Indonesi
an
Journa
l
of El
ect
ri
cal Engineer
ing
an
d
Comp
ut
er
Scie
nce
Vo
l.
12
,
No.
3
,
Decem
ber
201
8
, p
p.
1054
~
1062
IS
S
N: 25
02
-
4752, DO
I: 10
.11
591/ijeecs
.v1
2
.i
3
.pp
1054
-
1062
1054
Journ
al h
om
e
page
:
http:
//
ia
es
core.c
om/j
ourn
als/i
ndex.
ph
p/ij
eecs
Improve
d Chicke
n Swarm
Opti
mizati
on
Algo
rith
m to
S
olve
the Trav
elling Sa
lesman P
ro
bl
em
Fa
yça
l
C
he
bihi
, Mohamme
d
E
ssa
id
R
i
ff
i
,
A
mi
ne
A
ghar
ghor
, So
uk
ain
a
C
heri
f
B
ou
r
ki
S
eml
ali,
Ab
delf
att
ah
H
aily
Chouai
b
Doukka
li
Univ
ersity
,
M
or
o
c
c
o
Art
ic
le
In
f
o
ABSTR
A
CT
Art
ic
le
history:
Re
cei
ved
A
pr
22
, 201
8
Re
vised Ju
l
16
,
2018
Accepte
d
Aug
30
, 201
8
Thi
s
pape
r
pr
oposes
a
novel
discr
et
e
bi
o
-
inspire
d
chic
ken
sw
arm
opti
m
iz
ation
al
g
orit
hm
(CSO
)
to
solve
the
problem
of
the
tra
vel
in
g
sale
sm
an
proble
m
(TSP)
which
is
one
of
the
m
ost
know
n
proble
m
s
used
to
eva
luate
the
per
form
ance
of
the
new
m
et
ahe
urist
ic
s.
Thi
s
proble
m
is
solved
b
y
apply
ing
a
lo
ca
l
sea
rch
m
e
thod
2
-
opt
in
ord
er
to
improve
the
qu
al
ity
of
th
e
soluti
ons.
Th
e
DCS
O
as
a
sw
arm
s
y
stem
of
the
algorithm
inc
re
ase
s
the
le
v
e
l
of
dive
rsificat
io
n,
in
the
sam
e
wa
y
th
e
hie
r
archic
a
l
orde
r
of
t
he
chi
ck
en
sw
arm
and
the
beha
viors
of
chic
kens
inc
re
ase
th
e
le
ve
l
of
int
ens
ifi
c
at
ion
.
In
thi
s
cont
ribu
ti
on
,
we
red
efi
n
ed
th
e
basic
diff
ere
nt
oper
at
ors
and
op
era
t
ions
of
the
CS
O
a
lgorithm
.
The
per
fo
rm
anc
e
of
the
al
gor
it
hm
is
t
este
d
on
a
s
y
m
m
et
ric
TSP
benc
hm
ark
dataset
from
TSP
LIB
li
br
ar
y
.
Th
ere
fore
,
t
h
e
al
gorit
hm
provi
des
good
result
s
in
te
rm
s
of
bot
h
opti
m
iz
ation
a
cc
ura
c
y
and
robustness c
om
par
ing to
oth
er
m
et
ah
eur
isti
cs.
Ke
yw
or
ds:
2
-
OP
T
C
SO
DCSO
NP
-
H
ARD
TSP
TSPL
IB
Copyright
©
201
8
Instit
ut
e
o
f Ad
vanc
ed
Engi
n
ee
r
ing
and
S
cienc
e
.
Al
l
rights re
serv
ed.
Corres
pond
in
g
Aut
h
or
:
Fayç
al
Cheb
i
hi,
C
houaib
Do
ukkali U
niv
e
rsity
, Mor
o
c
c
o.
Em
a
il
:
cheb
ihi.
f@ucd
.ac.m
a
1.
INTROD
U
CTION
T
rav
el
in
g
sal
e
sm
an
pr
oble
m
(TSP
)
is
on
e
of
the
m
os
t
extensively
stud
ie
d
pro
blem
s
[1]
in
operati
onal
researc
h.
T
he
aim
of
the
prob
le
m
is
t
o
fin
d
th
e
shortest
pat
h
li
nkin
g
a
set
of
ci
ti
es
;
the
Sale
sm
an
sh
oul
d
cr
os
s
each
ci
ty
on
l
y
on
ce
a
nd
r
et
urn
to
the
c
it
y
of
de
par
tu
re
i
n
order
t
o
cl
os
e
the cyc
le
.
The
TS
P
is
NP
-
har
d
pro
blem
and
it
s
c
om
plexity
increa
se
de
pends
on
the
nu
m
ber
of
ci
ti
es
incl
ud
e
d.
If
we
c
onside
r
i
s the
num
ber
of cit
ie
s,
the
n
t
he
num
ber
of th
e possible
so
l
ut
ion
s is
:
(
1
)
!
2
n
(1)
T
akin
g
int
o
c
onside
rati
on
tha
t
a
com
pu
te
r
f
ind
s
a
possible
so
l
ution
in
1µ
s
tim
e
proces
s.
T
he
t
otal
tim
e to o
btain
al
l po
ssi
ble s
olu
ti
ons
is
descr
i
bed b
y t
he foll
ow
i
n
g
T
able
1
.
Table
1.
T
he
Com
pu
ta
ti
on
Ti
m
e Estim
at
ed
Nu
m
b
e
r
o
f
Cities
Nu
m
b
e
t of
T
o
u
rs
Ti
m
e
in Yea
r
20
6
,08
E+16
2
25
3
,1E+2
3
9
,84
E+6
30
4
,4E+3
0
1
,4E+1
4
35
1
,48
E+38
4
,68
E+21
40
1
,02
E+46
3
,23
E+29
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
Impr
oved Chic
ken
Swa
r
m O
pt
imizati
on Alg
ori
thm
t
o
So
lv
e
t
he
Tr
avell
in
g S
alesm
an…
(
Fa
yçal Che
bih
i
)
1055
This
ki
nd
of
pro
blem
is
ver
y
com
m
on
in
industry
busine
ss
[
2],
[
3
]
su
c
h
as
X
-
ray
cry
sta
ll
og
ra
ph
y
wh
il
e
analy
zi
ng
c
rysta
ls
struct
ur
e
or
e
ve
n
in
m
or
e
r
eal
li
fe
sit
uat
ion
s
lo
gisti
cs
[
4
]
su
c
h
as
school
trans
portat
ion.
The
di
f
fic
ulty
of
this
pro
bl
e
m
has
gen
e
rated
m
uch
interest
to
s
olv
e
TSP
,
sta
rti
ng
with
the
i
m
ple
m
ent
at
io
n
of
heurist
ic
s
m
e
tho
ds
suc
h
as
T
a
bu
Searc
h
(T
S)
[5
]
,
Gen
et
ic
Algorithm
(GA)
[6
]
,
Heurist
ic
App
ro
ac
h
[
7],
G
r
eedy
Ra
ndom
i
ze
Ad
a
ptive
Searc
h
Proce
dure
(
GR
AS
P
)
[
8],
an
d
Si
m
ula
te
d
Anneali
ng
(S
T
)
[9
]
.
Re
centl
y
,
se
ver
al
stu
dies
use
of
the
bio
-
insp
ire
d
al
gor
it
hm
s
us
in
g
swa
rm
intel
lig
ence
m
et
ho
ds
[10]
s
uch
a
s:
ant
col
on
ie
s
optim
iz
a
ti
on
(
ACO
)
[
1
1],
pa
rtic
l
e
sw
arm
op
tim
iz
a
tio
n
(P
S
O
)
[
12
]
,
[13]
bee
c
olonies
optim
iz
at
ion
(B
CO)
[14],
harm
on
y
search
a
lgorit
hm
(H
S)
[15],
[16],
ba
t
-
insp
i
red
al
go
rithm
(BA) [
17]
,
[
18]
,c
uc
koo sea
rc
h (CS)
[1
9]
,
[
20
]
an
d a
bio
-
ins
pire
d hunti
ng s
earch
alg
or
it
hm
(
HU
S) [
21
].
