riables a
r
e ex
pr
es
sed
b
y
=
[
1
2
…
]
,
=
[
1
2
…
]
(4)
The
init
ia
l
pr
oject
il
e
popu
la
ti
on
as
ca
nd
i
dat
e
so
luti
ons
is
r
andom
ly
gen
erated
by
assig
nin
g
rand
om
values
(
r
and
j
) wit
hin
[
0,
1] to
each b
oundary
as foll
ows:
,
(
=
1
)
=
+
.
(
−
)
,
=
1
,
2
,
…
,
;
=
1
,
2
,
…
,
(5)
v
θ
y
(
x
)
ψ
(
x
)
y
(
x
)
φ
(
x
)
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
S
N
:
2088
-
8708
In
t J
Ele
c &
C
om
p
En
g,
V
ol.
9
, N
o.
5
,
Oct
ober
20
19
:
3772
-
3778
3774
The
ta
r
get
po
pu
la
ti
on
is
ra
ndom
ly
created
by
per
tu
r
bin
g
a
ra
nd
om
l
y
sel
ect
ed
proj
ect
il
e
with
the d
i
ff
e
ren
ce
of the t
wo o
t
he
r
ra
ndom
ly
sel
ect
ed
pr
oj
ect
il
es. T
he
ta
r
get
popula
ti
on is
ge
ner
at
e
d by
,
(
)
=
1
,
(
)
+
0
.
5
(
2
,
(
)
−
3
,
(
)
)
,
=
1
,
2
,
…
,
;
=
1
,
2
,
…
,
(6)
wh
e
re
the
in
di
ces
a
1
,
a
2
,
an
d
a
3
{1,2,
…,
N
}
are
rand
oml
y
cho
sen
i
nteg
ers
an
d
m
us
t
be
diff
e
re
nt
fro
m
each
oth
e
r
a
nd all
ar
e also
dif
fer
e
nt
f
r
om
the b
ase
ind
e
x.
Evaluati
ng
the
fitness
value
s
of
the
pro
j
ect
il
es
and
the
ta
rget
s
are
carried
ou
t
by
us
i
ng
(
7)
and
(
8)
a
s
fo
ll
ows:
(
)
=
(
,
(
)
)
(7)
(
)
=
(
,
(
)
)
(8)
The best p
r
oje
ct
il
e
x
bestj
(
I
)
and it
s b
est
v
al
ue
f
best
(
I
)
are then
s
el
ect
ed
by co
m
par
i
ng
t
he
fitne
ss v
al
ue
s of
each
x
i,j
(
I
)
a
nd
t
i,j
(
I
)
, as
f
ollo
ws:
(
)
=
{
,
(
)
(
(
,
(
)
)
)
≤
(
(
,
(
)
)
)
,
(
)
ℎ
(9)
(
)
=
(
(
)
)
(10)
wh
e
re
I
=
1,2,…,
I
max
w
hich denote
s t
he
s
ubse
qu
e
nt
gen
e
r
at
ion
c
reated
f
or each
it
erati
on.
The
m
ai
n
pr
oc
ess
of
PT
S
optim
iz
at
ion
is
iterati
ng
the
pro
j
ect
il
e
in
or
de
r
to
reach
it
s
best
fitness
value. T
he pr
oj
ect
il
e search
m
od
el
f
or
c
onve
r
ging to
wa
rd
s
to
the
tar
get is a
s foll
ow
s:
for
i
=
1:
N
for
j
=
1:
D
If
r
and
<
0.5
,
(
+
1
)
=
1
,
(
)
−
(
(
,
(
)
)
′
(
,
(
)
)
)
(11)
Else
,
(
+
1
)
=
,
1
(
)
+
(
(
,
(
)
)
′
(
,
(
)
)
)
(12)
end
end
end
In
t
his
PS
A
optim
iz
at
ion
,
a
li
near
sh
a
pe
d
hill
with
a
slo
pe
of
m
is
us
e
d
as
the
im
p
act
functi
on.
The
im
pact fun
ct
ion
is
(
)
=
.
(13)
As
th
e im
pact fu
nctio
n has
b
e
en
s
pecified
, t
he
fun
ct
io
n
γ
(
x
) c
ou
l
d be
def
i
ne
d here
b
y
(
)
=
.
−
2
2
+
2
2
2
(14)
and
(
(
,
(
)
)
′
(
,
(
)
)
)
=
(
.
