Indonesi
an
Journa
l
of El
ect
ri
cal Engineer
ing
an
d
Comp
ut
er
Scie
nce
Vo
l.
1
3
,
No.
2
,
Febr
uar
y
201
9
, pp.
634
~
642
IS
S
N: 25
02
-
4752,
DOI: 10
.11
591/ijeecs
.v1
3
.i
2
.pp
634
-
642
634
Journ
al h
om
e
page
:
http:
//
ia
es
core.c
om/j
ourn
als/i
ndex.
ph
p/ij
eecs
Defense
behavio
r of r
eal t
ime str
ategy g
am
es: comp
ar
is
on
between
HFSM
an
d FSM
Rahma
t
F
au
z
i
1
, Moch
am
ad
Ha
ri
adi
2
, Sup
eno M
ardi S
u
siki
N
u
gr
oh
o
3
,
M
u
ha
r
ma
n L
ubis
4
1,4
School
of
Indu
stria
l
Engi
n
ee
r
in
g,
T
el
kom
Unive
rsit
y
(T
el
-
U),
I
n
donesia
2,3
Facul
t
y
of
Ind
ustria
l
Engi
n
ee
r
i
ng,
Insti
tut Te
kn
ologi
Sepu
luh
N
ovember
(IT
S),
I
ndonesia
Art
ic
le
In
f
o
ABSTR
A
CT
Art
ic
le
history:
Re
cei
ved
J
ul
24
, 2
018
Re
vised
N
ov
2
1
, 2
018
Accepte
d
Dec
3
, 2
018
RTS
Gam
e
is
on
e
of
th
e
popula
r
genr
e
in
PC
g
aming,
which
has
b
ee
n
play
e
d
b
y
var
ious
t
y
p
e
of
pl
a
y
ers
f
req
u
ent
l
y
.
In
RTS
g
ame,
NP
C
De
fe
nse
Buil
d
ing
(Towe
r)
has
at
t
ac
king
beh
avi
or
to
th
e
cl
osest
ene
m
y
withou
t
conside
rin
g
ce
rt
ai
n
ene
m
y
pa
ramet
ers.
Thi
s
c
ause
s
the
NP
C
T
ower to
b
e m
ore
pre
dictable
b
y
the oppon
ent
and ea
sil
y
def
eat
ed
if
NP
C
a
tt
a
ck
ed
b
y
en
emies
in
the group
.
Thus,
th
is
rese
a
rch
sim
ula
t
es
N
PC
Towe
r
usin
g
Hier
ar
chica
l
Finit
e
St
ate
Mac
hine
(HF
SM
) m
et
hod
compar
ed with F
inite S
t
at
e M
a
chi
n
e (F
SM
).
In thi
s
stud
y
,
NP
C
Towe
r
detec
ts
ene
m
i
es
b
y
se
ei
ng
at
f
our
par
amet
ers
n
amel
y
NP
C
Towe
r
Hea
lt
h,
E
nem
y
'
s
Hea
l
th,
E
nem
y
T
y
p
e,
and
Towe
r
Distan
ce
to
ene
m
ie
s
.
NP
C
Towe
r
wi
ll
at
t
ac
k
th
e
m
ost
d
ange
rous
ene
m
y
ac
cor
d
ing
to
the
‘Degre
e
o
f
Dange
r’
par
amete
r.
The
n
use
t
he
d
ecision
-
m
ak
ing
logic
of
th
e
rule
-
base
d
s
y
stem.
The
ou
tput
of
NP
C
T
ower
ar
e
three
t
y
p
e
of
b
eha
v
iors
name
l
y
Aggress
ive
Attacking,
Regular
A
tt
a
cki
ng,
and
At
t
ac
k
wi
th
Spec
ia
l
Skill
.
From
the
te
st
resul
ts
of
3
NP
C
Towe
r
,
Kam
anda
ka NP
C
Towe
r
with
H
FS
M
m
et
hod
is
winning
8
.
92
%
compar
e
to
Kam
anda
ka
To
wer
with
FS
M
m
et
hod.
For
Ga
y
at
ri
Towe
r
NP
C
obta
ine
d
e
qual
result
s
usi
ng
both
HF
SM
and
FS
M.
Mea
nwhile,
Adi
kar
a
NP
C
with
HF
S
M
m
et
hod
i
s
4.
62%
superio
r
to
Adikar
a
Towe
r
with
FS
M m
et
hod.
Ke
yw
or
d
s
:
Com
par
ison
Def
e
ns
e
be
havi
or
Finit
e stat
e m
a
chine
Re
al
tim
e strat
egy
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
:
Ra
hm
at
Fau
zi
,
School
of In
dustria
l En
gin
ee
ring
,
Tel
ko
m
U
ni
versi
ty
(
Tel
-
U)
,
Jal
an
Tel
e
ko
m
un
i
kasi
No.
1,
Ba
ndung,
4025
7
,
I
ndonesi
a.
Em
a
il
:
rah
m
a
tfauzi@te
lk
om
un
ive
rsity
.ac.id
1.
INTROD
U
CTION
Re
al
-
Ti
m
e
Strate
gy
(RTS)
i
s
a
genre
ga
m
e
that
has
a
ty
pical
war
gam
e
con
sist
in
g
of
buil
ding
const
ru
ct
io
n
a
nd
tro
op
stre
ng
t
h.
Re
s
ourc
e
colle
ct
ion
i
s
us
e
d
for
buil
ding
c
onstr
uction,
def
e
nse
an
d
stren
gth
e
ning
com
bat
troops.
The
RTS
ga
m
e
con
side
rs
aspects
of
e
c
onom
y,
resour
c
es
possesse
d,
tro
ops
’
stren
gth
,
at
ta
ck
ing
a
nd
def
e
ns
i
ng
strat
egies
.
T
his
ki
nd
of
ge
nre
ha
s
beco
m
e
extrem
el
y
po
pula
r
due
to
the
rising
of
fam
ou
s
A
ge
of
Em
pire,
Wa
rcr
a
ft
a
nd
D
ota
.
Cu
rr
e
ntly
,
on
e
fam
ou
s
exam
ple
of
RTS
ga
m
es
is
Cl
ash
of
Cl
ans
(C
oC),
w
hich
was
de
velo
pe
d
by
S
up
e
rcell
besides
oth
e
r
known
gam
e,
Boo
m
Be
ach
(
BB
)
that
play
e
d
by
thousa
nd
s
of
people
in
m
obil
e
phone
plat
form
.
In
C
oC
gam
es,
there
are
N
on
-
Play
a
ble
C
har
act
e
r
(N
PC
)
Def
e
ns
e
B
uilding
s
li
ke
N
PC
Ca
nnon,
Mo
rta
r,
A
rc
her
T
ow
er
an
d
W
i
za
r
d
To
wer
.
Ca
non
and
M
or
ta
r
at
ta
ck
the
enem
y
on
t
he
gro
und
w
hile
Archer
T
ow
e
r
a
nd
Wizar
d
To
w
er
s
hoot
enem
i
es
in
the
gr
ound
a
nd
water
(ai
r)
area
.
In
bo
t
h
CoC
and
BB
,
NP
C
be
ha
vior
ha
ve
def
e
ns
ive
pa
tt
ern
i
n
w
hic
h
do
strike
t
hro
ugh
at
ta
ck
near
est
.
Unfortu
nate
ly
,
this
ty
pe
of
str
ike
is
easi
ly
pe
netrated
or
a
ntici
pated
by
op
pone
nt.
Mo
reles
s,
if
t
hey
a
re
unit
ed
by
at
ta
cking
i
n
the
gro
up.
Th
eref
or
e
,
this
ki
nd
of
fixe
d
m
od
el
f
or
NP
C
To
wer
beh
a
vior
create
s
the
figh
t
or
course
i
n
the
gam
e
beco
m
e
less
va
ry
or
pre
di
ct
a
ble
to
the
oppo
nen
t.
To
overc
om
e
this
issues,
one
s
olu
t
ion
is
thr
ough
dynam
ic
NP
C
be
ha
vio
r
with
m
or
e
di
ver
se
an
d
le
ss
predict
a
ble
at
ta
ck
patte
r
n,
w
hich
can
be
ac
hiev
e
d
by
the
ad
aptive
and
i
ntell
igent
natu
re
of
the
be
hav
i
or
patte
r
n.