The
m
et
aheu
risti
c
Chicken
s
war
m
op
ti
m
iz
a
ti
on
(CS
O)
is
a
bio
-
ins
pire
d
beh
a
vior
of
c
hi
cken
.
It
was
introd
uced
i
n
2014
by
XIA
N
BI
NG
ME
NG
[22],
a
nd
prov
e
s
it
s
ef
fici
ency
to
so
lve
so
m
e
con
t
inu
e
d
op
ti
m
iz
ation
pro
blem
,
bu
t
it
is
no
t
possible
to
use
it
to
s
ol
ve
the
c
om
bin
at
ori
al
optim
i
zat
ion
prob
le
m
.
The
aim
of
this
pa
per
is
to
pro
pose
a
no
vel
a
dap
ta
ti
on
of
t
he
CSO
m
et
aheurist
ic
to
sol
ve
the
com
bin
at
ori
al
op
ti
m
iz
ation
pro
blem
s
by
r
edef
i
ning
opera
ti
on
s
an
d
op
e
r
at
or
s.
T
o
pro
ve
the
ef
fici
enc
y
of
the
pr
opos
e
d
adap
ta
ti
on,
th
e
ada
pted
CS
O
is
ap
plied
to
s
olv
e
s
om
e
be
nc
hm
ark
i
ns
ta
nces
of
th
e
tra
veling
sa
le
s
m
an
pro
blem
.
The
obta
ined
r
es
ults
are c
om
par
ed
t
o
the
b
e
st sol
ution
s
ex
it
in
g
in
the l
it
eratur
e
.
The
rest
of
t
he
pa
pe
r
is
or
ga
nized
as
f
ollo
ws:
t
he
sec
ond
sect
io
n
desc
ribes
the
T
ra
veling
Sale
sm
an
Pr
oble
m
,
the
thir
d
sect
io
n
int
rod
uces
the
C
hi
cken
S
wa
rm
Op
ti
m
iz
ation
Algorithm
,
the
forth
sect
io
n
pr
ese
nts
our
pro
posed
adap
ta
ti
on
of
t
he
CS
O
al
go
ri
thm
to
so
lve
TSP
,
t
he
fifth
sect
ion
sho
ws
the
e
xp
e
rim
ent
al
an
d
nu
m
erical
r
esu
lt
s,
an
d
fi
nally
the last
secti
on
is a c
on
cl
us
io
n.
2.
TRA
VELIN
G
SA
LE
S
M
AN
PROBLE
M
The
tra
veling
s
al
es
m
an
prob
le
m
[23]
was
fir
st
introdu
ce
d
by
the
Ital
ia
n
Ma
them
atici
an
Kar
l
Me
ngre
in
1930
as
a
giv
en
li
st
of
ci
ti
es
al
ong
with
th
e
cost
of
tra
vel
between
eac
h
pair
of
them
.
T
he
aim
of
this
stud
y
is
to
fin
d
t
he
s
hortest
pat
h,
w
hich
al
lo
we
d
vi
sit
ing
each
ci
t
y
on
ce
sta
rtin
g
an
d
en
ding
i
n
the
sam
e
ci
t
y.
Give
n
this
si
m
ple
form
ulati
on
,
it
m
i
gh
t
be
po
ss
ible
that
the
pro
blem
cou
ld
hav
e
a
n
e
qu
al
ly
si
m
ple
so
luti
on
.
It
app
ea
rs
that
this
is
no
t
the
case
even
if
th
e
pr
oble
m
is
e
asy
to
exp
res
s
and
inte
rpreted
.
To
the
pr
e
sen
t
day
there
is
no
ef
fici
ent
so
l
utio
n
to
t
he
TS
P
has
bee
n
f
ound.
H
owev
er,
the
pr
ob
le
m
has
ins
pire
d
s
ever
a
l
m
at
he
m
at
icians,
com
pu
te
r
s
ci
entist
s
and
a
host
of
non
-
pro
fessio
nal
researc
he
rs.
The
pro
blem
can
be
m
at
he
m
at
ic
a
lly
fo
rm
ulate
d
as:
Let
G
=
(V
, E,
W
)
be
a w
ei
gh
te
d
grap
h
with
V
=
{v
1,
v
2
…
vn},
the
n
th
e
TS
P
in G can
b
e
r
e
presente
d
as
:
Let
G
=
(V
,
E,
W
)
be
a
w
ei
gh
te
d
gr
a
ph
with
V
=
{v
1,
v2
…
vn},
t
hen
the
T
SP
in
G
can
be
represe
nted
as
;
m
ini
m
iz
e
∑
(
)
.
∈
subje
ct
to:
(
{
}
)
()
2,
2
,
,
{
0
,
1
}
,
i
ei
ev
e
eS
e
x
v
V
x
S
V
S
x
e
E
(2)
3.
CHICKE
N
S
WA
RM OPTI
MIZ
ATION
The
C
hicke
n
s
w
arm
op
ti
m
iz
a
ti
on
(C
SO)
off
ers
a
s
war
m
optim
iz
at
ion
base
d
on
t
he
c
hick
en’
s
nat
ur
al
beh
a
vior
i
n
a
s
war
m
.
A
hiera
rch
ic
al
order
is
est
ablishe
d
in
the
swa
rm
.
Th
e
chicke
ns
wit
h
the
hi
gh
est
fi
tness
values
a
re
ide
ntifie
d
as
roost
ers
an
d
th
os
e
with
the
worst
fitnes
s
values
are
chic
ks
.
Me
anwhil
e,
th
os
e
in
the
m
idd
le
are
he
ns
.
T
he
s
warm
is
div
ided
into
gro
up
s;
ea
ch
gr
oup
c
on
t
ai
ns
a
r
ooste
r,
a
co
up
le
of
he
ns
a
nd
chicks
and is c
reated
rand
om
l
y as desc
ribe
d
earli
er [
22]
.
The ro
os
te
r
w
it
h
the
b
e
st fit
ne
ss v
al
ue
ca
n
lo
ok f
or fo
od in a
w
ide
r ran
ge of
p
la
ces.
12
,,
(
1
(
0
,
)
)
tt
i
j
i
j
x
x
R
a
n
d
n
(3)
2
()
||
1,
1
,
,
,
ki
i
ij
ff
f
if
f
f
k
N
k
i
e
othe
rw
ise
(4)
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.
12
, N
o.
3
,
Dece
m
ber
2
01
8
:
1054
–
1062
1056
Hen
s
ca
n
ra
nd
om
l
y st
eal
g
oo
d foo
d
f
r
om
the o
the
r
c
hicke
ns.
