Indonesi
an
Journa
l
of El
ect
ri
cal Engineer
ing
an
d
Comp
ut
er
Scie
nce
Vo
l.
9,
No.
2,
Februa
ry 2
018,
pp.
335
~
341
IS
S
N: 25
02
-
4752, DO
I: 10
.11
591/ijeecs
.v9.i
2.pp
335
-
341
335
Journ
al h
om
e
page
:
http:
//
ia
es
core.c
om/j
ourn
als/i
ndex.
ph
p/ij
eecs
Fuzzy
Cogniti
ve M
aps
B
ased
G
am
e
B
alanc
ing
S
yste
m
in
R
eal
T
ime
Pat
el
Ka
lp
ana
D
h
anji
1
,
S
ant
ho
sh
K
u
mar
S
ingh
2
1
Resea
r
ch
S
cho
l
ar,
Inform
at
ion
Te
chn
o
log
y
,
AM
ET
Univer
si
t
y,
Chenn
ai
2
Depa
rtment
of
C
om
pute
r
S
ci
en
ce
,
T
agor
e
Coll
e
ge
of
s
ci
en
ce a
n
d
comm
erc
e, Mu
m
bai
Art
ic
le
In
f
o
ABSTR
A
CT
Art
ic
le
history:
Re
cei
ved
Oct
21
, 201
7
Re
vised
Dec
2
9
, 2
01
7
Accepte
d
Ja
n
20
, 2
01
8
Pla
y
ers
m
a
y
sto
p
play
ing
a
picked
amus
ement
sooner
tha
n
ant
ic
ip
ated
for
som
e
rea
sons
.
A
standout
amongs
t
the
m
ost
vital
is
id
ent
if
ie
d
wit
h
the
w
a
y
amus
ement
planners
and
desi
gner
s
adj
ust
d
ive
rsion
ch
alle
nge
le
v
el
s
.
Prac
ticall
y
spea
king,
pl
a
y
ers
h
a
ve
disti
n
ct
iv
e
ab
il
ity
le
v
el
s
and
m
a
y
discov
er
comm
on
fore
orda
ine
d
troubleso
m
e
le
ve
ls
as
too
sim
ple
or
too
h
ard
,
g
et
t
ing
to
be
noti
c
ea
b
l
y
disappo
inted
or
exha
usted.
The
outc
om
e
m
ight
be
diminished
insp
ira
ti
on
to
conti
nue
pl
a
y
ing
the
dive
rsion
,
whi
ch
impli
es
dec
re
ase
d
enga
g
ement.
A
wa
y
t
o
deal
with
a
ll
e
via
t
e
th
is
issue
is
d
y
n
ami
c
a
m
usem
ent
troubl
e
ad
justi
ng,
which
is
a
proc
ed
ure
that
a
lt
ers
di
ver
sion
p
l
a
y
par
amete
rs
prog
ressively
as
indicated
b
y
th
e
pre
sent
play
er
ap
ti
t
ude
le
ve
l.
In
thi
s
pape
r
we
p
ropose
a
const
a
nt
answer
for
DG
B
uti
li
z
ing
Evol
uti
on
a
r
y
Fuzz
y
Cogni
ti
v
e
Maps,
for
prog
ressively
adjus
ti
ng
a
dive
rsi
on
troubl
e
,
givi
ng
a
ver
y
m
uch
adj
usted
le
v
el
of
te
st
to
the
play
er
.
Tra
nsform
at
iv
e
Fuz
z
y
Cognit
ive
Maps
depe
nd
on
idea
s
tha
t
spe
ak
to
s
et
ti
ng
dive
rsion
fac
tors
and
are
conn
ecte
d
b
y
f
luff
y
and
pr
obabi
li
st
ic
ca
us
al
conn
ections
t
hat
c
an
b
e
ref
reshe
d
progr
essively
.
W
e
t
al
k
abou
t
a
fe
w
re
-
ena
c
tment
tri
es
th
a
t
uti
lization
our
a
nsw
er
in
a
runne
r
sort
amus
ement
to
m
ake
al
l
th
e
m
or
e
ca
pt
iva
t
ing and d
y
nami
c
div
ersi
on
enc
oun
te
rs.
Ke
yw
or
d
s
:
Re
al
-
tim
e Strat
egy
Gam
e D
ifficult
y B
al
ancing
Fu
z
zy
C
ogniti
ve
Ma
ps
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
:
Pate
l Kalpa
na Dh
a
nji,
Re
search
S
c
hola
r,
In
form
at
io
n
Tec
hnol
og
y,
AMET
Un
i
versi
ty
,
Chen
nai
.
1.
INTROD
U
CTION
Dive
rsion
play
in
c
om
pu
te
riz
ed
recreati
ons
include
s
a
fe
w
com
ponen
ts,
f
or
exam
ple,
ac
ti
viti
es
and
diff
ic
ulti
es
that
play
ers
m
us
t
e
m
br
ace
to
finish
am
us
e
m
ent
e
xer
ci
ses
.