The
refor
e
,
dynam
ic
beh
av
io
r i
s
a
ty
pe o
f p
at
te
rn
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
Defense
b
e
havi
or
of re
al ti
me st
ra
te
gy g
am
e
s
:
c
omp
ar
iso
n b
et
we
en
HF
SM
and FSM
(
Ra
hmat
Fa
uz
i
)
635
that
ta
kes
into
account
se
ver
a
l
factor
s
t
hat
infl
uen
ce
t
he
de
ci
sion
m
aking
of
NP
C.
For
e
xam
ple
NP
C
com
bat
def
e
ns
e
will
not
al
ways
us
e
strong
at
ta
cks
that
re
quire
en
or
m
ou
s
ene
r
gy
if
the
healt
h
c
onditi
ons
a
re
in
go
od
conditi
on
or
do
not
al
ways
use
norm
al
at
ta
c
ks
t
hat
require
le
ss
ene
r
gy
if
healt
h
c
onditi
ons
are
wea
k.
In
sho
rt,
the syst
em
sh
ould
exhibit
t
he whole
range
of
f
le
xi
bili
ty
si
m
ply as a
r
e
su
lt
of cha
nges i
n
t
he NPC
att
rib
ut
es for
par
ti
cula
r gam
es that m
igh
t r
ai
se certai
n p
ossi
bili
ti
es.
In
ge
neral
,
Fi
ni
te
Stat
e
Ma
ch
ine
(
FSM)
is
c
omm
on
m
e
thod
a
pp
li
e
d
in
th
e
gam
e
to
det
erm
ine
the
beh
a
vior
beca
us
e
it
s
sim
plicity
to
be
im
ple
m
ented
in
va
riou
s
ty
pe
of
ga
m
es,
al
though
at
certai
n
po
i
nt
it
got
hindere
d i
f t
he
agen
t
has m
ultip
le
ap
proac
he
s
in
the
N
PC
be
hav
i
or
[1
]
.
It
al
so
ha
s
ne
gative
po
i
nts
due
to
i
t
has
a
la
rg
e
num
ber
of
sta
te
s
an
d
t
ran
sit
io
ns
s
o
th
at
representat
i
on
an
d
a
naly
sis
bec
om
e
diff
ic
ult
[2,
9].
I
n
orde
r
to
desig
n
t
he
dynam
ic
beh
avi
or
of
NP
C
that
has
certai
n
c
har
act
erist
ic
of
beh
a
vior,
it
is
sugg
e
ste
d
t
o
us
e
Hierarc
hical
Fi
nite
Stat
e
Ma
c
hin
e
(
HF
SM
)
m
et
ho
d
to
im
prov
e
the
pr
ocess
to
determ
ine
th
e
be
ha
vio
ral
re
sp
onse
in
the
fluctuat
e
cha
nges
of
conditi
on,
w
hi
ch
us
e
d
certai
n
kind
of
r
ule
syst
em
.
Me
anwhil
e,
HFS
M
as
a
dev
el
op
m
ental
m
e
th
od
of
FS
M
has
t
he
a
dv
antage
of
redu
ci
ng
t
he
c
om
plexity
of
desi
gnin
g
a
gen
t
be
ha
vior,
wh
ic
h
can
sim
plify
the
com
pu
ta
ti
on
al
pr
oce
ss
an
d
kee
p
the
con
st
raint
by
m
ini
m
iz
ing
a
num
ber
of
c
odin
g
bits
and
ad
justi
ng
t
he c
od
e
s
f
or
th
e
sta
te
s
[2,
3].
A
detai
le
d an
d
rea
li
sti
c
si
m
ula
ti
on
of
dynam
ic
be
hav
i
or
re
quires
a
pr
eci
se
m
od
el
ing
of
de
ci
sion
-
m
aking
pr
oces
s
re
ga
rd
i
ng
act
ion
pri
ori
ti
zat
i
on
a
nd
sel
ect
io
n,
w
hich
is
c
ap
able
to
dev
el
op
go
al
di
rected
be
hav
i
or
al
te
ndencies
an
d
ta
ct
ic
al
m
od
el
ba
sed
on
dem
and
s
a
nd
i
nput
pa
ram
e
te
r
[
4].
Im
po
rtantl
y,
in
form
ation
is
a
basic
requirem
ent
f
or
the
devel
op
m
ent
proc
ess
as
a
par
t
of
the
ec
os
yst
em
[5
]
,
wh
ic
h
can
be
util
iz
ed
le
ad
to
the
s
ub
sta
ntial
saving
s
i
n
powe
r
co
nsum
ption
[6
]
,
differentia
te
disti
nc
ti
ve
at
tribu
te
s
of
be
hav
i
or
[
7],
de
vel
op
c
ollab
orat
ion
at
ta
ct
ic
al
le
vel
[8,
20
]
,
et
c.
In
this
stud
y,
the
res
e
arch
e
r
dev
el
op
e
d
RT
S
gam
e
entit
led
D
WI
P
A
Y
U
DHA,
in
w
hic
h
it
s
c
oncept
ha
s
sam
e
ty
pe
a
nd
re
fers
t
o
t
he
RTS
genre
li
ke
Cl
ash
of
Cl
a
ns
.
It
ha
s
se
ver
al
c
ompone
nts,
w
hich
are
(1)
Tr
oops
(
2)
C
om
bat
D
efen
se
B
uildin
g
an
d
(3)
Re
s
ource
s.
Like
wise,
thi
s
re
searc
h
al
so
de
vel
op
D
WIPA
Y
U
DHA
to
ha
ve
th
os
e
c
om
po
ne
nts,
bu
t
this
researc
h
em
phasi
zed m
or
e on
b
e
hav
i
or of
N
PC Com
bat D
e
fen
se
Buildi
ng.
2.
RESEA
R
CH MET
HO
D
Im
po
rtantl
y,
the
m
et
ho
d
sta
ges
determ
ine
the
su
cce
ss
of
the
researc
h
to
assu
re
it
s
validit
y
and
reli
abili
ty
[1
6
-
19,
24
-
31]
.
Fig
ur
e
1
is
a
flo
w
char
t
t
hat
desc
ribe
t
he
detai
l
of
the
ste
ps
un
der
ta
ken
to
co
m
ple
te
this
rese
arc
h,
w
hich
is
incl
ude
NP
C
A
gen
t
De
sign,
Desig
n
of
En
vir
onm
ent
,
Desig
n
of
E
ne
m
y
,
Desig
n
of
HF
S
M
for
NP
C
Ag
e
nt
and
Desig
n of
FSM
f
or
E
nem
ie
s
, th
e
n
c
on
ti
nue
with
var
i
ous test
scena
rios.
Figure
1.
Ef
fec
ts of sel
ect
ing
diff
e
re
nt sw
it
c
hing
unde
r dyn
a
m
ic
co
nd
it
io
n
2.1.
Desi
gn o
f
N
P
C
T
ower
Ag
e
nt
The
f
our
NP
C
Com
bat
Def
e
ns
e
are
NP
C
T
ow
e
r
Kam
andaka,
Gayat
ri
N
PC
T
ower,
NPC
Ga
ndewa
To
wer
a
nd
NPC
Ad
i
kar
a
. E
ac
h
NP
C
T
ower
has se
ve
ral
pa
r
a
m
et
ers t
hat
ch
aracte
rize
them
su
c
h as
the
val
ue
of
Dam
ag
e
,
Att
ac
k
Sp
ee
d
,
Att
ack
Ra
ng
e
,
Hitp
oi
nt
,
Att
ack
Ty
pe
and
N
PC
Beh
avior
.
Th
e
val
ue
s
of
e
ach
para
m
et
er
of each
NPC
T
ow
e
r
a
re
descr
i
bed in
Table
1
.
Desig
n
of
NPC Agen
t
Desig
n
of
E
n
e
m
y
Desig
n
of
Env
iron
m
en
t
Scen
ario
Desig
n
of
HFSM
f
o
r
NPC Ag
en
t Beh
av
io
r
Desig
n
of
FSM
f
o
r
Ene
m
y
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.
1
3
, N
o.
2
,
Fe
bru
ary
201
9
:
634
–
642
636
Table
1.