,
(
)
−
2
2
+
2
2
.
(
,
(
)
)
2
+
2
.
,
(
)
)
(15)
The dist
ance
bet
ween
pro
j
ect
il
e and tar
get at t
he
c
urren
t i
te
r
at
ion
is cal
c
ulate
d
as
fo
ll
ows:
,
(
)
=
[
(
(
,
(
)
)
−
(
,
(
)
)
)
]
1
/
(16)
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
Project
il
e
-
tar
ge
t searc
h alg
ori
thm: a stoc
has
ti
c m
et
ah
e
ur
ist
ic
o
ptimiz
atio
n t
echn
i
qu
e
(
Ay
ong Hien
dro
)
3775
Ther
e
a
re
2
pa
ram
et
ers
fo
r
con
t
ro
ll
in
g
the
distance
(
d
),
tho
se
a
re
r
and
s
,
wh
e
re
(
r
,
s
)
>
0.
Othe
r
c
on
tr
ol
pa
ram
et
er f
or (1
1) an
d (12
)
is
de
fine
d by
=
,
=
(
1
−
)
,
q
>
0
(17)
The
new
ta
rge
t
popula
ti
on
(
t
k,j
(
I
+1)
)
is
al
so
c
reated
in
he
re.
Finall
y,
the
be
st
pro
j
ect
il
e
fo
r
t
he
nex
t
it
erati
on
is
(
+
1
)
=
{
,
(
+
1
)
(
(
,
(
+
1
)
)
)
≤
(
(
,
(
+
1
)
)
)
,
(
+
1
)
ℎ
(18)
(
+
1
)
=
(
(
+
1
)
)
(19)
The
pr
ocesses
are
rep
eat
e
d
un
ti
l
f
best
(
I
+1)
meet
s
it
s
desired
accu
racy
le
ve
l
(
ɛ
)
or
the
it
erati
on
has
reache
d
I
m
ax
an
d resu
lt
s i
n
x
bestj
(
I
+1)
as the sati
sfied pr
oject
il
e land
i
ng onto
th
e target.
4.
RESU
LT
S
A
ND AN
ALYSIS
Ther
e
are
si
x
ben
c
hm
ark
s
a
nd
their
sh
i
fted
functi
ons
us
ed
to
eval
uate
pe
rfor
m
ance
of
the
propos
e
d
PTS
al
gorithm
.
The
m
a
the
m
a
ti
cal
fo
rm
ulati
on
of
the
be
nc
hm
ark
functi
ons
are
giv
e
n
in
Table
1.
In
order
t
o
ver
ify
the
perf
or
m
ance
of
th
e
propose
d
PT
S
al
gorithm
,
it
is
com
par
ed
to
ot
her
al
gorit
hm
s
wh
ic
h
ha
ve
bee
n
repor
te
d
in
[
5
-
10
]
.
T
o
ca
rr
y
ou
t
t
he
com
pari
so
n
of
al
gorit
hm
per
f
or
m
ance,
the
a
ppr
oac
h
usi
ng
is
t
o
c
om
par
e
the
accu
racies
for
a
fixe
d
num
ber
of
it
era
ti
on
s.
I
n
the
e
xp
e
rim
ent,
100
dim
ension
s
(
D
=
10
0)
,
30
searc
h
agen
ts
(
N
=
30
),
an
d
1000
it
erati
on
s
(
I
max
=
1000)
a
re
im
pl
e
m
ented
f
or
ea
ch
al
gorithm
.
The
sta
ti
sti
cal
resu
lt
s
after
30
in
de
pe
nd
e
nt
ex
per
i
m
ent
s
are
eval
uated.
T
he
m
e
an
an
d
sta
ndar
d
de
viati
on
(SD)
values
of
t
he
best
so
luti
ons
from
the last
it
erati
ons a
re
pu
t a
s th
e m
e
tric
s to
ass
ess the
pe
rform
ance of alg
or
it
h
m
s.
Table
1.