1
,
,
1
,
,
2
,
,
1
(
)
2
(
)
t
t
t
t
i
j
i
j
r
j
i
j
tt
r
j
i
j
x
x
S
R
a
n
d
x
x
S
R
a
n
d
x
x
(5)
her
e
(
1
)
(
|
|
)
1
ir
i
ff
f
Se
(
6)
And
2
()
2
ri
ff
Se
(7)
Chicks l
ook f
or fo
od ar
ound t
heir
m
oth
ers
1
,
,
,
,
()
t
t
t
t
i
j
i
j
m
j
i
j
x
x
F
L
x
x
(8)
Finall
y, the al
gorithm
o
f
C
SO i
s r
e
pr
ese
nted
as foll
ow
:
Ch
ick
en
Swar
m
O
p
ti
m
i
zatio
n
A
lg
o
ri
th
m
Init
ialize a
po
p
u
latio
n
of
N
ch
ick
en
s
an
d
def
in
e the re
lated
para
m
e
ters
Evalu
ate the
N
ch
i
ck
en
s’ f
itn
ess
values,
t=0
W
h
ile (
t <
Max_
G
en
eration
)
If
(
t %
G
=
= 0)
Ran
k
the ch
ic
k
en
s’ f
itn
ess
valu
e
s
an
d
estab
lish
a
h
ierar
ch
ical
o
rder in
th
e swar
m
Div
id
e the sw
ar
m
in
to
dif
f
erent gro
u
p
s, and
determ
in
e the
relation
sh
ip
s b
etween ch
ick
s an
d
m
o
th
er
h
en
s
in
a
g
rou
p
End
I
F
Fo
r
i=1
:
N
If
i==
r
o
o
ster
Up
d
ate its so
lu
tio
n
/lo
catio
n
us
in
g
equation
(
3
)
End
If
If
i==
hen
s U
p
d
ate its so
lu
tio
n
/l
o
catio
n
us
in
g
equ
a
tio
n
(
5
)
End
If
If
i==
chick
s Up
d
ate its so
lu
tio
n
/lo
catio
n
us
in
g
equation
(
8
)
End
If
Evalu
ate the
n
ew so
lu
tio
n
s
If
the n
ew so
l
u
tio
n
is better than
its prev
io
u
s o
n
e up
d
ate it
End
f
o
r
End
W
h
ile
4.
A NO
VEL DI
S
C
RETE
C
H
I
CKEN SW
A
R
M
O
PTIMIZ
ATIO
N
TO
S
OLVE T
RAV
EL
ING
SA
LE
S
MAN
PROBLE
M
This
pa
per
proposes
a
no
ve
l
adap
ta
ti
on
of
the
Chicken
Sw
arm
Op
ti
m
i
zat
ion
(CS
O)
to
so
lve
th
e
Trav
el
in
g
Sale
sm
an
Pr
ob
le
m
.
This
adap
ta
ti
on
is
est
ablished
by
the
red
e
fi
niti
on
of
opera
tors
an
d
operat
ions
and
res
pects
the
ge
ne
ral
process
of
the
CSO
al
gorith
m
.
This
sect
i
on
desc
ribes
t
he
di
ff
e
ren
t
pro
pose
d
i
m
pr
ovem
ents o
f
operat
or an
d o
per
at
io
ns
.
4
.
1.
O
pera
t
or
Impr
ov
e
me
nt
The
po
sit
io
n
of
a
sel
ect
ed
chicken
i
is
the
so
l
ution
represe
nted
by
a
Ham
il
to
nians
cy
cl
e
of
the
ci
ti
es
=
{
1
,
2
,
…
,
}
.
=
[
1
,
2
,
…
,
]
ℎ
(
∈
[
1
,
…
,
]
)
4
.
2
.
Oper
at
i
on
s
Impr
ov
e
ment
4
.
2
.
1.
The
Chi
cken M
ovem
e
nt
Ther
e
a
re
3
ty
pes
of
chicke
ns
(h
e
ns
,
chic
ks
and
r
oo
ste
rs).
Each
on
e
has
a
deter
m
ined
process
of
m
ov
e
m
ent that can
be dete
rm
i
ned as
fo
ll
ows:
a)
Roo
ste
rs
R
oo
ste
rs
that
hav
e
bette
r
fi
tness
value
c
an
lo
ok
f
or
f
ood
i
n
a
la
r
ge
r
s
pace.
T
his
co
ncep
t
i
s
represe
nt
ed by
E
quat
ion (
3).
Ba
sed on t
hese
preps,
we
ca
n defi
ne
this m
ovem
ent as
:
12
(
0
)
tt
ii
x
x
R
andn
(9)
Wh
e
re
⨂
m
e
ans
a
sel
f
-
pe
r
m
uta
ti
on
a
nd
(
,
2
)
de
fines
a
rand
om
nu
m
ber
of
pe
rm
ut
at
ion
s
.
An ex
am
ple of
this
op
e
rati
on
is represe
nted as f
ollows:
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
Impr
oved Chic
ken
Swa
r
m O
pt
imizati
on Alg
ori
thm
t
o
So
lv
e
t
he
Tr
avell
in
g S
alesm
an…
(
Fa
yçal Che
bih
i
)
1057
Let
’s
=
[
1
,
2
,
3
,
4
,
5
]
an
d
Ra
nd =
2
Figure
1. Exa
m
ple o
f
m
ov
em
ent o
f
a ro
os
t
er
The ne
w
s
ol
ution bec
om
es:
+
1
=
[
1
,
3
,
2
,
5
,
4
]
b)
Hen
s
Hen
s
f
ollow
t
heir
group
-
m
ate
rooster
t
o
lo
ok
f
or
foo
d.
Howe
ver,
they
rand
om
l
y
steal
food
from
oth
e
r
chic
ken
s
.
These
m
ov
em
ents
are
def
ine
d
in
E
qu
at
io
n
(
2).
I
n
this
ada
pt
at
ion
,
we
repr
esent
the
m
ov
e
m
ent
of
he
n
s
f
ollo
wing
the
group
-
m
at
e
ro
os
te
r
as
a
local
search
arou
nd
the
r
oo
ste
r.
In
the
sa
m
e
way,
we
rep
rese
nt
the
m
ov
em
ent
of
he
ns
w
hile
ste
al
ing
foo
d
f
r
om
oth
er
chicke
ns
as
a
local
searc
h
arou
nd
t
he
c
hi
cken.
The
m
ov
em
ent equat
ion o
f he
ns
bec
om
e:
1
1
2
()
()
t
t
t
t
i
i
r
i
tt
ri
x
x
R
a
n
d
x
x
R
a
n
d
x
x
!
!
(10)
Wh
e
re
A
⊖
B
presents
a list
of
p
e
rm
utati
on
s to go f
ro
m
so
lut
ion
B
to
s
olu
ti
on A.
Exam
ple:
Let
’s
A=
[1,2,
3,4,5]
and B=
[ 2,
4,3
,1
,
5]
Step1
:
(
1
2)
Step2
:
(
4
2)
Figure
2. Exa
m
ple o
f
m
ov
em
ent o
f
a
hen
s
ste
p 1 a
nd
s
te
p 2
Finall
y, to
go fro
m
so
luti
on
A
to s
olu
ti
on B t
he
li
st o
f perm
utati
on
is:
A
⊖
B = [{
1,2},
{
4,2}]
.
The
⊕
ope
r
at
or
i
nd
ic
at
es
that
we
ad
d
the
ne
xt
operati
on
t
o
t
he
new
s
olu
t
ion
c
reated
by
the
la
st op
e
rati
on.
c)
Chicks
The
c
hicks
fi
nd
the
f
ood
a
r
ound
their
m
oth
e
r.
T
his
c
on
ce
pt
pr
e
sents
a
fata
l
disad
va
ntage
because
the
chicks
ca
n
on
ly
le
arn
fr
om
their
m
oth
er
s.
Theref
or
e
,
the
chicks
ca
n
easi
ly
fall
i
nto
local
m
ini
m
u
m
.
Dinghui
[24]
pro
poses
a
new
equ
at
io
n
to
go
thr
ough
this
prob
le
m
by
a
dd
i
ng
the
prob
a
bili
ty
of
le
a
rn
i
ng
from
the m
ai
n
roost
er a
nd a s
el
f
-
le
arn
i
ng operati
on. T
he
n
e
w
e
quat
ion bec
om
es:
1
()
()
t
t
t
t
i
i
m
i
tt
ri
x
w
x
F
L
x
x
C
x
x
!
!