NP
C
To
wer
Stat
ist
ics
No
Para
m
eter
NPC
Ka
m
an
d
ak
a
Gan
d
ewa T
o
we
r
Gay
at
ri
Tower
Ad
ik
ara
1
Da
m
ag
e
10
7
13
18
2
Attack
Speed
0
.9 s
0
.5 s
1
.2 s
4
s
3
Attack
Ran
g
ed
10
11
8
15
4
Attack
T
y
p
e
Sin
g
le Attack
Sin
g
le Attack
Sp
lash
Ar
ea
Sp
lash
Ar
ea
5
Hit Poin
t
450
325
570
400
6
NPC Beh
av
io
r
No
r
m
al
Attack,
Ag
g
ressiv
e Attack,
an
d
Uniq
u
e Skill
No
r
m
al
Attack,
Stron
g
Attack,
an
d
Uniq
u
e Skill
No
r
m
al
Attack,
Stron
g
Attack, and
Un
iq
u
e Skill
No
r
m
al
Attack,
an
d
Str
o
n
g
Attack
2.2.
Desi
gn o
f
N
P
C
Ene
my
Ag
e
nt
To
c
reate
NP
C
To
wer
as
a
C
om
bat
Def
e
ns
e
Buil
ding,
this
stud
y
al
s
o
desi
gn
e
d
t
he
E
nem
y
Ag
e
nt
to
te
st
the
stren
gth
of
NP
C
T
ower
to
s
urvive
i
n
spe
ci
fic
sit
ua
ti
on
.
T
here
are
three
(
3)
ty
pe
s
of
Enem
y
fo
r
NP
C
Ag
e
nts,
wh
ic
h
are
NP
C
E
ne
m
y
of
Sw
or
ds
t
ha
t
belong
to
t
he
cat
egory
of
str
ong
e
nem
ie
s,
NP
C
E
nem
y
of
Arro
ws
are
cat
e
gorized
as
m
od
erate
e
nem
ie
s
and
N
PC
E
nemy
of
Ma
gic
wh
ic
h
is
a
wea
k
ty
pe
of
e
nem
y.
Ag
e
nt
NP
C
Ag
e
nt
has
se
ve
ral p
a
ram
et
ers
su
c
h
as
d
am
ag
e, m
ov
em
ent sp
eed
(
M
S)
,
att
ack
range
, Hit
point.
F
ull sta
ti
sti
cs
of
Ag
e
nt
NP
C
Agent p
a
ram
et
er v
al
ues
are
at
ta
ched (
See Ta
bl
e 2).
Table
2.
NP
C
Enem
y St
at
isti
cs
No
Para
m
eter
Ene
m
y
of
Sword
s
Ene
m
y
of
Ar
rows
Ene
m
y
of
M
ag
ic
1
Da
m
ag
e
10
7
5
2
Ar
m
o
r
5
2
2
3
Hitp
o
in
t
60
40
40
4
Attack
sp
eed
1
/s
1
/s
1
/s
5
MS
1
2
4
6
Tr
ain
in
g
ti
m
e
3
0
s
2
0
s
3
0
s
7
Attack
r
an
g
e
1
4
4
2.3.
Desi
gn o
f
HFS
M NP
C T
ower
This
resea
rc
h
us
es
Hierarc
hi
cal
Finit
e
State
Ma
chine
(
H
FSM)
m
et
ho
d
to
desig
n
NPC
beh
a
vior.
Be
sides
that,
it
al
so
us
e
d
R
ul
e
Ba
sed
Syst
e
m
as
a
decisi
on
m
aking
f
or
NP
C
w
hen
fi
ghti
ng
with
var
i
ou
s
ty
pes
of
e
nem
ie
s
who
at
te
m
pt
to
e
nt
er
the
ra
nge
of
t
he
NP
C
T
ower
.
In
a
dd
it
io
n,
H
FSM
NP
C
T
ow
e
r
is
sym
bo
li
zed
by a circle
a
s a
sig
n of
a
stat
e,
which t
he
a
rro
w
li
ne
a
s a si
gn
of inter
-
sta
te
trans
fer
a
s s
hown in Fi
gure
2.
Figure
2.
H
FS
M Ag
e
nt
NP
C
To
wer
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
Defense
b
e
havi
or
of re
al ti
me st
ra
te
gy g
am
e
s
:
c
omp
ar
iso
n b
et
we
en
HF
SM
and FSM
(
Ra
hmat
Fa
uz
i
)
637
2.4.
N
PC
T
ow
er Be
havior
Design
The
be
hav
i
or
of
NP
C
C
om
bat
Def
e
ns
e
is
desi
gn
e
d
us
i
ng
r
ule
-
base
d
decisi
on
-
m
aking
m
et
h
od
s
[
9
-
13]
.
In
this
ca
se,
t
he
ru
le
-
base
d
has
seve
ral
pa
ram
et
ers
that
infl
uen
ce
the
de
ci
sion
m
akin
g
process
s
uch
as
t
he
dista
nce
, t
he
e
nem
y
ty
pe
, t
he
NPC
'
s
li
fe
and
the
ene
my
'
s
li
fe
. M
ean
wh
il
e,
the
N
PC
be
ha
vi
or
a
re
Str
ong
Att
ack
,
Normal At
tack
and
T
hr
owi
ng
Trap
. I
n
a
dd
it
ion, the
NPC
C
om
bat D
efe
ns
e
Beha
vio
r
cons
ist
o
f:
1.
Stron
g
Atta
ck
.
It
is
the
at
ta
ck
ing
beh
a
vi
or
of
N
PC
with
it
s
fu
ll
of
a
vaila
bl
e
power
i
n
w
hi
ch
NP
C
T
owe
r
can
destr
oy enem
y fo
rces
on a
larg
e
scale
or r
e
du
ce
the live
s of e
nem
y’s h
ero in
gr
eat
da
m
ages.
2.
Norm
al
Atta
ck.
It
is
a
NP
C
at
ta
ck
be
hav
i
or
with
m
od
e
rate
stren
gth
,
wh
ic
h
NP
C
T
ow
e
r
can
only
destr
oy
enem
ie
s o
n
a
s
m
al
l scal
e o
r
lo
wer dam
age com
par
e to str
ong
at
ta
ck.
3.
Un
i
qu
e
S
kill
.
I
t
is
the
at
ta
cki
ng
be
hav
i
or
wit
h
s
pecial
s
kill
s
that
on
ly
ce
rtai
n
N
PC
has,
w
hich
this
kind
of
beh
a
vior
bec
om
es secret wea
pon
i
n
s
pecific
conditi
on a
nd to
s
pecific t
ar
ge
t [14,
15].
4.
Sil
ence
.
It
is
t
he
NP
C'
s
be
ha
vior
w
hen
it
s
conditi
on
is
in
preca
rio
us
co
nd
it
io
n
or
be
sq
ue
ezed
that
is
un
li
kely
to
w
i
n. T
his p
a
rtic
ula
r beha
vior
on
ly
for
NP
C
Arc
he
r To
we
r
a
nd
Wi
za
rd T
ower
.
2.5.
Desi
gn o
f
R
ule
-
based
N
PC
T
ower
Atta
ck
C
on
t
ro
l
le
r
has
a
functi
on
f
or
co
ntr
olli
ng
the
at
ta
ck
be
hav
i
or
of
NPC
To
wer.
I
n
thi
s
stu
dy,
it
us
ed
ru
le
-
base
d
lo
gic
[
21
-
23
]
t
o
det
erm
ine
the
NPC
To
we
r
be
havi
or
,
w
hic
h
is
e
m
bed
ded
in
St
at
e
Co
ntro
ll
er
Atta
ck.
By
giv
en
t
he
r
ule
-
based
knowle
dge,
eac
h
NP
C
can
res
pond
t
o
the
c
hanges
in
i
npu
t
va
riables
into
be
hav
i
or
s
that
ha
ve
bee
n
desig
ne
d
t
o
us
e
HFSM
.
From
each
bu
il
di
ng
com
bat
de
fen
se
,
t
he
NPC
has
4
(
f
our)
ty
pe
par
am
et
ers
tha
t
bec
om
e
ru
le
-
base
d
i
nput,
w
hich
are
N
PC
bu
il
di
ng
hea
lt
h
,
e
nem
y
hea
lt
h
,
distance
a
nd
ene
m
y
status
.