Be
nc
hm
ark
fu
nctio
ns
Fu
n
ctio
n
Fo
r
m
u
la
Test r
an
g
e
(x)
Sh
if
t
p
o
sitio
n
(
o
)
G
lo
b
al op
ti
m
u
m
x*
F
(x*
)
Sp
h
ere
1
(
x
)
=
∑
[
2
=
1
]
[
-
1
0
0
,
1
0
0
]
D
-
[
0
,
0
,
…,
0
]
0
Ro
sen
b
rock
2
(
x
)
=
∑
[
100
(
+
1
−
2
)
2
+
(
−
1
)
2
−
1
=
1
]
[
-
3
0
,
3
0
]
D
-
[
1
,
1
,
…,
1
]
0
Sch
wef
el
3
(
x
)
=
−
∑
[
sin
(
√
|
|
=
1
)]
[
-
5
0
0
,
5
0
0
]
D
-
[
s, s,
…,
s
]
S =
4
2
0
.96
8
7
4
6
-
4
1
8
9
8
.2
8
8
7
Ras
trigin
4
(
x
)
=
∑
[
2
=
1
−
10
cos
(
2
)
+
10
]
[
-
5
.12
,
5
.12
]
D
-
[
0
,
0
,
…,
0
]
0
Ack
ley
5
(
x
)
=
1
+
20
−
20
.
exp
(
−
1
5
√
1
∑
2
=
1
)
−
exp
(
1
∑
cos
(
2
=
1
)
)
[
-
3
2
,
3
2
]
D
-
[
0
,
0
,
…,
0
]
0
Griewa
n
k
6
(
x
)
=
1
+
1
4000
∑
2
=
1
−
∏
(
√
)
=
1
[
-
6
0
0
.
6
0
0
]
D
-
[
0
,
0
,
…,
0
]
0
Sh
if
ted
Sp
h
ere
7
(
x
)
=
∑
[
2
=
1
],
z
=
x
-
o
[
-
1
0
0
,
1
0
0
]
D
[
-
30,
-
3
0
,
…,
-
30]
[
-
3
0
,
-
3
0
,
…,
-
30]
0
Sh
if
ted
Ro
sen
b
rock
8
(
x
)
=
∑
[
100
(
+
1
−
2
)
2
+
(
−
1
)
2
−
1
=
1
],
z =
x
-
o
[
-
3
0
,
3
0
]
D
[
-
15,
-
1
5
,
…,
-
15]
[
-
1
4
,
-
1
4
,
…,
-
14]
0
Sh
if
ted
Sch
wef
el
9
(
x
)
=
−
∑
[
sin
(
√
|
|
=
1
)]
,
z
= x
-
o
[
-
5
0
0
,
5
0
0
]
D
[
-
3
0
0
,
-
3
0
0
,
…,
-
300]
[
s, s,
…,
s
]
S =
1
2
0
.96
8
7
4
6
-
4
1
8
9
8
.2
8
8
7
Sh
if
ted
Ras
trigin
10
(
x
)
=
∑
[
2
=
1
−
10
cos
(
2
)
+
10
],
z =
x
-
o
[
-
5
.12
,
5
.12
]
D
[
-
2
,
-
2
,
…
,
-
2]
[
-
2
,
-
2
,
…
,
-
2]
0
Sh
if
ted
Ack
ley
11
(
x
)
=
1
+
20
−
20
.
exp
(
−
1
5
√
1
∑
2
=
1
)
−
exp
(
1
∑
cos
(
2
=
1
)
)
,
z
=
x
-
o
[
-
3
2
,
3
2
]
D
[
-
5
,
-
5
,
…
,
-
5]
[
-
5
,
-
5
,
…
,
-
5]
0
Sh
if
ted
Griewa
n
k
12
(
x
)
=
1
+
1
4000
∑
2
=
1
−
∏
(
√
)
=
1
,
z =
x
-
o
[
-
60
0
.
6
0
0
]
D
[
-
4
0
0
,
-
4
0
0
,
…,
-
400]
(
-
4
0
0
,
-
4
0
0
,
…,
-
400)
0
The
pa
ram
et
er
s
nee
de
d
t
o
dr
i
ve
t
he
P
TS
al
gorithm
are:
gr
a
vitat
ion
al
acce
l
erati
on
(
g
)
,
i
niti
al
velocit
y
of
pro
j
ect
il
e
(
v
),
sl
op
e
of
ta
rget
’s
s
urface
(
m
),
q
,
r
,
an
d
s
.
Gr
a
vitat
ion
al
a
ccel
erati
on
is
a
co
ns
ta
nt
of
g
=
9.
81
.