(11)
Wh
e
re
w
is
a
sel
f
-
per
m
utati
on
pa
ram
et
er
th
at
ind
ic
at
es
the
nu
m
ber
of
pe
rm
utati
on
s
an
d
FL
is
a
le
arn
in
g
factor
wh
ic
h
m
eans
that
the
chick
le
arn
f
r
om
it
s
m
o
ther
in
the
sa
m
e
way
C
is
a
lear
ni
ng
fact
or
f
ro
m
the
rooster
.
A
=
B =
1
2
3
4
5
2
4
3
1
5
A
=
B’ =
1
2
3
4
5
1
4
3
2
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.
12
, N
o.
3
,
Dece
m
ber
2
01
8
:
1054
–
1062
1058
4.3.
The
Neig
hb
or
hood
To
im
pr
ove
th
e
so
luti
on
qual
it
y,
neighb
orh
ood
m
et
ho
ds
a
re
re
quire
d.
T
hi
s
arti
cl
e
pro
poses
a
2
-
opt
local
searc
h
t
o im
pr
ove the
qu
al
it
y of
the
so
l
ution p
r
opos
e
d by the
DSC
O
al
gorithm
.
Figure
3. Mo
ve
f
r
om
so
luti
on
(
B)
to
s
olu
ti
on
(
A
) wit
h
a
sim
ple p
e
rm
utati
on
he
2
-
opt
m
ov
e
m
ent
causes
a
s
m
al
l
per
tur
bation
t
o
the
so
luti
on
in
or
der
to
fin
d
a
good
so
l
ution
in
it
s
neig
hborh
ood.
This
op
e
rati
on
is dem
on
strat
e
d
in
Fig
ure
3.
4
.
4
.
Discrete
CS
O
A
l
go
ri
thm
St
ep
1:
In
it
ia
li
ze
a
po
pu
la
ti
on
of
N
c
hick
ens
a
nd
def
i
ne
relat
ed
pa
ra
m
et
ers
(N
the
num
ber
of
chicke
ns
in
t
he
swar
m
,
Ra
nd
[0,
1],
r1
[1…
N]
is
an
in
de
x
of
rooster
r
2
r
1
[1,
…,
N]
is
a
n
in
dex
of
chic
kens,
FL [0,1]
,
C a
nd
w.
St
ep
2:
Use
a
2
-
opt
local
sea
rch to im
pr
ove
the quali
ty
o
f
s
olu
ti
ons
St
ep
3:
E
valua
te
Th
e
N
c
hick
en’
s
f
it
nes
s
values
at
t=
0
a
nd
save t
he glo
bal b
est
s
olu
ti
on.
St
ep
4:
Ra
nk
the
chick
ens
a
nd
est
ablish
a
hi
erarch
al
order
in
the
swar
m
and
im
pr
ove
the
so
l
utio
n
us
in
g 2
-
opt l
oc
al
se
arch.
St
ep
5:
Ra
nd
om
ly
div
ide
the
swar
m
into
diff
e
ren
t
gro
ups
and
determ
ine
the
relat
ion
s
hi
p
betwee
n
the ch
ic
ks
a
nd
the m
oth
er
-
he
ns i
n
a
grou
p.
St
ep
6:
Fi
nd
a
new
s
olu
ti
on
by
updatin
g
t
he
po
sit
io
n
of
eac
h
rooster
,
hen
s
an
d
c
hicks
us
i
ng
the
ne
w
E
quat
ion
s
(9
)
, (1
0) an
d (
11).
St
ep
7:
Update
the
new s
olu
ti
on whe
n
it
is
be
tt
er th
an
the
previ
ou
s
one.
St
ep
8:
Re
tu
rn to ste
p 4 until
the m
axi
m
u
m
nu
m
ber
of it
er
at
ion
s is
reac
he
d.
5.
E
X
PERI
MEN
TAL RES
UL
TS
This
sect
io
n
il
lustrate
s
perfor
m
ance
te
sts
of
CSO
al
gorith
m
on
Eu
cl
idea
n
insta
nces
of
TSPL
IB.
All
exp
e
rim
ents
are
pe
rfor
m
ed
on
a
n
In
te
l
c
ompu
te
r
process
or
(R
)
C
or
e
(T
M)
i7
-
65
00
CPU
@
2.5
GH
z
@
2.
60
GH
z
a
nd
16
G
B
of
RAM.
Th
e
pr
og
ram
is
c
od
e
d
in
the
pr
ogram
m
ing
la
ngua
ge
C
#
Visu
al
Stud
i
o
20
15
a
nd
for
eac
h
in
sta
nc
e TSPL
IB
w
e
te
st 100
ti
m
es.
To
r
un
the
pro
gr
am
,
a
li
st
of
par
am
et
ers
has
be
en
s
et
.
Table
2
s
hows
the
val
ues
of
th
e
par
am
et
ers
us
e
d:
Table
2.
T
he
P
aram
et
ers
Valu
es
Para
m
eters
Valu
es
PS
(
p
o
p
u
la
tio
n
siz
e)
100
RN
(
Nu
mb
er o
f ro
o
ster
s)
2%
HN
(
Nu
mb
er o
f he
n
s)
20%
CN
(
Nu
mb
er o
f chicks
)
78%
G
(
Nu
mb
er o
f tou
rs
to u
p
d
a
te the a
lg
o
rith
m)
2
C
(
Roo
ster
lear
n
in
g
facto
r)
0
,4
FL
(Hens lea
rn
in
g
facto
r)
0
,4
W
(self
-
l
ea
rn
in
g
fa
cto
r)
0
.9
5
.
1
.
P
ar
ametr
ic
A
n
aly
sis
Ther
e
a
re
ei
gh
t
par
am
et
ers
i
n
the
DCSO
a
lgorit
hm
.
The
pur
po
s
e
of
t
he
al
go
rit
hm
is
t
hat
the
he
ns
m
us
t
loo
k
for
a
ne
w
s
olu
ti
on
ar
ou
nd
a
go
od
one
repres
ented
by
the
r
oo
ste
r.
As
a
r
esult,
the
num
ber
of
roosters
m
us
t
be
higher
tha
n
t
he
nu
m
ber
of
he
ns
(N
R
>
N
H
).
Sim
il
arly
,
the
chicks
are
s
e
ekin
g
a
new
sol
ution
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
Impr
oved Chic
ken
Swa
r
m O
pt
imizati
on Alg
ori
thm
t
o
So
lv
e
t
he
Tr
avell
in
g S
alesm
an…
(
Fa
yçal Che
bih
i
)
1059
arou
nd
the
ir
m
oth
e
rs,
an
d
f
or
a
good
inte
ns
if
ic
at
ion
,
the
nu
m
ber
of
chicks
m
us
t
be
hig
he
r
than
the
nu
m
ber
of
hens
(
NC
>
N
H)
.
If
t
he
val
ue
of
G
is
ve
ry
high,
the
al
gorithm
cann
ot
converge
qui
ckly
to
the
op
tim
a
l
so
luti
on.
Ot
herwise,
if
the
val
ue
of
G
is
ve
r
y
low,
the
al
gorithm
m
ay
fall
into
a
l
ocal
optim
al
.
Af
te
r
s
ever
al
te
sts,
G=2
m
a
y
achieve
a
good
re
su
lt
in
a
m
uch
red
uce
d
tim
e.
As
fo
r
G
,
the
value
of
this
par
am
et
er
has
a
sign
ific
a
nt
im
p
act
on
t
he
res
ul
t.
Fu
rt
her
m
ore,
to
a
vo
i
d
the
pro
blem
of
fa
ll
ing
into
m
inim
u
m
op
ti
m
a
l
i
n
the
chick
’s
m
ov
e
m
ent,
the
sel
f
-
le
arn
in
g
pa
ram
et
er
gi
ves
t
o
th
e
chic
k
the
po
ssibil
it
y
to
fin
d
a
good
so
l
ution
in
a
bigger
s
pace
of
so
l
utions
by
the
value
of
w
=0,9.