Me
an
w
hile,
the
ou
t
put
from
ru
le
-
bas
ed
of
NP
C
C
om
bat
Def
ense
beh
a
vior
t
hro
ugh
3
(thr
ee
)
ty
pe
of
respo
ns
es
to
wa
rd
s
the
e
nem
y
i
n
t
he
f
or
m
of
A
ggressi
ve
(S
tr
ong
Atta
ck),
Ordina
ry
(
N
or
m
al
Atta
ck
)
a
nd
S
pecial
(Uniq
ue
s
kill
).
Each
of
th
os
e
par
am
et
ers
ha
s
an
inter
val
r
ang
e
of
value
s
,
w
hich
is
div
ided
i
nto
th
ree
par
ts
nam
ely
Weak,
Mo
dera
te
a
nd
Stron
g
f
or
th
e
To
wer
a
nd
En
e
m
y
Life
'
s
li
fe
par
am
et
ers.
Me
anwhil
e
the
di
sta
nc
e
par
am
et
er
has
on
ly
tw
o
pa
rts
nam
ely
the
sh
ort
an
d
lo
ng
dist
ance
pa
ram
et
ers.
The
pa
ram
eter
s
of
inte
rv
al
s
sh
ow
n
in Ta
ble 3
,
4
a
nd
5.
Table
3.
Param
et
er Inter
val T
ow
e
r
Li
fe
Interval Valu
e
No
tatio
n
0
–
33
W
eak
34
–
66
Mod
erate
67
–
100
Stron
g
Table
4.
Param
et
er Inter
val E
nem
y Li
fe
Interval Valu
e
No
tatio
n
0
–
33
W
eak
34
–
66
Mod
erate
67
–
100
Stron
g
Table
5.
Param
et
er Inter
val E
nem
y Dist
ance
Interval Valu
e
No
tatio
n
0
–
50
Near
51
–
100
Far
3.
RESU
LT
S
A
ND AN
ALYSIS
NP
C
C
o
mb
at
D
e
fens
e
Bui
ld
ing
s
th
at
h
av
e
b
een
de
si
gn
e
d
an
d
d
e
sc
ri
b
e
d
in
p
r
evi
ou
s
ch
a
pt
er
a
r
e
impl
e
m
ent
e
d
in
si
mul
ati
o
ns
usi
ng
U
nity
G
a
me
E
ngi
n
e.
Thi
s
t
est
is
ai
me
d
t
o
e
v
al
uat
e
t
h
e
d
e
ve
lop
e
d
ap
p
ro
a
ch
m
et
h
od
a
nd
an
al
y
z
e
th
e
r
es
ult
o
bt
ain
e
d.
Th
e
r
e
fo
r
e,
th
e
e
xp
e
ri
me
nt
w
as
do
n
e
b
y
co
m
pa
ri
n
g
FS
M
me
th
od
an
d
H
FS
M
met
h
od
,
w
hi
ch
o
bt
ain
e
d
t
he
ti
m
e
res
ults
a
n
d
t
he
n
um
b
er
o
f
e
n
e
mi
es
kill
e
d.
T
he
r
e
a
r
e
se
v
eral
te
st
sc
en
a
ri
os t
h
at
h
a
ve
b
e
e
n d
o
ne
i
e.
NP
C
Kam
and
a
ka
a
gainst
W
ea
k
te
am
,
Mod
er
at
e
te
a
m
an
d
S
tro
ng team
.
NP
C
Gande
wa
To
wer agai
ns
t
W
ea
k
te
am
,
M
od
e
rate t
eam
and
S
t
ron
g
te
a
m
.
NP
C
Ad
i
kar
a
a
gainst
W
ea
k
te
a
m
,
M
od
e
rate t
ea
m
an
d
S
tr
ong
te
am
.
Th
e
f
ir
st
sc
e
n
a
rio
b
y
t
e
stin
g
1
(
on
e
)
NPC
To
w
e
r
a
g
ain
st
st
ro
ng
t
ro
o
ps
obt
ai
n
ed
th
e
a
ve
r
a
ge
ti
m
e
of
kil
lin
g
1
e
n
em
y
tr
oo
ps
3
.8
1s
o
n
t
esti
n
g
w
ith
H
FS
M
a
nd
3.8
4
5s
w
ith
F
SM
.
M
e
an
w
hi
le,
b
y
a
ddi
ng
1
(
on
e
)
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.
1
3
, N
o.
2
,
Fe
bru
ary
201
9
:
634
–
642
638
NP
C
T
o
w
er
as
a
s
e
c
on
d
s
cen
a
rio
fo
r
t
e
stin
g
2
(t
wo
)
NPC
To
w
e
r
a
g
ain
st
str
on
g
tr
o
ops
obt
ai
n
ed
th
e
a
ve
r
a
ge
of
k
ill
in
g
1
en
em
y
tr
o
op
s
2.
78
s
on
HFS
M
a
n
d
3.
11
45
s
wit
h
FS
M.
I
n
this
t
est
NPC
To
w
e
r
with
H
FS
M
me
th
od
c
an
ki
ll
a
ma
xi
mu
m
of
10
e
ne
m
y
tro
o
ps
w
hil
e
t
he
N
PC
To
w
e
r
with
FS
M
h
av
e
m
axi
m
um
of
9
en
e
my
t
ro
op
s.
Fu
rt
he
r
mo
r
e,
in
t
h
e
t
hir
d
sc
en
a
ri
o,
by
t
est
ing
1
(
on
e
)
N
PC
To
w
e
r
a
g
a
inst
m
od
e
r
at
e
tro
o
ps
obt
ai
n
ed
on
a
ve
r
a
ge
t
im
e
k
ill
in
g
1
e
ne
m
y
tr
o
op
1.
00
5
s
o
n
HFS
M
a
nd
1.
00
75
s
o
n
FS
M
te
st.
T
h
en
,
b
y
ad
di
ng
1
(
on
e
)
NP
C
T
o
w
er
a
s
t
h
e
f
ou
rt
h
s
c
en
a
ri
o,
with
t
e
stin
g
2
(
t
wo
)
NP
C
T
o
w
e
r
a
gai
ns
t
m
od
e
rat
e
tr
o
ops
ha
d
t
h
e
av
e
r
ag
e
ti
m
e
o
f
kill
i
ng
1
t
ro
o
p
1
.6
9s
on
H
FS
M
a
nd
1.
95
s
on
F
SM
t
e
stin
g.
I
n
this
te
st
NP
C
To
w
e
r
wit
h
HF
S
M
m
eth
o
d
c
an
kill
a
m
axi
m
um
o
f
11
en
e
my
tr
o
op
s
w
hil
e
th
e
NP
C
To
w
e
r
w
ith
F
S
M
ma
xi
mu
m
of
1
0
en
e
m
y t
r
oo
ps.
Fo
r
th
e
fi
fth
s
c
en
a
rio
,
th
is
st
ud
y
w
as
t
est
in
g
1
(
on
e
)
NP
C
T
o
w
er
a
g
ai
nst
t
h
e
w
eak
t
ro
op
s
w
hic
h
r
esu
lte
d
th
e
a
v
e
rag
e ti
m
e t
o
kill
1
e
n
em
y
1
.6s
o
n
H
FS
M
an
d
1.
6
05
o
n
FS
M.
O
n
th
e
oth
e
r
h
an
d,
b
y
a
dd
in
g
oth
e
r
1
(
o
ne
)
NP
C
T
o
w
er
a
s
t
h
e
sixt
h
s
c
e
na
ri
o,
wit
h
t
e
s
ting
2
(t
w
o)
N
PC
T
o
w
er
ag
a
inst
th
e
w
e
ak
tro
o
ps
gai
n
ed
t
h
e
a
ve
r
ag
e
tim
e
o
f
k
ill
in
g
1
t
r
oo
p
2.2
7s
on
HFS
M
te
sti
ng
an
d
2.3
4s
on
FS
M
te
sti
ng
.
In
t
hi
s
t
es
t
NP
C
To
w
e
r
wit
h
HF
S
M
me
th
od
h
av
e
c
ap
a
bili
t
y
to
k
ill
a
m
axi
m
u
m
o
f
13
e
n
em
y
tr
o
op
s
whi
le
th
e
NP
C
To
w
e
r
wit
h
F
SM
c
an
r
e
a
ch
until
1
2
e
n
e
my
t
r
oo
ps.