The
init
ia
l
velocit
y
of
pro
j
e
ct
il
e,
the
slope
of
the
ta
r
get’s
surface
,
an
d
q
giv
e
im
po
rtant
co
ntribut
ion
s
to
the con
verge
nc
e proces
ses,
w
hile
r
a
nd
s
a
re
contr
olled the
searchi
ng pr
oc
ess of t
he
PT
S
al
gorithm
.
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
S
N
:
2088
-
8708
In
t J
Ele
c &
C
om
p
En
g,
V
ol.
9
, N
o.
5
,
Oct
ober
20
19
:
3772
-
3778
3776
To
facil
it
at
e
t
he
analy
s
is,
se
le
ct
ed
functi
ons
are
us
e
d
as
exp
e
rim
ent
ob
j
ect
s
.
Fig
ur
e
3
desc
ribes
the
influ
e
nce
of
v
on
c
onve
rgence
proce
ss
of
the
Ac
kley
f
un
ct
io
n
under
fixe
d
m
co
ndit
ion
,
wh
il
e
Fig
ure
4
is
it
s
per
f
orm
ance
with
var
ia
ti
ons
of
m
at
v
f
ixed
c
onditi
on
.
As
seen
in
Figu
re
3,
v
ha
s
a
str
ong
e
ffec
t
to
conve
rg
e
nce
s
peed
of
the
P
TS
al
gorithm
.
The
c
onverge
nc
e
process
run
s
relat
ive
slo
w
ly
wh
en
v
has
la
rg
e
values
. On
the o
the
r
ha
nd,
m
has
sli
gh
tl
y eff
ect
to
the co
nv
erg
e
nce speed
.
H
oweve
r,
the
final f
it
ness va
lues o
f
the searc
hi
ng
proces
s
var
y si
gnific
antly
w
it
h
bo
t
h
v
a
nd
m
a
s seen i
n Table
2
.
Othe
r
c
on
tr
ol
par
am
et
er
of
the
P
TS
al
gorithm
is
q
,
as
show
n
i
n
Fi
g
ure
5.
I
ncr
easi
ng
the
value
of
q
will
sp
ee
d
up
the
c
onve
r
ge
nce
process
of
the
PT
S
al
gorithm
.
This
pa
ram
et
er
al
so
giv
e
s
ve
ry
hi
g
h
con
t
rib
ution
t
o
the
best
fi
nal
f
it
ness
value
w
hich
c
ou
l
d
be
r
eached
by
the
al
gorithm
.
Table
3
pr
e
sents
th
e
best
final
fitnes
s
va
lue
of
S
ph
e
re
f
un
ct
io
n
acc
ord
ing
to
it
s
q
values.
H
ow
e
ver,
no
t
al
l
pro
ble
m
s
beh
ave
the
sam
e
way
as
Ack
le
y
and
S
ph
e
re
f
unct
ions.
F
or
ex
a
m
ple,
Schw
e
f
el
fu
nctio
n
wil
l
su
ff
e
r
from
s
ta
gn
at
io
n
co
nd
it
ion
and
lose
it
s
c
onve
rg
e
nce
t
o
the
global
optim
u
m
if
q
i
s
set
to
a
hi
gh
val
ue,
as
sh
ow
n
in
Fig
ure
6.
The
par
am
et
er
q
s
peed
s
up
c
onve
rg
e
nce
process
for
al
l
te
st
functi
ons,
but
Schwe
fel
functi
on
nee
ds
r
el
at
ive
sm
a
ll
sp
eed i
n order
to g
o
to
the
global o
pti
m
u
m
p
at
h.
Figure
3
.
Fit
ne
ss v
al
ue
c
urves
w
it
h va
riat
ions o
f
v
Figure
4
.
Fit
ne
ss v
al
ue
c
urves
w
it
h va
riat
ions o
f
m
Table
2
.
Influe
nce
of
v
a
nd
m
upon
the
b
e
st f
inal fit
ness
v
al
ue
Perf
o
r
m
an
ce of
the
Ackley
f
u
n
ctio
n
with
variatio
n
s o
f
m
at
v
m
= 1
0
v
0
.1
1
5
20
50
2
.66
E
-
15
5
.82
E
-
13
3
.76
E
-
12
4
.95
E
-
11
4
.13
E
-
09
v
=
0.1
m
0
.1
1
5
20
50
1
.48
E
-
13
3
.82
E
-
14
2
.66
E
-
15
2
.66
E
-
15
2
.66
E
-
15
Figure
5
.
Fit
ne
ss v
al
ue
c
urves
w
it
h va
riat
ions o
f
q
Figure
6
.