T
he
FL
par
am
et
ers
give
the
chick
s
th
e
po
s
sibil
it
y
to
le
arn
from
the
m
oth
er.
Mo
reove
r,
the
best
val
ue
m
us
t
be
a
ran
dom
nu
m
ber
betwee
n
0,4
and
1
(FL
∈
[
0.4,1]
).
In
o
r
der
t
o
give
m
or
e
rob
us
tn
ess
to
the
al
gor
it
h
m
,
the
chick
s
can
al
so
le
ar
n
from
ro
os
te
rs
us
in
g
a
C
par
a
m
et
er
with a lea
rn
i
ng f
act
or
rand
oml
y cho
se
n betw
een
0.4 a
nd 1 (
C
∈
[
0.4,1]
).
Figure
4
s
how
s
the
ev
olu
ti
on
of
the
ave
ra
ge
r
unni
ng
ti
m
e
fo
r
100
e
xe
cutions
w
hile
var
yi
ng
th
e
par
am
et
er
G
and
the
siz
e
of
the
popula
ti
on
us
in
g
TSPL
IB
instance
with
a
diff
e
ren
t
num
ber
of
the
ci
ty
(S
T
70
and Kr
oA1
00).
Figure
4. Ru
n
t
i
m
e o
btaine
d b
y varyin
g G
pa
ram
et
er
Figure
5. Ru
n
t
i
m
e o
btaine
d b
y varyin
g p
op
ul
at
ion
siz
e
Fig
ure
6. Ru
n
t
i
m
e o
btaine
d b
y varyin
g
t
he n
um
ber
of
roost
ers
The
re
su
lt
s
re
veal
that
there
is
an
increa
se
in
the
exec
ution
ti
m
e
wh
en
increasin
g
the
value
of
t
he
par
am
et
er G
on Fi
g
ure
2
a
nd
th
e size
of
t
he pop
ulati
on
on
Figure
5.
Fo
r
m
or
e
detai
ls,
oth
e
r
ex
per
i
m
ents
hav
e
be
en
pe
rfor
m
ed
to
detect
the
be
st
value
of
the
par
am
et
ers
RN, HN
and C
N.
The
e
xp
e
rim
e
nts
in
Fi
gure
6
s
how
that
t
he
pe
rce
ntage
of
t
he
r
ooste
r
sho
uld
be
l
ow
t
o
e
ns
ure
faster c
onve
rg
e
nce.
Table
3
sho
ws
the
num
erical
resu
lt
s
ob
ta
i
ne
d
w
he
n
a
pp
l
yi
ng
the
DCS
O
to
t
he
TS
P
us
in
g
s
om
e
TSPL
IB
instances.
T
he
fi
rst
colum
n
con
ta
ins
the
nam
e
of
the
i
ns
ta
nc
e,
the
seco
nd
colum
n
con
ta
i
ns
the
nu
m
ber
of
node
s
in
the
i
ns
ta
nce,
t
he
thir
d
colum
n
co
ntains
the
opti
m
a
l
so
luti
on
fou
nd
in
T
SPL
IB
Lib
ra
ry,
and
t
he
f
ourt
h
colum
n
co
ntains
the
best
s
olu
ti
on
obta
in
ed
by
th
e
DC
SO
al
go
rithm
.
More
ov
e
r,
t
he
fifth
colum
n
co
ntains
the
w
orst
s
olu
ti
on,
t
he
si
xth
c
olu
m
n
co
ntains
th
e
ave
rag
e
so
l
ution,
the
seve
nt
h
co
lum
n
denotes
the
sta
nd
a
r
d
de
viati
on
of
s
olu
ti
on
obt
ai
ned
over
30
in
dep
e
ndent
runs,
t
he
ei
ghth
col
um
n
con
t
ai
ns
the
per
ce
ntage
de
viati
on
of
the
aver
a
ge
so
l
ution
over
30
i
nd
e
pe
nd
e
nt
runs,
t
he
nin
th
colum
n
co
ntains
th
e
per
ce
ntage
de
viati
on
of
the
best
so
luti
on
i
n
30
ind
e
pe
ndent
runs,
the
t
enth
col
um
n
i
s
repres
ente
d
by
tow
par
am
et
ers.
Th
e
nu
m
ber
f
or
optim
al
so
luti
on
(C
opt
)
an
d
the
nu
m
ber
of
s
olu
ti
on
(ove
r
30
runs)
f
or
wh
ic
h
the
dev
ia
ti
on
from
optim
al
so
luti
on
is
le
ss
t
ha
n
or
e
qual
to
1,
a
nd
the
la
st
colum
n
re
pr
e
sents
the
best
tim
e
0
200
400
600
2
5
10
Ti
me
(
s)
st70
KroA100
0
100
200
300
100
200
400
Ti
me
(
s)
st70
KroA100
0
200
400
2
5
10
Ti
me
(
s)
st70
KroA100
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.
12
, N
o.
3
,
Dece
m
ber
2
01
8
:
1054
–
1062
1060
ob
ta
ine
d
by
th
e
DCS
O
al
go
rithm
in
30
i
nd
e
pe
nde
nt
r
uns.
The
al
gorithm
stop
s
w
he
n
the
best
s
olu
ti
on
i
s
f
ou
nd
or if the
ex
e
cut
ion
ti
m
e exceed
s
3000s.
Table
3.
N
um
e
rical
Re
su
lt
s Obtai
ned b
y
DC
SO
A
pp
li
ed
to Som
e TSP
I
ns
t
ances
of
TS
PL
IB
Ins
tan
ce
Size
Op
t
Bes
t Sol
W
o
rst Sol
Av
erage
PDav
(
%)
PDb
est (%)
C
1%
/C
opt
Ti
m
e
(
s)
Eil51
51
426
426
426
426
0
.00
0
.00
3
0
/3
0
0
,32
Berlin
5
2
52
7542
7542
7542
7542
0
.00
0
.00
3
0
/3
0
0
,09
S
t7
0
70
675
675
675
675
0
.00
0
.00
3
0
/3
0
0
,04
Eil76
76
538
538
541
2
3
9
.5
0
.27
0
.00
3
0
/2
1
5
,78
Pr7
6
76
1
0
8
1
5
9
1
0
8
1
5
9
1
0
8
1
5
9
1
0
8
1
5
9
0
.00
0
.00
3
0
/3
0
8
,35
Rat9
9
99
1211
1211
1211
1211
0
.00
0
.00
3
0
/3
0
1
1
,79
Kr
o
A10
0
100
2
1
2
8
2
2
1
2
8
2
2
1
2
8
2
2
1
2
8
2
0
.00
0
.00
3
0
/3
0
0
,05
Kr
o
B10
0
100
2
2
1
4
1
2
2
1
4
1
2
2
1
4
1
2
2
1
4
1
0
.00
0
.00
3
0
/3
0
3
,17
Kr
o
C1
0
0
100
2
0
7
4
9
2
0
7
4
9
2
0
7
4
9
2
0
7
4
9
0
.00
0
.00
3
0
/3
0
2
,68
Kr
o
D1
0
0
100
2
1
2
9
4
2
1
2
9
4
2
1
2
9
4
2
1
2
9
4
0
.00
0
.00
3
0
/3
0
7
,49
Kr
o
E10
0
100
2
2
0
6
8
2
2
0
6
8
2
2
1
5
6
2
2
1
1
2
0
.19
0
.00
3
0
/0
5
1
1
,52
Rd1
0
0
100
7910
7910