O
n
the
ot
he
r
h
an
d,
i
n
th
e
s
e
v
e
nth
sc
e
n
ar
i
o,
t
esti
n
g
1
(o
n
e)
N
PC
G
a
nd
e
w
a
T
o
w
e
r
ag
ai
nst
str
o
ng
tro
o
ps
h
a
v
e
ob
tai
ne
d
th
e
av
e
r
ag
e
ti
m
e
o
f
kil
ling
1
e
n
em
y
t
ro
op
5.6
3s
wit
h
H
FS
M
an
d
5.6
0s
wit
h
FS
M.
T
he
n,
in
c
r
ease
th
e
nu
m
be
r
wit
h
1
NP
C
To
w
e
r
as
th
e
eig
ht
h
sc
e
na
ri
o,
t
h
en
testi
n
g
2
(t
w
o
)
NP
C
T
o
w
e
r
ag
ai
nst
st
ro
ng
tro
o
ps
got
a
n
av
e
r
ag
e
r
es
ul
t
of
kill
i
n
g
1
e
ne
m
y
tro
o
p
3.
93
s
o
n
HFS
M
t
est
in
g
an
d
4.
5
2s
o
n
t
esti
n
g
wi
th
FS
M.
In
t
his
t
est
NP
C
T
o
w
e
r
w
ith
HFS
M
met
ho
d
c
an
kill
a
m
a
x
imu
m
o
f 8
en
e
my
tr
o
ops
wh
i
le t
h
e
NP
C
To
w
e
r
wit
h F
S
M
m
axi
m
um
o
f
8
en
e
my
f
o
rce
s.
In
t
h
e
ne
xt
sc
e
na
ri
o,
w
hi
ch
i
s
ni
nth
,
t
his
st
ud
y
w
as
t
e
stin
g
1
(
on
e
)
NP
C
G
a
nd
e
w
a
To
w
e
r
ag
ai
nst
mo
d
erat
e t
ro
o
ps
resu
lte
d t
he
av
e
r
ag
e ti
m
e
of
kill
i
ng
1
(o
ne
)
en
e
m
y tr
o
op
4.
27
s o
n t
e
stin
g
wit
h
HF
SM
a
nd
4.1
4s
w
ith
F
S
M.
M
or
e
ov
e
r,
by
s
u
ppl
e
m
e
ntin
g
1
(
on
e
)
NP
C
To
w
e
r
a
s
th
e
t
ent
h
s
c
e
na
ri
o,
t
h
en
t
es
ting
2
(t
wo
)
NP
C
To
w
e
r
a
g
ain
st
m
od
e
r
ate
t
r
oo
ps
gai
n
ed
t
h
e
a
v
e
rag
e
r
e
sult
of
kill
in
g
1
e
n
e
my
t
r
oo
p
2.
7
9
s
on
HF
S
M
t
esti
ng
an
d
3.
23
s
wit
h
FS
M.
In
thi
s
t
est
N
P
C
T
o
w
e
r
wit
h
HFS
M
me
th
od
c
a
n
k
i
ll
a
m
axi
m
um
of
10
en
e
my
t
ro
o
ps,
whi
c
h
h
av
e
e
qu
al
v
al
u
e
wit
h
NP
C
T
o
w
er
wit
h
FS
M.
I
n
t
he
ne
xt
st
ep
, t
he
el
ev
e
nth
s
c
en
a
ri
o
w
as
t
esti
ng
1
(o
n
e)
NP
C
G
a
nd
e
w
a
T
o
w
er
ag
ai
nst
t
he
w
e
ak
t
ro
o
ps
g
ai
n
ed
t
he
av
e
r
ag
e
t
im
e
o
f
kill
i
ng
1
en
e
my
tr
o
op 4
.45
s on
t
e
stin
g
wit
h HF
S
M a
nd 4.
45
s
o
n
t
e
stin
g wit
h
FS
M. A
d
diti
n
g 1
(o
n
e) NP
C
T
o
w
e
r a
s
the
t
w
el
ft
h
S
c
en
a
ri
o,
th
e
n
t
e
stin
g
2
(t
w
o)
NP
C
To
w
e
r
a
gai
ns
t
wea
k
tr
oo
ps
obt
ai
n
ed
the
a
v
era
g
e
r
e
sult
of
kill
i
ng
1
e
n
em
y
tr
oo
p
2.9
2s
on
HF
S
M
te
st
ing
a
nd
3.
93s
wit
h
FS
M
.
I
n
this
t
est
,
NPC
To
w
e
r
wit
h
H
FS
M
me
th
od
c
a
n
ki
ll a
m
a
xim
u
m
of
1
0
en
e
m
y t
r
oo
ps,
s
o
did
th
e
NP
C
T
o
wer
wit
h F
S
M.
In
th
e
t
hi
rt
eent
h
s
c
en
a
ri
o,
t
hi
s
st
ud
y
w
as
te
s
ting
1
(
on
e
)
N
PC
To
w
e
r
A
di
ka
r
a
a
gai
nst
st
ro
ng
tr
o
op
s
obt
ai
n
ed
th
e
a
ve
r
a
ge
o
f
tim
e
to
kill
1 (o
n
e)
e
n
em
y
t
r
oo
p 0.0
3s
wit
h H
F
SM
an
d 0.
13
s wit
h FS
M.
Ad
din
g
1
(o
n
e)
NP
C
T
o
wer
a
s
t
h
e
f
ou
rt
e
ent
h
sc
e
na
ri
o,
th
en
t
e
stin
g
2
(t
wo
)
NP
C
T
o
wer
a
gai
ns
t
st
r
on
g
tro
o
ps
obt
ai
n
ed
th
e
a
ve
r
a
ge
resul
t
of
ki
lli
n
g
1
e
n
em
y
tr
o
op
0.
2
15
5s
o
n
H
FS
M
a
nd
0.4
9
8s
wit
h
FS
M.
I
n
this
t
es
t
NP
C
T
o
w
e
r
w
ith
HFS
M
me
t
ho
d
c
a
n
kill
a
ma
xi
mu
m
o
f
1
6
e
n
em
y
tr
o
op
s
w
hil
e
th
e
NP
C
T
o
w
er
wit
h
FS
M
15
tr
oo
ps
.
M
e
an
w
hil
e,
in
th
e
fi
fte
nt
h
s
c
e
na
ri
o,
a
g
ain
t
esti
n
g
1
(
o
ne
)
N
PC
T
o
w
er
A
dik
a
r
a
ag
ai
n
st
th
e
mo
d
erat
e
tr
o
o
ps
g
ai
ne
d
t
h
e
a
ve
r
a
ge
r
e
sult
o
f
kill
i
ng
1
en
e
my
t
ro
op
0
.0
7
s
wit
h
HF
S
M
an
d
0.
10
7s
wit
h
F
S
M.
O
nc
e
a
g
ai
n,
b
y
ad
di
ng
1
(
o
ne
)
N
PC
T
o
w
e
r
as
th
e
sixt
e
ent
h
s
cen
a
rio
,
th
en
t
esti
ng
2
(t
wo
)
N
PC
To
w
e
r
ag
ai
nst
mo
d
erat
e
t
r
oo
ps
o
bt
a
ine
d
th
e
av
e
rage
resu
lt
of
kil
ling
1
en
e
m
y
t
ro
op
0.4
2s
on
HF
S
M
t
es
t
an
d
0.
5
3
wit
h
F
S
M. In t
his
t
e
st NP
C T
o
wer w
ith H
F
SM a
nd
FS
M me
th
od
c
a
n ki
ll
a m
ax
im
um
of 2
1 en
e
my tr
oo
ps.
In
th
e
se
v
en
te
ent
h
s
cen
ari
o,
th
e
t
est
a
lso
us
e
on
e
N
PC
To
w
e
r
A
di
ka
r
a
ag
ai
nst
t
h
e
w
e
ak
tr
o
ops
,
w
hi
ch
ob
tai
n
e
d
th
e
a
ve
r
a
ge
t
im
e
o
f
kill
i
ng
1
e
n
e
my
t
r
oo
p
s
0.