Fit
ne
ss v
al
ue
c
urves
w
it
h va
riat
ions o
f q
for
Sc
hwefel
f
un
ct
io
n
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
Project
il
e
-
tar
ge
t searc
h alg
ori
thm: a stoc
has
ti
c m
et
ah
e
ur
ist
ic
o
ptimiz
atio
n t
echn
i
qu
e
(
Ay
ong Hien
dro
)
3777
Table
3
.
Influe
nce
of
q
up
on the
best f
i
nal f
it
ness value
q
0
.1
1
5
10
20
30
40
50
0
.83
2
7
4
9
1
.37
E
-
08
1
.94
E
-
28
2
.85
E
-
47
3
.76
E
-
69
8
.18
E
-
90
1
.17
E
-
102
1
.14
E
-
128
It
has
bee
n
m
entione
d
t
hat
pa
ram
et
ers
v
,
m
,
an
d
q
in
flue
nc
e
to
the
co
nver
gen
ce
s
peed
an
d
the
cl
os
est
op
ti
m
u
m
valu
e
that
c
ou
l
d
be
reac
hed
by
t
he
P
TS
sea
rc
hi
ng
proces
ses.
Howe
ver,
the
conve
rg
e
nce
t
ow
a
r
ds
global o
pti
m
um
will
b
e fail
ed
if the
pro
j
ect
il
es are trapped i
nto
local o
ptim
a in the
searc
h
sp
a
ce. F
ur
t
he
rm
or
e,
the
proj
ect
il
es
gen
e
rated
by
it
erati
on
proces
ses
are
al
so
de
pende
nt
to
dist
ance
c
ontrol
pa
ram
et
ers:
r
and
s
.
These
param
eter
s
are
ve
ry
im
po
rtant
in
order
t
o
e
nh
a
nce
pro
bab
il
it
y
of
the
new
pro
j
e
ct
il
es
to
avo
i
d
from
local
opti
m
a.
Com
par
ison
s
of
al
gorithm
s’
perf
or
m
ances
are
s
umm
ariz
ed
in
Table
4
an
d
Ta
ble
5.
The
obta
in
resu
lt
s
f
or
benchm
ark
s
F
1
-
F
6
are
sh
ow
n
in
Table
4
.
As
su
m
m
arized
in
Table
4,
PTS
ou
tpe
rfo
rm
s
h
yb
rid
firef
ly
al
gorith
m
(H
FA),
vel
ocity
-
base
d
ar
ti
fici
al
bee
colon
y
al
gorith
m
(V
ABC),
a
lt
ern
at
ive
dif
f
eren
ti
al
evo
l
ution
al
gorithm
(A
DE)
and
opposit
io
n
-
base
d
m
agn
et
ic
op
tim
iz
at
i
on
al
gorithm
(F
MO
A),
excl
ud
i
ng
m
od
ifie
d
m
on
key
al
gorithm
(MM
A),
on
F
1
.
O
nce
aga
i
n,
P
TS
ha
s
th
e
best
pe
rform
ance
on
F
2
and
F
3
.
Investi
gatio
n
on
F
4
shows
tha
t
AD
E
is
the
be
st,
bu
t
PTS
a
nd
V
ABC
are
al
m
os
t
as
go
od.
On
f
unct
io
n
F
5
,
PTS
is
the
best,
but
MM
A,
FA
,
and
A
DE
pe
rfor
m
alm
os
t
as
well
as
P
TS.
T
he
c
omparati
ve
res
ults
al
s
o
dem
on
strat
e
th
at
PTS,
fire
fly
al
gorithm
(F
A),
an
d
A
DE
ha
ve
the
best
pe
rfor
m
ances
on
F
6.
Ex
per
im
ents
on
F
7
and
F
12
sho
w
that
PTS
c
ou
l
d
find
t
heir
best
global
op
ti
m
a
po
i
nts
as
seen
in
Table
5.
F
or
s
hifted
func
ti
on
s
F
7
-
F
12
, PTS
pe
rfor
m
s
m
uch
be
tt
er th
an
m
oth
-
flam
e o
pti
m
izati
on
alg
ori
thm
(
MF
O)
[
6].
Table
4
.