An
a
m
us
e
m
ent
plann
er
m
ay
chan
ge
the
div
e
rsion
m
ec
han
ic
s
to
m
ake
chall
eng
es
si
m
pler
or
ha
r
de
r
to
com
pr
ehe
nd,
giv
i
ng
pr
e
def
i
ned
tr
oubl
e
le
vels,
for
e
xam
ple,
"sim
ple",
"ordin
ary",
a
nd
"ha
rd".
N
onet
heless,
these
m
od
ific
at
ion
s
a
re
sta
ti
c
an
d
m
igh
t
be
m
ade
in li
gh
t
of a
dis
creti
on
a
ry be
nc
hm
ark
, whic
h i
s not a
pprop
riat
e for
al
l cl
ie
nt
s.
Pr
act
ic
al
ly
sp
eakin
g,
play
er
s
ha
ve
div
er
s
e
abili
ty
and
ex
per
ie
nc
e
le
vels
an
d
m
ay
disco
ver
foreor
daine
d
tro
uble
so
m
e
lev
el
s
as
"t
oo
s
i
m
ple"
or
"t
oo
hard",
getti
ng
to
be
noti
ceably
disap
po
i
nted
or
exh
a
us
te
d.
T
he
ou
tc
om
e
m
igh
t
be
dim
inished
ins
pirati
on
to
co
ntin
ue
pla
yi
n
g
the
am
us
e
m
ent,
wh
ic
h
im
pl
ie
s
le
ssened en
ga
gem
ent.
An
a
nswer
f
or
ada
pt
to
the
se
issues
is
to
progressi
vely
change
the
di
ver
sio
n
tr
oubl
e
le
vels
as
ind
ic
at
ed
by
the
pr
e
sent
pla
yi
ng
set
ti
ng
,
wh
ic
h
inco
r
porates
ob
se
r
ving
play
er
act
iv
it
ie
s,
m
ist
akes,
and
exec
utio
n
i
n
t
he
am
us
em
ent
.
The
w
riti
ng
al
lud
es
t
o
ar
ra
ng
em
ents
in
li
gh
t
of
this
th
ought
as
"
dynam
ic
div
e
rsion
tro
uble
ad
j
us
ti
ng
a
nd
"
dynam
ic
t
rou
ble
al
te
ratio
n
(DD
A)
".
T
her
e
a
re
a
fe
w
work
s
t
hat
ap
proac
h
DG
B
a
nd
relat
ed
issues
.
For
instance
Tij
s
and
c
o
-
c
reato
rs
[1
]
pro
pose
d
to
ad
just
tro
ub
l
e
le
vels
util
iz
i
ng
t
he
play
er'
s
passionat
e
sta
te
.
Be
that
as
it
m
a
y,
the
wo
r
k
by
Tijs
an
d
co
-
c
r
eat
or
s
[
1]
sh
ows
a
few
do
w
ns
ides
.
In
it
ia
ll
y,
their
appr
oach
nee
ds
to
get
s
om
e
inf
or
m
at
ion
ab
ou
t
his/her
pas
sion
at
e
sta
te
a
m
id
the
am
us
e
m
ent.
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.
9
,
No.
2
,
Fe
bruary
2
01
8
:
335
–
341
336
Also
,
t
heir
ap
proac
h
does
no
t
hav
e
a
le
giti
m
a
te
ly
us
efu
l
basi
c
le
ader
s
hip
f
r
a
m
ewo
r
k.
I
n
a
no
t
her
r
el
at
ed
work,
Hunicke
[
2]
an
al
yz
ed
how
dy
nam
ic
trouble
al
te
rati
on
infl
ue
nced
play
er
adv
a
nce
w
hile
le
adin
g
analy
se
s
tha
t
con
t
ro
ll
ed
fr
ee
m
ark
et
act
iv
it
y
of
di
ff
e
ren
t
thi
ng
s
i
n
the
am
us
em
ent.
Vasconcel
os
de
Me
de
iros
[
3]
pro
posed
a
sta
ti
c
le
vel
adjustin
g,
i
n
li
ght
of
t
he
i
nput
of
genuine
gam
ing
e
ncou
nters.
This
a
ppr
oach
is
fascinati
ng
i
n
li
ght
of
t
he
fact
that
the
tr
ouble
le
vel
is
dis
play
ed
util
iz
ing
ge
n
uin
e
i
nfor
m
at
i
on
(
rathe
r
tha
n
util
iz
ing
a
n
ir
regular
and
s
ubj
ect
ive
gauge
).
Be
tha
t
as
it
m
ay
,
this
ar
rangem
ent
is
not
dynam
ic
an
d
t
he
tr
oubl
e
le
vels
c
onti
nue
as
befor
e
am
id the who
le
am
us
e
m
ent.
Fig
ure
1.
Tim
e
O
ver Gam
e
In
this
pap
e
r,
we
pro
po
se
a
te
chn
iq
ue
t
o
c
hange
t
he
tr
ou
ble
le
vels
pow
erfull
y
and
co
ntinuo
us
ly
,
wh
ic
h
de
pends
on
play
er
as
s
ociat
ion
data,
set
ti
ng
factors,
an
d
E
voluti
on
ary
F
uzzy
Co
gnit
ive
Ma
ps.