7910
7910
0
,00
0
.00
3
0
/3
0
1
0
,55
Eil10
1
100
629
629
637
6
3
2
.43
0
.54
0
.00
3
0
/0
5
1
6
.9
Lin
1
0
5
105
1
4
3
7
9
1
4
3
7
9
1
4
3
7
9
1
4
3
7
9
0
.00
0
.00
3
0
/3
0
2
,2
Pr1
0
7
107
4
4
3
0
3
4
4
3
0
3
4
4
3
2
6
4
4
3
1
4
,5
0
,02
0
,00
3
0
/2
5
2
0
,02
Pr1
2
4
124
5
9
0
3
0
5
9
0
3
0
5
9
0
3
0
5
9
0
3
0
0
,00
0
,00
3
0
/3
0
1
,9
Bier1
2
7
127
1
1
8
2
8
2
1
1
8
2
8
2
1
1
8
6
5
7
1
1
8
4
6
9
,5
0
,15
0
,00
3
0
/7
6
4
,62
Ch
1
3
0
130
6110
6110
6155
6
1
2
4
,1
0
,23
0
,00
3
0
/5
1
5
,05
Pr1
3
6
136
9
6
7
7
2
9
6
7
7
2
9
7
4
6
8
9
6
9
9
5
0
,23
0
,00
3
0
/4
2
0
,75
Pr1
4
4
144
5
8
5
3
7
5
8
5
3
7
5
8
5
3
7
5
8
5
3
7
0
,00
0
,00
3
0
/3
0
2
,01
Ch
1
5
0
150
6528
6528
6584
6
5
5
0
,3
0
,34
0
,00
3
0
/2
2
3
,7
Kr
o
A15
0
150
2
6
5
2
4
2
6
5
2
4
2
6
6
4
9
2
6
5
6
0
,2
0
,13
0
,00
3
0
/7
2
0
,02
Kr
o
B15
0
150
2
6
1
3
0
2
6
1
3
0
2
6
2
6
6
2
6
1
4
6
,63
0
,06
0
,00
3
0
/7
2
1
,21
Pr1
5
2
152
7
3
6
8
2
7
3
6
8
2
7
3
8
1
8
7
3
7
5
9
,06
0
,10
0
,00
3
0
/2
0
1
3
,4
Rat1
9
5
195
2323
2324
2360
2
3
4
0
,7
0
,76
0
,04
2
0
/0
-
D1
9
8
198
1
5
7
8
0
1
5
7
8
0
1
5
8
7
0
1
5
8
0
2
,83
0
,14
0
,00
3
0
/3
4
5
,07
Kr
o
A20
0
200
2
9
3
6
8
2
9
3
6
8
2
9
7
4
0
2
9
4
4
9
,23
0
,27
0
,00
3
0
/2
4
9
,05
K
ro
B20
0
200
2
9
4
3
7
2
9
4
4
8
2
9
8
1
9
2
9
5
4
2
.49
0
.29
0
.03
2
8
/0
-
Gil2
6
2
262
2378
2382
2410
2
3
9
0
,7
0
,53
0
,08
2
6
/0
-
A28
0
280
2579
2579
2611
2
5
8
6
,8
3
0
,30
0
,00
3
0
/6
1
0
2
,17
Pr2
9
9
299
4
8
1
9
1
4
8
1
9
1
4
8
5
5
2
4
8
3
1
1
,7
0
,25
0
,00
3
0
/2
1
2
5
,11
Lin
3
1
8
318
4
2
0
2
9
4
2
1
5
4
4
2
7
1
3
4
2
4
6
2
.
16
1
.03
0
.29
1
2
/0
-
Rd4
0
0
400
1
5
2
8
1
1
5
3
3
6
1
5
5
7
4
1
5
4
6
5
.3
1
.20
0
.35
7
/1
0
-
Nr
w1
3
7
9
1379
5
6
6
3
8
5
8
9
5
1
5
9
8
3
7
5
9
3
4
9
.53
4
.78
4
.08
0
/0
-
The result
s i
n Table
3 co
nfi
r
m
that t
he
DC
SO
al
gorithm
c
an
s
olv
e m
os
t of
t
he
TS
PL
IB
instances i
n
a
ver
y
fast e
xecut
ion
ti
m
e.
To
pr
ov
e
th
e
r
obus
t
ness
of
the
al
gorithm
,
Table
4
c
om
par
es
the
a
ver
a
ge
so
l
ution
pr
opos
e
d
by
the
DS
CO al
gorith
m
w
it
h
oth
e
r m
et
ho
ds
rece
ntly
u
sed
a
nd
w
hi
ch
are a
pp
li
ed
to
so
l
ve
TSP u
sing
T
SPL
IB li
br
a
ry.
Table
5
c
om
par
es
the
be
st
ex
ecuti
on
ti
m
e
a
chieve
d
by
the
al
go
rithm
com
par
ed
to
ne
w
bio
-
i
nspired
al
gorithm
that sol
ves
th
e
TSP.
T
able
4.
T
he
A
ver
a
ge
Re
s
ults
Ob
ta
ine
d
by
S
om
e Me
tho
ds
on
S
om
e
TSPLI
B
I
ns
ta
nce
s
Ins
tan
ce
Op
t
DSCO
ACO
[
2
5
]
[
2
6
]
PSO [
1
3
]
GA
[
2
7
]
[
2
8
]
Eil51
426
426
430
4
3
6
,9
429
Berlin
5
2
7542
7542
7594
7832
7738
St7
0
675
675
750
6
9
7
,5
-
Eil76
538
538
5
5
2
,6
5
6
0
,4
-
KroA1
0
0
2
1
2
8
2
2
1
2
8
2
2
1
4
7
5
-
2
1
4
4
5
Accor
ding
to
the
res
ults
in
Table
4,
the
ave
rag
e
ti
m
e
ob
ta
ined
by
the
D
SCO
al
gorith
m
giv
es
good
resu
lt
s
t
h
an
th
e
ot
her
m
et
ho
ds
.
Ta
ble
5,
F
ig
ure
7,
a
nd
F
ig
ure
8
com
pare
the
a
ve
rag
e
t
i
m
e
ob
ta
ine
d
by
the
DS
CO
al
gorith
m
co
m
par
ed
to
n
e
w bio
-
ins
pir
ed
al
go
rithm
s
.
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
Impr
oved Chic
ken
Swa
r
m O
pt
imizati
on Alg
ori
thm
t
o
So
lv
e
t
he
Tr
avell
in
g S
alesm
an…
(
Fa
yçal Che
bih
i
)
1061
Table
5.
T
he
Avera
ge
Tim
e Result
s Obtai
ne
d by Bi
o
-
i
nspir
ed
Me
th
ods
on So
m
e TSPL
IB Instances
Ins
tan
ce
Op
t
DSCO
CS [
2
0
]
BA [
1
8
]
HUS [
2
1
]
Eil51
426
0
,32
1
,16
0
,20
0
,51
Berlin
5
2
7542
0
,09
0
,09
0
,03
0
,16
St7
0
675
0
,04
1
,56
0
,43
1
,42
KroA1
0
0
2
1
2
8
2
0
,05
2
,70
1
,36
7
,18
KroB
1
0
0
2
2
1
4
1
3
,17
8
,74
3
,35
1
1
,25
KroC
1
0
0
2
0
7
4
9
2
,50
3
,36
2
,51
6
,48
Figure
7. Com
par
i
ng ex
ec
ution t
i
m
e o
f bio
-
insp
ire
d
al
gorithm
in
a sm
a
ll
TSP
inst
ances
Figure
8. Com
par
i
ng ex
e
c
ution t
i
m
e o
f bio
-
insp
ire
d
al
gorithm
in
a large
TSP i
ns
ta
nces
The
ex
pe
rim
en
ta
l
resu
lt
s
sh
ow
that
exec
ution
ti
m
e
of
DS
CO
al
gorithm
is
bette
r
than
the
othe
r
bi
o
-
insp
ire
d
al
gorithm
s
in
m
os
t
of
TSPL
IB
in
sta
nces
us
e
d
to
te
st
the
perform
a
nce
of
the
al
gorithm
.