Algori
thm
p
erf
orm
ance m
e
tric
s f
or
basic f
unct
io
ns,
D
=
100
F
PTS
MM
A
[
6
]
HFA/FA
[
7
]
VABC
[
8
]
ADE
[
9
]
FMOA
[
1
0
]
Mean
SD
Mean
SD
Mean
SD
Mean
SD
Mean
SD
Mean
SD
F
1
9
.85
E
-
177
0
.00
E+00
0
.00
E+0
0
.00
E+00
2
.64
E
-
171
0
.00
E+00
1
.05
E
-
25
2
.34
E
-
25
6
.37
E
-
45
1
.12
E
-
44
1
.67
E
-
01
8
.00
E
-
02
F
2
2
.62
E
-
19
4
.77
E
-
19
-
0
.07
7
1
5
2
0
.16
1
8
3
1
.60
E
-
11
2
.62
E
-
11
8
.90
E+01
3
.46
E+01
9
.80
E+01
3
.21
E
-
03
F
3
-
3
9
2
1
2
.9
3
4
0
4
4
.1
3
-
-
12439
1
3
3
.24
-
-
-
6
.17
E+03
2
.42
E+03
F
4
2
.31
E
-
15
2
.39
E
-
15
-
3
.39
E
-
08
7
.29
E
-
09
3
.34
E
-
14
7
.47
E
-
14
0
.00
E+00
0
.00
E+00
8
.74
E
-
02
4
.93
E
-
02
F
5
4
.03
E
-
15
2
.07
E
-
15
4
.4E
-
15
0
.00
E+00
1
.25
E
-
14
3
.36
E
-
15
1
.50
E
-
05
3
.33
E
-
05
6
.21
E
-
15
0
.00
E+00
6
.58
E
-
02
1
.34
E
-
02
F
6
0
.00
E+00
0
.00
E+00
-
0
.00
E+0
0
.00
E+00
-
100
0
.00
E+00
0
.00
E
+0
0
0
.00
E+00
1
.15
E+00
7
.98
E
-
02
Table
5
. Alg
or
i
thm
p
erf
orm
ance m
e
tric
s f
or
s
hifted
f
un
ct
i
ons
,
D
=
100
Fu
n
ctio
n
PTS
MFO
[
1
1
]
Mean
SD
Mean
SD
F
7
0
.00
E+00
0
.00
E+00
0
.00
0
1
1
7
0
.00
0
1
5
F
8
3
.21
4
0
4
8
E
-
22
6
.74
0
4
7
7
E
-
22
1
3
9
.1487
1
2
0
.2607
F
9
-
4
2
6
0
0
.9
4
9
5
7
7
1473.
5
1
7
3
9
8
-
8
4
9
6
.78
7
2
5
.8737
F
10
1
.92
8
1
5
0
E
-
06
1
.22
2
0
5
9
E
-
06
8
4
.60
0
0
9
1
6
.16
6
5
8
F
11
1
.17
0
7
9
5
E
-
07
9
.66
8
9
5
7
E
-
08
1
.26
0
3
8
3
0
.72
9
5
6
F
12
0
.00
E+00
0
.00
E+00
0
.01
9
0
8
0
.02
1
7
3
2
5.
CONCL
US
I
O
N
Fr
om
exp
erim
ents
on
sel
ect
ed
well
-
know
n
be
nch
m
ark
s
and
their
s
hi
fted
f
un
ct
i
on
s
,
it
has
been
dem
on
strat
ed
that
PT
S
al
go
rithm
is
an
eff
ec
ti
ve
opti
m
iz
at
i
on
al
gorithm
to
deal
wit
h
hig
h
dim
ension
s
globa
l
op
ti
m
iz
ation
pro
blem
s.
It
is
al
so
pro
ve
n
to
be
a
ver
y
c
om
pet
it
ive
al
go
rithm
co
m
par
ed
to
ot
her
well
-
kn
own
m
et
aheu
risti
c a
lgorit
hm
s
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
S
N
:
2088
-
8708
In
t J
Ele
c &
C
om
p
En
g,
V
ol.
9
, N
o.
5
,
Oct
ober
20
19
:
3772
-
3778
3778
RE
FERE
NCE
S
[1]
Y.D.
Serge
y
e
v,
D.E
.
Kv
asov,
M
.
S.
Mukham
et
zh
anov,
“
On
Th
e
Eff
iciency
of
N
at
ure
-
Inspir
ed
Meta
heur
isti
cs
i
n
Expe
nsive
Glob
al
Opt
imiza
t
ion w
it
h
L
imite
d
Bu
dget
,
”
S
ci
en
ti
f
ic
Re
ports
,
vo
l.