6
4s
wit
h
H
FS
M
a
n
d
0
.9
9
s
with
FS
M.
A
ddi
ng
1
(
on
e
)
NP
C
T
o
wer
as
t
h
e
ei
ght
e
en
th
sc
e
n
a
rio
,
th
e
n
b
y
t
es
ting
2
(t
wo
)
N
PC
To
w
e
r
ag
a
inst
t
h
e
w
eak
t
ro
op
s
obt
ai
n
ed
t
h
e
a
ve
r
a
ge
tim
e
o
f
kill
i
n
g
1
e
ne
my
tr
oo
ps
0.
7
5s
on
H
FS
M
t
esti
n
g
an
d
1
.1
4s
wit
h
F
S
M.
In
thi
s
test
NP
C
T
o
w
e
r
wit
h
HFS
M
m
et
ho
d
can
k
ill
a
m
a
xim
u
m
o
f
2
8
e
n
em
y
t
ro
op
s
whil
e
t
h
e
NPC
To
w
e
r
wit
h
FS
M als
o 27 t
ro
op
s. Fu
rt
he
r
mo
r
e,
t
h
e res
u
lt
o
f c
o
m
p
ar
i
s
on te
st b
et
w
een
FS
M
a
n
d H
FS
M
m
et
h
od f
ro
m 18
(
eig
ht
e
ent
h)
sc
en
a
ri
os
o
f
eac
h
3
NP
C
T
o
w
e
r
a
g
ain
st
3
(th
r
e
e)
t
yp
e
of
t
ro
op
s
w
e
r
e
p
r
es
e
nte
d
i
n
g
rap
h
f
or
ms
as
sh
o
w
n i
n t
h
e Fi
g
ur
e
3,
Fi
g
ur
e
4
a
nd
Fi
gu
r
e 5
.
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
Defense
b
e
havi
or
of re
al ti
me st
ra
te
gy g
am
e
s
:
c
omp
ar
iso
n b
et
we
en
HF
SM
and FSM
(
Ra
hmat
Fa
uz
i
)
639
Figure
3.
Sim
ulati
on
Result
s
of 2 NPC
K
am
and
a
ka
a
gai
ns
t 3 Ty
pe of
Tro
op
s
The
first
te
st
usi
ng
1
NP
C
T
ower
kam
and
a
ka
an
d
3
ty
pes
of
e
nem
y
troops
that
ha
ve
diff
e
ren
t
at
ta
c
k
powe
rs
nam
ely
stron
g
tro
op
s
(
sw
ord
),
m
od
erate
tr
oops
(
arrow
s
)
an
d
w
eak
tro
ops
(m
agic).
In
the
c
om
bat
def
e
ns
e
c
onditi
on
s
that
only
use
1
(
on
e
)
T
ow
er
ob
ta
ine
d
a
ba
la
nced
re
s
ult
betwee
n
HF
S
M
an
d
F
SM
m
et
hod.
It
was
fou
nd
that
sin
gle
NPC
Tow
e
r
was
able
to
withsta
nd
6
(six)
str
ong
t
roo
p
at
ta
ck
s.
F
or
m
od
erat
e
tro
op
s
,
sing
le
N
PC
T
ower
wa
s
a
ble
t
o
withsta
nd
a
m
od
erate
at
ta
ck
by
m
axi
m
um
8
(eig
ht)
tr
oops.
As
f
or
t
he
wea
k
tro
op
s
,
a
gai
n
s
ing
le
NP
C
T
ower
was
a
ble
t
o
s
urvive
from
9
(
ni
ne)
tr
oops.
Me
an
wh
il
e,
by
ad
ding
on
e
NPC
Kam
and
aka
To
wer
pro
vid
es
be
tt
er
com
bat
de
fen
se
res
ults
in
the
H
FSM
m
eth
od
c
om
par
ed
to
the
F
SM
m
e
thod
.
In
t
he
com
bat
def
e
ns
e
c
ondit
ion
with
2
(tw
o)
NP
C
T
ower
with
H
FSM
c
an
with
sta
nd
a
tt
ack
from
10
strong
tro
op
s
,
wh
ic
h
i
s
10%
m
or
e
superi
or
rathe
r
t
ha
n
T
ower
with
FSM.
O
n
t
he
oth
e
r
hand,
T
ower
with
H
FS
M
can
withstan
d
a
n
at
ta
ck
of
11
m
od
erate
tro
ops,
w
hich
is
9.0
9%
m
or
e
su
pe
rior
com
par
e
to
To
wer
with
FSM
.
At
la
st,
To
wer
wit
h
H
FSM
a
ble
to
hold
13
wea
k
t
r
oops
that
is
7.
69%
m
or
e
s
up
erior
c
om
par
ed
to
To
we
r
with
FSM.
Fr
om
these
3
(t
hr
ee
)
te
sts,
an
a
ver
a
ge
sc
or
e
of
8.92
%
was
ob
t
ai
ned
f
or
H
FS
M
m
or
e
superi
or
c
om
par
ed
to
FSM
m
et
ho
d
.
In
the
ne
xt
te
st,
the
seco
nd
one
us
i
ng
1
(
on
e)
Gayat
ri
To
wer
NP
C
an
d
3
ty
pes
of
ene
m
y
fo
rces
tha
t
hav
e
dif
fer
e
nt
at
ta
ck
powe
r,
wh
ic
h
are
str
ong
(s
wor
d)
,
tr
oops
(a
rro
ws)
,
tro
ops
w
eak
(m
agic).
I
n
the
c
om
bat
def
e
ns
e
c
onditi
on
s
that
only
use
1
T
ow
e
r,
it
has
ob
ta
ine
d
a
n
eq
ual
re
su
lt
betwee
n
H
FS
M
m
et
ho
ds
a
nd
FSM
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.
1
3
, N
o.
2
,
Fe
bru
ary
201
9
:
634
–
642
640
m
et
ho
d.
A
fter
wards,
it
was
f
ound
that
1
NPC
To
wer
was
a
ble
to
withstan
d
the
at
ta
c
k
of
3
(t
hr
ee
)
str
ong
tro
ops
.
Fo
r
t
he
m
od
e
r
at
e
tro
ops,
1
N
PC
T
ow
e
r
is
a
ble
to
withstan
d
t
he
at
ta
cks
f
r
om
4
(fo
ur)
tr
oops.
As
for
t
he
wea
k
tro
op
s
,
1
NP
C
Tow
e
r
able
to
su
r
v
ive
from
5
tro
ops.
T
he
add
it
io
n
num
ber
of
NP
C
Ga
ya
tri
Tow
e
r
pr
ov
i
ded
balance
d
def
e
nse
res
ults
bet
w
een
the
H
FSM
com
par
ed
t
o
t
he
F
SM.
As
a
resu
lt
,
Gayat
ri
towe
r
with
H
FSM
m
et
ho
d
is
supe
rior
in
te
rm
s
of
the p
aram
et
er
of
t
he
rest o
f
li
fe
afte
r
fi
gh
ti
ng
a
gainst
t
he
e
nem
y
co
m
par
e
to
the
Gayat
ri
To
wer
with
FSM
m
e
thod.
At
la
st,
the
thir
d
te
st
usi
ng
1
NP
C
Adikara
T
ow
e
r
a
nd
3
ty
pes
of
enem
y
tro
op
s
that
al
s
o
ha
ve
dif
fer
e
nt att
ack
po
wer
li
ke
pr
e
vious
te
s
t.
I
n
t
he
c
om
bat
def
e
ns
e
co
ndit
ion
t
hat
only
use
1
To
wer
A
dik
a
r
a
p
resen
te
d
H
FSM
m
et
ho
d
i
s
m
or
e
superi
or
to
FSM
m
et
ho
d.
It
was
f
ou
nd
that
1
NP
C
To
wer
with
HFSM
a
ble
to
wit
hs
ta
nd
st
ron
g
tr
oop
at
ta
cks
by
8
t
roo
ps
,
m
eant
12.
5%
m
or
e
superi
or
tha
n
t
he
FSM
To
wer
.
F
or
the
m
od
erate
tr
oops
,
1
N
PC
T
ower
is
able
to
withstan
d
12
tr
oo
p
at
ta
c
ks
by
8.33%
m
or
e
s
uperi
or
than
the
FSM
To
wer
.