Especial
ly
in
la
rg
e
insta
nce,
our
al
gorithm
can
fi
nd
the b
e
st
so
luti
on
in
t
he
best
ti
m
e,
wh
ic
h
will
be
use
d
to
s
olv
e o
th
er
NP
-
hard c
om
bin
at
or
y
pro
blem
s lik
e T
SP.
6.
CONCL
US
I
O
N
This
pap
e
r
pr
e
sents
a
ne
w
di
screte
Chicke
n
s
war
m
op
ti
m
iz
at
ion
(D
C
SO
)
al
gorithm
to
s
olv
e
t
he
sy
m
m
e
tric
Trav
el
ing
sal
esm
a
n
pr
ob
le
m
(TSP
)
by
ap
plyi
ng
a
l
ocal
sea
rch
t
o
im
pr
ove
the
qual
it
y
of
t
he
so
luti
ons.
T
he
adap
ta
ti
on
of
the
al
gorithm
is
based
on
the
beh
a
viors
of
chicke
ns
in
a
swar
m
.
The
al
go
rithm
has
been
te
ste
d
on
a
set
of b
en
chm
ar
k
instanc
es
of
TSP
LIB.
Its
pe
rfor
m
ance
excee
ds
the re
cent
m
et
ho
ds
u
s
e
d
to so
l
ve
t
he
TS
P
su
c
h
a
s
ACO
, P
S
O
a
nd
GA,
the e
ff
ect
ive
ne
ss of th
e
alg
ori
th
m
is
du
e
to
t
he dive
rsificat
ion o
f
op
e
rati
ons a
nd
op
e
rato
rs base
d on the
dif
fer
e
nt k
i
nd of c
hic
kens (r
oo
ste
rs,
hens a
nd ch
ic
ks).
In
the
f
uture
r
esearche
s
we
will
i
m
pr
ove
t
he
DCS
O
al
go
rithm
in
orde
r
to
obta
in
a
bet
te
r
res
ults
by
app
ly
in
g
an
hy
br
i
d
ap
proach
e
s.
More
ov
e
r,
t
he
DCS
O
r
obust
ness
an
d
ra
pid
it
y
enco
ura
ge
the
us
e
of
al
gorithm
to s
olv
e
oth
e
r
c
om
bin
at
or
ia
l o
pti
m
iz
at
ion
p
r
oble
m
s.
REFERE
NCE
S
[1]
S.
Arora,
“
Poly
nom
ia
l
ti
m
e
a
pproximati
on
sc
hemes
for
Euc
l
ide
an
tr
aveling
sale
sm
an
and
othe
r
geometric
proble
m
s,”
J.
A
CM,
vol. 45, no. 5, pp. 753
–
782,
Sep.
1998
.
[2]
R.
G.
Bla
nd
an
d
D.
F.
Shall
cro
ss
,
“
La
rge
travel
li
ng
sal
esm
an
p
roble
m
s
ari
s
ing
from
expe
riments
in
X
-
ra
y
cr
y
st
allogra
ph
y
:
A pre
li
m
ina
r
y
re
port
on
computa
ti
on,
”
Oper
.
R
es.
Lett.,
vo
l. 8, no. 3, pp. 125
–
128,
Jun.
1989.
[3]
H.
D.
Rat
l
iff
an
d
A.
S.
Rosenth
al
,
“
Order
-
Pick
i
ng
in
a
Re
ct
ang
ula
r
W
are
house
:
A
Solvabl
e
Cas
e
of
the
T
rav
e
ling
Sale
sm
an
Proble
m
,
”
Oper. Re
s.
,
vol.
31
,
no
.
3
,
pp
.
507
–
521
,
Jun.
1983.
[4]
J.
K.
Le
nstra
an
d
A.
H.
G.
R.
Kan,
“
Som
e
Sim
p
le
Applicati
ons
of
the
Tra
v
el
l
ing
Sale
sm
an
Problem,”
J.
Oper.
Re
s.
Soc.
,
vol. 26, no. 4, pp. 717
–
733,
Dec
.
1975.
[5]
M.
Za
ch
ariasen
and
M.
Dam
,
“
T
abu
Sear
ch
on
th
e
Geom
et
ric
Travel
ing
Sal
esm
an
Problem,”
in
M
et
a
-
Heur
isti
cs
,
I.
H.
Os
m
an
and
J.
P.
Ke
lly
,
Eds.
B
oston,
MA
:
Spri
nger
US
,
1996
,
p
p.
571
–
587
.
[6]
J.
-
Y.
Potvin,
“
G
ene
t
ic
al
gor
it
hm
s
for
the
tra
vel
i
ng
sale
sm
an
proble
m
,
”
Ann.
Oper.
Res.
,
vo
l.
63
,
no.
3,
pp.
337
–
370,
Jun.
1996.
[7]
Abid,
M.
M.,
&
Muhammad,
I.
(2015).
Heuri
sti
c
Approac
hes
to
Solve
Tra
velin
g
Sale
sm
an
Pro
ble
m
.
Indone
sia
n
Journal
of El
ec
tr
ic
a
l
Eng
ineeri
ng
and
Com
puter S
ci
en
ce,
15(2)
,
39
0
-
396.
[8]
Y.
Marina
kis,
A.
Migdal
as,
and
P.
M.
Pard
al
os,
“
Expa
nding
Nei
ghborhood
GRASP
for
the
Tra
v
el
ing
Sale
sm
an
Problem,”
Com
put.
Opt
im.
Appl
.
,
vol
.
32
,
no
.
3
,
p
p.
231
–
257
,
De
c
.
2005
.
[9]
X.
Geng,
Z
.
Che
n,
W
.
Yang
,
D
.
Shi,
and
K.
Zhao,
“
Solving
th
e
t
rav
eling
sal
esm
an
proble
m
b
ase
d
on
an
ad
apt
iv
e
sim
ula
te
d
an
ne
aling
a
lgori
thm wi
th
gre
ed
y
se
arc
h
,
”
Appl.
Soft
Co
m
put.
,
vol
.
11,
n
o.
4
,
pp
.
3680
–
3
689,
Jun.
2011.
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.
12
, N
o.
3
,
Dece
m
ber
2
01
8
:
1054
–
1062
1062
[10]
Al
-
Obaidi
,
A.
T
.
S.,
Abdulla
h,
H.
S.,
&
Ahm
ed,
Z.
O.
(2018).
Mee
rka
t
Cla
n
Algori
thm:
A
New
Swa
rm
Inte
ll
ig
ence
Algorit
hm
.
Indo
nesia
n
Journa
l
o
f
Elec
tr
ical E
ngi
nee
ring
and
Co
m
pute
r
Scie
n
ce, 10(1),
354
-
360
.
[11]
M.
Manfrin,
M.
Bira
ttari
,
T.
Stüt
zl
e
,
and
M.
Dori
go,
“
Para
ll
e
l
Ant
Colon
y
Optimiz
at
ion
for
the
Travel
ing
Sal
esm
an
Problem,”
in
Ant
Colon
y
Opti
m
iz
at
ion
and
S
warm
Inte
lligen
ce
,
vol.
4150,
M.
Dorigo,
L.
M.
Gam
bar
del
l
a
,
M
.
Bira
ttari
,
A.
Marti
nol
i,
R
.
Poli
,
and
T
.
Stützl
e,
Eds.
B
erl
in
,
Heide
lb
erg
:
Springer
Ber
li
n
Heide
lb
erg
,
200
6,
pp.
224
–
234
.
[12]
M.
Cle
rc
,
“
Discre
t
e
Particl
e
Sw
arm
Optimiza
t
ion,
illus
trated
b
y
th
e
Tr
ave
l
in
g
Sale
sm
an
Problem,”
in
New
Optimiza
ti
o
n
T
ec
hniqu
es
in
E
ngine
er
ing,
vo
l.
141,
Ber
li
n
,
Heide
lb
erg
:
Spr
inge
r
Be
rli
n
H
ei
de
lbe
rg,
2004
,
pp.