8
,
p
p.
1
-
9,
2018.
[2]
C.
Blum
,
A
.
Rol
i,
“
Meta
h
eur
ist
ics
in
Com
bina
tor
ia
l
Optimization
:
Overvi
ew
and
Conce
ptu
al
Co
m
par
ison,”
AC
M
Computing
Surv
ey
s
,
vo
l. 35, pp.
268
–
308,
2003
.
[3]
H.
Maka
s,
N.
Yum
usak,
“
Bal
anc
ing
Expl
or
at
i
on
and
Expl
oita
ti
on
b
y
Us
ing
Sequent
i
al
Exec
uti
on
Cooperati
on
Bet
wee
n
Art
ifi
c
ia
l
Be
e
Colon
y
and
Migrat
ing
Bird
s
Optimiza
ti
on
Algori
thms
,
”
Tur
ki
sh
Journal
of
El
e
ct
ric
a
l
Engi
ne
ering
&
Computer
Scien
ce
s
,
vo
l. 24, pp. 4935
–
4956
,
20
16.
[4]
M.
Baće,
S.
Il
ijić,
Z.
Nara
n
cić,
L.
Bistri
ci
ć
,
“
The
Enve
lop
e
o
f
Projectile
Tr
ajec
tor
ie
s,
”
Europ
ean
Journal
of
Phy
sics
,
vol
.
23
,
pp.
6
37
-
642
,
20
02.
[5]
S.
Mirja
l
il
i
,
“
Moth
-
Flame
Opti
m
iz
at
ion
Algo
ri
thm:
A
Novel
Natur
e
-
Inspire
d
Heuri
stic
Pa
rad
i
gm
,
”
Knowle
dg
e
-
Based
Syst
ems
,
vol.
89
,
pp
.
228
-
249,
2015
.
[6]
R.
H.
Abi
y
ev
,
M.
Tunay
,
“
Ex
per
imental
Stud
y
of
Speci
f
ic
Benc
hm
ark
ing
Functi
ons
for
Modifie
d
Monk
e
y
Algorit
hm
,
”
Proce
dia
Computer
Sci
en
ce
,
vol
.
10
2,
pp
.
595
-
602,
2
016.
[7]
L.
Zha
ng
,
L.
L
i
u,
X.S.
Yang,
Y.
Dai,
“
A
Novel
Hy
br
id
Firefly
Algorit
hm
for
G
loba
l
Optimiz
at
i
on,
”
PLoS
One
,
vol.
11
,
e016323
0,
2016
.
[8]
N.
Im
ani
an,
M.
E.
Shiri
,
P.
Moradi,
“
Velocit
y
Based
Ar
ti
fi
ci
a
l
Bee
Colon
y
Algorit
hm
for
High
Dim
ensiona
l
Conti
nuous Opti
m
iz
at
ion
Proble
m
s,”
Engi
n
ee
rin
g
Applications o
f
Artifi
ci
al
Int
el
l
i
genc
e
,
vo
l. 36, p
p.
148
-
163,
2014
.
[9]
A.W.
Moham
ed,
H.Z.
Sabr
y
,
M.
Khors
hid,
“
An
al
te
rna
ti
v
e
Diffe
r
ent
i
al
Evol
uti
on
Algo
rit
hm
fo
r
Global
Optimiza
ti
o
n,
”
J
ournal
of Adv
an
ce
d
Re
search
,
v
ol.
3
,
pp
.
149
-
16
5,
2012
.
[10]
M.
Aziz,
M.H.
Tay
a
ran
i
-
N
,
“
Oppos
it
ion
-
Base
d
Magne
tic
Opt
imiza
ti
on
Algor
it
hm
with
Para
m
et
er
Adapt
at
io
n
Strat
eg
y
,
”
Swarm
and
Ev
o
lut
ion
ary
Computati
on
,
vol
.
26
,
pp
.
97
-
119,
2016
.
[11]
E
.
I.
Buti
kov
,
“
T
he
Envelope
of
Bal
li
st
ic
Tra
j
ect
orie
s
and
El
l
ipt
i
c
Orbits,
”
Ame
ri
can
Journal
o
f
Phy
sics
,
vo
l.
83,
pp.
952
-
958,
201
5.
Evaluation Warning : The document was created with Spire.PDF for Python.