Me
an
wh
il
e,
as
for
t
he
wea
k
t
roo
ps
,
1
N
PC
T
ow
e
r
with
H
FSM
a
ble
to
s
urvive
f
ro
m
13
tro
op
s
, 15.3
8%
m
or
e superi
or
com
par
e to t
he
FS
M T
ower.
Figure
4.
Sim
ulati
on
Result
s
of 2 NPC
G
a
ndewa
T
ow
e
r
a
gainst
3
Ty
pe of
Tro
op
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
Defense
b
e
havi
or
of re
al ti
me st
ra
te
gy g
am
e
s
:
c
omp
ar
iso
n b
et
we
en
HF
SM
and FSM
(
Ra
hmat
Fa
uz
i
)
641
Figure
5.
Sim
ulati
on
Result
s
of 2 NPC
Adik
ara T
ow
e
r
a
gai
ns
t
3
Ty
pe
of
Troo
ps
4.
CONCL
US
I
O
N
The
inc
reases
total
nu
m
ber
of
N
PC
Kam
and
aka
To
we
r
pr
ov
i
ded
bette
r
r
esult
of
com
bat
def
e
ns
es
com
par
e
to
H
FSM
m
e
tho
ds,
if
us
in
g
F
SM
m
et
ho
ds.
The
refor
e
,
with
2
(tw
o)
NP
C
T
ower
with
H
FS
M
can
protect
the
at
ta
ck
from
17
st
rong
tr
oops
with
le
ading
sc
ore
5.8
8
%
c
om
par
e
to
the
to
we
r
wi
th
F
SM.
Me
an
wh
il
e,
the
to
wer
with
HF
SM
ca
n e
nd
ur
e
t
he a
tt
ack
f
ro
m
22
m
ediu
m
tro
ops
with
l
eadin
g sc
or
e
4.
54
%
c
om
par
e
to
the
towe
r wit
h FS
M.
On the
oth
e
r ha
nd, t
he
to
w
er
with
HFSM
can
h
a
ndle
29
weak t
r
oops
wi
th lea
ding
sco
r
e 3.
44
%
com
par
e
to
the
to
wer
with
FSM.
F
ro
m
3
(th
ree)
s
im
ulati
on
s,
the
ave
r
age
sc
ore
of
H
FSM
is
4.6
2
%
m
or
e
su
pe
rio
r
c
om
par
e
t
o
F
SM.
It
can
be
c
oncl
uded
based
on
t
he
e
vid
e
nce
th
at
HFSM
m
eth
od
can
m
ake
N
PC
Com
bat
Def
e
nse
To
we
r
perform
bette
r
res
ult
in
18
ty
pe
of
s
cenari
os
c
om
par
e
to
FSM
m
e
thod.
I
n
t
hi
s
s
t
u
dy
t
h
e
us
e
of
H
F
S
M
m
e
t
ho
ds
im
pl
em
e
nt
e
d
i
n
t
he
N
P
C
C
om
ba
t
D
e
f
e
ns
e
B
ui
l
di
ng
(
T
ow
e
r
)
s
t
i
ll
ha
s
t
he
po
s
s
i
b
i
l
it
y
t
o
be
de
ve
l
op
e
d
f
ur
t
he
r
f
o
r
di
f
f
e
r
e
n
t
c
at
e
go
r
i
e
s
of
c
om
po
ne
nt
s
or
ot
he
r
va
r
i
a
nt
t
yp
e
of
ga
m
e
.
F
or
e
xa
m
pl
e
,
by
ad
di
n
g
t
im
e
pa
r
am
et
e
r
s
i
nt
o
R
T
S
ga
m
e
s
uc
h
a
s
i
n
t
he
D
oT
A
ga
m
e
a
nd
t
he
n
um
be
r
o
f
t
he
N
P
C
bu
l
l
e
t
s
.
O
n
t
h
e
ot
he
r
ha
nd
,
f
o
r
t
he
N
P
C
C
om
ba
t
D
e
f
e
ns
e
B
ui
l
di
ng
(
T
ow
e
r
)
a
l
s
o
c
a
n
be
i
m
pl
em
e
nt
e
d
di
f
f
e
r
e
nt
l
y
s
uc
h
a
s
t
he
us
e
o
f
F
uz
z
y
L
og
i
c
m
e
t
ho
d
or
t
he
op
t
im
il
i
z
at
i
on
w
i
th
G
e
ne
t
i
c
A
l
go
r
i
t
hm
(
G
A
)
,
N
S
G
A
-
I
I
m
et
h
od.
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.
1
3
, N
o.
2
,
Fe
bru
ary
201
9
:
634
–
642
642
REFERE
NCE
S
[1]
Mill
ingt
on
I.
Art
if
icial
Intelli
g
en
ce
for Games
.
S
an
Franc
isco: Morgan
Kaufm
an P
ubli
sher.
2006
.
[2]
Gira
ult
A
,
L
ee
B,
Lee
EA.
Hier
archic
a
l
Finit
e
Sta
te
Mac
hin
es
with
Multi
ple
Concur
ren
c
y
Mode
ls
.
I
E
EE
Tr
ansacti
ons
on
Computer
-
Aided
Design
of
In
te
grated
Circuits and
Syst
ems
,
18
(6),
742
-
760
,
19
99.
[3]
Cze
rwinski
R
,
K
ani
a
D.
S
y
nthe
si
s
Method
of
Hig
h
Speed
Fini
te
State
Mac
h
ine
s.
B
ull
etin
of
the
Pol
ish
Aca
dem
y
of
Scie
nc
es,
Te
chnical
S
ci
en
ce
s
,
58(
5),
635
-
644
,
201
0.
[4]
Kiel
ar
PM
,
H
an
del
O
,
B
ie
d
ermann
DH
,
Borr
m
a
nn
A.
Concur
ent
Hier
arc
hi
cal
Fi
nit
e
State
Mac
hi
nes
for
Modeli
n
g
Pedestri
an
Beh
a
viora
l
Te
nd
encie
s.
Tr
anspor
tat
ion
Re
search
Proc
edi
a
,
2,
576
-
584,
2014.
[5]
Lubi
s
M,
Kus
uma
sari
TF,
Hak
im
L.
The
Indone
si
a
Public
Inform
at
i
on
Disclosure
Act
(UU
-
KIP
):
Its
Chal
le
ng
e
s
an
d
Responses.
Int
.
J
.
of
E
le
c
tric
al
an
d
Computer
Eng
ine
ering
,
8(1)
,
9
4
-
103,
2018
.
[6]
Chow S
-
H,
Ho
Y
-
C,
Hw
ang
T.
Low
Pow
er
Rea
l
iz
a
ti
on
of
Fini
te
Stat
e
Ma
chi
nes
-
A Dec
om
positi
on
Approac
h.
ACM
Tr
ansacti
on
on
Design
Aut
omat
i
on
of El
e
ct
roni
c Syste
ms
,
1
(3),
3
15
-
340,
Jul
y
199
6.
[7]
DeDeo
S.
Confl
i
ct
and
Com
puta
t
ion
on
W
iki
p
edia:
A
Fin
it
e
-
St
at
e
Mac
h
ine
Ana
l
ysis
of
Editor
Interac
t
ions.
Fut
ur
e
Inte
rnet
,
8(x)
,
2
016.
[8]
Arif
YM
,
Hari
a
di
M,
Nugroho
SM
S.
Inte
gra
si
Hier
arc
h
y
Fin
ite
Stat
e
M
ac
hin
e
dan
Logi
k
a
Fuz
z
y
untuk
Desa
in
Strat
egi NP
C
Ga
m
e.
Matics
4(3)
,
Mare
t
2011.
[9]
Le
e
B,
Lee
EA
.
Inte
ra
ct
ion
of
Fi
nit
e
St
ate
Mac
h
i
nes
and
Con
cur
r
ency
Mode
ls.
Ci
rcuit
s,
S
yste
ms
a
nd
Computers
2,
1998.
[10]
Mlaki
c
D,
Nikol
ovski
S,
Alib
asic E.
Designing
A
utomati
c
Met
er
Rea
ding
S
y
s
te
m
using Ope
n
Source
Hardware
and
Software
.
Int. J. of
E
le
c
tric
al
and
Computer
Eng
i
nee
ring
,
7(6), 32
82
-
3291,
De
c. 2
017.