219
–
239
.
[13]
X.
H.
Shi,
Y.
C.
Li
ang,
H.
P.
L
e
e,
C.
Lu
,
and
Q.
X.
W
ang,
“
Parti
cl
e
sw
arm
opti
m
iz
a
ti
on
-
base
d
algorithms
for
TS
P
and
gen
erali
z
ed TSP
,
”
Inf
.
Proc
e
ss
.
Le
t
t.,
vo
l. 10
3,
no
.
5
,
pp
.
169
–
176,
Aug.
2007
.
[14]
D.
Te
odorovic,
P.
Luc
ic,
G.
Markovic
,
and
M.
D.
Orco,
“
Bee
Colon
y
Optimiz
ation:
Princi
pl
es
a
nd
Applic
ations,
”
2006,
pp
.
151
–
1
56.
[15]
Zong
W
oo
Gee
m
,
Joong
Hoon
Kim
,
and
G.
V.
Loga
na
tha
n,
“
A
New
Heuri
stic
Optimiza
ti
o
n
Al
gorit
hm
:
Harm
o
n
y
Sear
ch,”
SIM
UL
ATION
,
vol. 76, no. 2, pp. 60
–
68
,
Feb
.
2001
.
[16]
M.
BOU
ZIDI
and
M.
E.
RIFF
I,
“
AD
AP
TATI
O
N
OF
THE
HARMON
Y
SEAR
CH
ALGO
RIT
HM
TO
SO
LV
E
THE
TRAVEL
LING
SA
LE
SMAN
PR
OBLEM
,
”
Journa
l
of
T
heor
etical
and
Applie
d
Inform
at
ion
Technol
o
g
y
,
04
-
Oc
t
-
2014.
[17]
X.
-
S.
Yang,
“
A
New
Meta
heur
isti
c
Bat
-
Inspir
ed
Algorit
hm
,
”
in
Natur
e
Inspire
d
Cooperati
ve
Strat
eg
ie
s
for
Optimiza
ti
o
n
(NICS
O
2010),
vol.
284,
J.
R.
Gonzá
l
ez,
D.
A.
Pelta,
C.
Cruz
,
G.
T
err
azas,
and
N.
Krasnogor,
Eds.
Berl
in
,
He
ide
lb
e
rg:
Springer
Ber
l
in
Heid
el
ber
g
,
2
010,
pp
.
65
–
74
.
[18]
Saji
,
Y.
,
&
R
iff
i,
M.
E
.
(2016)
.
“
A
novel
discr
et
e
b
at
al
gori
th
m
for
solving
the
tra
v
el
l
ing
sal
esm
an
proble
m
,
”
Neura
l
Com
put
.
Appl.
,
vol. 27, n
o.
7
,
pp
.
1853
–
1
866,
Oct
.
2016
.
[19]
X.
-
S.
Yang and
Suash Deb,
“
Cu
ckoo
Sear
ch
v
ia
L&
#x00E9;
v
y
fl
ight
s,”
2009,
pp.
210
–
214.
[20]
A.
Ouaa
rab
,
B.
Ahiod,
and
X.
-
S
.
Yang,
“
Discre
t
e
cuc
koo
s
ea
rch
al
gori
thm
for
the
tra
v
el
l
ing
sal
esm
an
proble
m
,
”
Neura
l
Com
put
.
Appl.
,
vol. 24,
n
o.
7
–
8
,
pp
.
1659
–
1669,
Jun.
201
4.
[21]
amine
AG
HA
R
GH
OR
and
M.
E.
RIFF
I,
“
HUNT
ING
SEARC
H
ALGO
RITH
M
TO
SO
LVE
THE
TRAVELI
NG
SA
LE
SM
AN
P
ROBLEM.,
”
Jou
rna
l
of
Th
eor
etic
al
and
Appl
ie
d
I
nform
at
ion
T
ec
h
nolog
y
,
10
-
Apr
-
2015.
[22]
X.
Meng,
Y.
Li
u,
X.
Gao,
and
H.
Zha
ng,
“
A
New
Bio
-
inspire
d
Algorit
hm
:
C
hic
ken
Sw
arm
Optimiza
ti
o
n,
”
i
n
Advanc
es
in
Sw
arm
Inte
lligence,
vol.
8794,
Y.
T
an,
Y.
Shi,
and
C.
A.
C.
Coe
ll
o
,
Eds.
Cham:
Springer
Inte
rn
at
ion
a
l
Publishing,
2014
,
pp
.
86
–
94
.
[23]
J.
Monnot
and
S.
Tou
louse,
“
The
Tra
v
el
ing
Sale
s
m
an
Proble
m
an
d
it
s
Var
ia
t
ions,
”
in
Pa
rad
igms
of
Com
bina
toria
l
Optimiza
ti
o
n,
V.
Th. Paschos,
Ed
.
Hoboken
,
NJ
,
US
A:
John W
il
ey
&
Sons
,
Inc
.
,
2
013,
pp
.
173
–
21
4.
[24]
D.
W
u,
F.
Kong,
W
.
Gao
,
Y.
She
n,
and
Z
.
Ji
,
“
Im
prove
d
ch
ic
k
en
s
warm
opti
m
iz
ati
on,
”
2015
,
pp
.
6
81
–
686.
[25]
C.
Ts
a
i,
“
A
new
h
y
br
id
h
eur
isti
c
appr
oa
ch
for
solving
l
arg
e
tr
aveling
sal
esm
an
pr
oble
m
*1,
”
In
f.
S
ci
.
,
vo
l.
166
,
no
.
1
–
4,
pp
.
67
–
81
,
Oct.
2004
.
[26]
A.
Puris,
R.
Bel
l
o,
Y.
Martí
n
ez
,
and
A.
Now
e,
“
Two
-
Stage
Ant
Colon
y
Optimi
z
at
ion
for
Solvin
g
the
Tra
v
el
ing
Sale
sm
an
Pr
oble
m
,
”
in
Natur
e
Inspire
d
Problem
-
Solving
Methods
in
Kno
wledge
Engi
ne
eri
ng,
v
ol.
4528,
J.
Mira
and
J.
R.
Álvar
e
z,
Eds.
B
erlin,
Heidelbe
rg:
Sprin
ger
Ber
li
n
Heidelbe
rg, 2007, pp.
307
–
316.
[27]
S.
-
M.
Soak
and
B.
-
H.
Ahn,
“
New
Gene
tic
Cro
ss
over
Opera
tor
for
the
TSP
,
”
i
n
Artifi
cial
Int
ellige
n
ce
and
Soft
Com
puti
ng
-
ICAIS
C
2004,
vol.
3070,
L
.
Rutk
ows
ki,
J.
H.
Sie
km
ann,
R.
T
adeus
ie
wicz,
and
L
.
A.
Z
ade
h,
Eds
.
Berl
in
,
He
ide
lb
e
rg:
Springer
Ber
l
in
Heid
el
ber
g
,
2
004,
pp
.
480
–
48
5.
[28]
I.
-
H.
Kuo,
S.
-
J.
Horng,
T.
-
W
.
K
ao,
T
.
-
L
.
Li
n
,
C.
-
L.
Lee,
Y.
-
H
.
Chen,
Y.
Pan
,
a
nd
T.
T
era
no
,
“
A
h
y
brid
sw
arm
int
ellige
n
ce
al
go
rit
hm
for
the
tr
a
vel
li
ng
s
al
esm
an
proble
m
:
A
h
y
b
rid
sw
arm
int
el
l
i
genc
e
al
gori
thm
for
the
tr
avelli
ng
sale
sm
an
probl
e
m
,
”
Exp
ert
S
y
st.
,
vol
.
27
,
no
.
3
,
p
p.
166
–
179
,
Jun.
2010.
Evaluation Warning : The document was created with Spire.PDF for Python.