[11]
Mall
ouk
W
,
Clu
a
E
.
An
Obj
ec
t
-
Orien
te
d
Approac
h
for
Hier
arc
h
ic
a
l
Sta
te
M
ac
hi
ne.
Proceedi
ngs
of
the
S
BGame
s
Confe
renc
e
in
C
omputing
.
R
ec
i
f
e,
Br
azil. 2006.
[12]
Maj
id
NK
,
H
ariadi
M,
Nugroho
SM
S.
Variasi
Per
il
aku
Tar
ung
N
on
P
lay
er
Chara
ct
er
Prajurit
Be
r
basis
Distribus
i
Gaus
sian p
ada
Re
al
Time
S
trategy
.
Suraba
y
a
:
Ju
rusan
T
ekni
k
Elektro, Institut
Teknologi
Sepu
luh
Nopem
ber
.
201
0.
[13]
Arif
YM
,
Har
ia
d
i
M,
Nugroho
S
MS
.
Strate
g
i
Me
nye
rang
Pada
G
ame
FP
S
M
engg
unakan
Hierarc
hic
al
Fi
n
it
e
Sta
t
e
Mac
hine dan
Lo
gik
a
Fuzzy
.
Sura
ba
y
a: Jurus
an
T
e
knik
E
le
ktro
,
Ins
ti
tut T
eknol
ogi
Sepuluh
Nopem
b
er.
2010
.
[14]
Tre
m
blay
J
,
Dra
ger
t
C
,
Verbrugg
e
C.
T
arg
et Sel
e
ct
ion
for
AI Co
m
pani
ons i
n
FP
S Gam
es.
Int.
C
on.
on
FDG
,
201
4.
[15]
Tre
m
blay
J
,
To
rr
es
PA
,
Verbrugg
e
C.
Mea
suring
Risk i
n
St
ea
l
th G
ames.
Int. Con. on
FDG
,
2014.
[16]
Dohert
y
D,
O’Ri
orda
n
C.
The
D
e
sign
Goals
and
Imple
mentation
o
f
AI
in
Mode
rn
Computer
Gam
es
.
Galwa
y
:
D
.
of
IT,
N
at
ion
al
Uni
ver
sit
y
of
Ire
la
n
d.
2006
.
[17]
W
ahono
RM.
P
e
ngantar
Soft
war
e
Ag
ent
:
Teori
d
an
Aplikasi
.
IEC
I
Chapter
Jap
an S
eri
es,
3(1
), 200
1.
[18]
Adisus
il
o
AK
.
Optimizati
on
Ag
en
t
Be
ha
vi
our
on
B
oxi
ng
Gam
e
Us
ing
Gen
et
i
c
Al
go
rithm
.
Sur
abay
a
:
Jurus
an
Teknik
El
ektro,
Instit
u
t Te
knologi Sepul
uh
Nopem
ber
.
2
013.
[19]
W
id
y
astuti
I.
Pem
il
ihan
Pe
ri
laku
NPC
Pada
G
ame
P
ertarungan
Jarak
De
kat
Me
nggunakan
F
uzz
y
Coordinato
r
.
Suraba
y
a: Jurus
an
Te
kn
ik El
ek
tro
,
Insti
tut
Te
kno
l
ogi
Sepuluh
Nopem
ber
.
2012
.
[20]
Benve
nist
e
A,
B
err
y
G.
The
S
y
n
chr
onous
Approac
h
to
React
ive
and
Real
-
Ti
m
e
Sy
stems
.
Proc
ee
d
ings
of
the
IE
EE
,
79
(9), Septe
m
b
er
1991.
[21]
Alur
R,
K
anna
n
S,
Yann
aka
kis
M
Com
m
unic
at
ing
Hier
arc
hi
cal
St
at
e
Ma
chi
n
e
s.
Int
.
Col
loquim
of
Au
tomata,
Languages and Progr
amm
ing,
Springer
pp
.
169
-
178,
1999
.
[22]
Risle
r
M.
Be
ha
vi
or
Control
fo
r
Singl
e
and
M
ult
iple
Au
tonomous
Age
nts
Bas
ed
on
Hi
erarchic
al
Fi
ni
te
State
Mac
hine
s
.
Doct
ora
l
Diss
ertati
on
,
T
ec
hn
ische
n
Univer
sit
at
Darm
stadt,
2009
.
[23]
Kurt
A,
Ozgune
r
U.
Hier
arc
h
ica
l
Fini
te
Sta
te
M
ac
hin
es
for
Aut
onom
ous
Mobile
S
y
s
te
m
s.
Cont
rol
Eng
ine
ering
Pract
i
ce
,
2
1
(
2),
pp.
184
-
194
.
Fe
brua
r
y
2013
.
[24]
Hakim
L,
Kus
um
asa
ri
TF,
Lub
is
M.
Te
x
t
Mining
of
UU
-
ITE
Im
ple
m
ent
ation
in
In
donesia
.
J.
of
Ph
ysic
s:
Confe
re
n
c
e
Serie
s
1007
(1).
[25]
Pinhei
ro
AC,
Adenil
so
S,
An
a
Mari
a
A.
FS
M
-
Based
Te
st
Case
Gene
r
atio
n
Methods
Ap
pli
ed
to
Te
st
t
he
Com
m
unic
at
ion
Software
on
B
oar
d
the
ITASA
T
Univ
ersity
S
a
te
llite:
A
Case
Stud
y
.
Sao
Jos
e
dos
C
ampos
.
J.
Ae
ros
psace
Tech
nology
Manag
e
ment
,
6(4
), 447
-
461,
Oct
-
D
ec
20
14.
[26]
Chow
TS.
Te
st
ing
Softwar
e
Design
Modelle
d
b
y
Fin
it
e
-
St
a
te
Ma
chi
n
es.
I
EE
E
Tr
ansacti
o
ns
on
Sof
tware
Engi
ne
ering,
4(3
),
178
-
187
.
1978
[27]
Fujiwara
S,
v
an
Bochmann
G,
K
hende
k
F,
Am
alou
M,
Gheda
m
si
A.
T
est
Se
le
c
tion
base
d
on
Fini
te
State
Mode
ls
.
IEE
E
Tr
ansac
ti
o
ns on
Soft
ware
E
ngine
ering
,
17(6
),
591
-
603
.
1991
.
[28]
Santi
ago
V,
Vi
j
a
y
kum
ar
NL,
Guim
ara
es
D,
Am
ara
l
AS
,
Ferr
e
ira
E.
An
Env
ir
onm
ent
for
Aut
om
at
ed
T
est
C
a
se
Gene
ration
fro
m
Stat
e
cha
r
t
a
nd
Fini
te
Sta
te
Mac
h
ine
-
b
ase
d
Beh
avi
or
al
M
odel
s.
Proce
edings
of
the
I
E
EE
Inte
rna
ti
ona
l
Co
n.
on
Softwar
e
Te
sting
Veri
f
ica
ti
on
and
Val
idation
W
orkshop.
W
ashingt
on
.
D
C,
US
A.
IEEE
Computer
Societ
y
,
pp
.
63
-
72
,
200
8.
[29]
Sala
u
y
ou
V
and
Grze
s
T.
FS
M
stat
e
assignm
ent
m
et
hods
for
lo
w
-
power
design.
Int.
Con
.
Comp
ute
r
Informatio
n
Syste
ms
and
Ind
ustrial
Manage
ment
App
li
at
ions
,
pp
.
345
-
350
,
2
007.
[30]
La
l
a
PK
.
An
Alg
orit
hm
for
th
e
St
at
e
As
signm
ent
of
S
y
nchr
onous
Sequent
i
al
Cir
cu
it
s.
Elec
tronic
s
Lett
ers
14(6)
,
pp
.
199
-
201,
1978
.
[31]
Ahlan
AR,
Lubi
s
M,
Lub
is
AR
.
In
form
at
ion
Secur
i
t
y
Aw
are
ness
a
t
t
he
Know
le
dge
-
b
ase
d
Instit
u
ti
on:
Its
Antecede
nts
and
Mea
sur
es.
P
roce
dia
Comput
er
Scienc
e
72
,
p
p.
361
-
373
,
201
5.
Evaluation Warning : The document was created with Spire.PDF for Python.