P
la
ye
r
coope
rati
on
s
c
on
ta
in
s
esse
ntial
act
ivit
ie
s
in
an
am
us
em
ent,
f
or
exam
ple,
"
hoppin
g",
"eat
i
ng
"
,
a
nd
"r
unni
ng
"
,
char
act
e
rized
i
n
the
di
ver
si
on
co
nf
ig
urat
ion
orga
nize.
Sett
ing
facto
rs
a
re
identifie
d
with
div
e
rsion
sta
t
e
an
d
Sale
n
an
d
Zim
m
er
m
an
[
4
]
ch
aracte
rize
"am
us
em
ent
sta
te
"
as
the
pr
e
sent
sta
te
of
the
di
ver
si
o
n
at
a
ny
giv
e
n
m
inu
te
.
Con
si
der
f
or
insta
nc
e
a
So
ccer
Gam
e.
In
it
s
a
m
us
em
ent
s
tate
co
m
po
ne
nts
we
cou
ld
loc
at
e
the
accom
pan
yi
ng
set
ti
ng
facto
rs
:
the
hal
f
ti
m
e
bein
g
play
ed,
the
rest
of
t
he
t
i
m
e,
group
dat
a,
cu
rr
e
nt
sc
ore
an
d
current cli
m
at
e
cond
it
io
ns
.
Tra
nsf
or
m
at
ive
Fu
zzy
Co
gn
i
ti
ve
Ma
p
(E
-
F
CM
),
is
a
displ
ay
ing
instr
ume
nt,
pro
pose
d
by
[
5
],[
6
]
,
in
view
of
F
uzzy
Cognit
ive
Ma
ps
,
with
t
he
disti
nction
t
hat
in
E
-
FCM
eac
h
sta
te
is
de
velo
ping
in
li
gh
t
of
no
n
-
determ
inist
ic
o
utside
cau
sal
it
i
es
progressi
vely
.
Our
ap
pr
oac
h
m
akes
an
E
-
FCM
in
view
of
div
er
sio
n
se
tt
ing
factors,
wh
ic
h
is
la
te
r
cha
ng
e
d
to
i
ncor
porat
e
play
er
c
ommun
ic
at
io
ns
,
f
or
exam
ple,
hop,
eat
and
r
un;
wh
ic
h
dep
e
nd
of
the
a
m
us
em
ent
ou
tl
ine.
T
he
E
-
FC
M
refresh
es
al
l
set
ti
ng
facto
rs
pro
gr
es
sively
rely
ing
upon
pl
ay
er
com
m
un
ic
at
ion
s,
w
hich
c
ha
ng
e
s
the
am
us
e
m
ent
trouble
le
vels
w
hile
a
div
e
rsion
sess
ion
is
goin
g
on.
We
util
iz
e
E
-
FCM
s
on
the
gro
unds
that
they
ar
e
eff
ect
ive
ap
pa
ratuses
to
help
with
thi
nk
i
ng
an
d
basic
le
a
der
s
hip
form
s.
The
wr
it
ing
giv
e
s
cases
of
util
i
zi
ng
E
-
FCM
s
in
a
few
uniq
ue
zo
ne
s,
for
exam
ple,
po
li
ti
cal
e
m
er
gen
c
y
adm
inist
rati
on
and
poli
ti
cal
basic
le
ader
s
hip
[
7
]
an
d
intel
li
gen
t
narrati
ng
[
6
]
.
An
ou
tc
om
e
of
pe
rio
dized
sm
al
l
side
gam
es
with
an
d
wit
hout
m
ental
i
m
age
ry
on
play
in
g
abili
ty
a
m
on
g
intercol
le
giate
le
vel
so
cce
r
pl
ay
ers
exp
la
ine
d
i
n [
8
]
. Review
of
c
ogniti
ve
rad
io
Netw
ork
is al
s
o
s
hows
the
[
9
].
2.
EVOLUTI
ONAR
Y
F
UZ
Z
Y
COGNITI
VE
MAP
Disp
la
yi
ng
a
dy
nam
ic
fr
am
e
work
ca
n
be
ha
rd
i
n
a
com
pu
ta
ti
on
al
se
ns
e
.
Furthe
rm
or
e,
plan
ning
a
sci
entifi
c
m
odel
m
igh
t
be
tr
oubles
om
e,
exo
r
bitant
an
d
at
tim
es
even
unim
aginab
le
.
These
m
et
ho
dolo
gies
offer
t
he
upsid
e
of
eval
uated
res
ults
ye
t
en
dure
a
fe
w
disadv
a
ntage
s,
for
exam
ple,
the
pr
e
re
qu
isi
te
t
o
have
par
ti
cula
r
le
ar
ning
outsi
de
th
e
area
of
prem
ium
A
co
m
par
at
ive
stud
y
b
et
ween
visibil
it
y
-
base
d
r
oa
dm
a
p
pat
h
plan
ning
al
gor
it
h
m
s
[1
0
]
.
Fluffy
Co
gn
it
iv
e
Ma
ps
are
a
subj
ect
ive
opti
on
way
to
deal
with
dy
nam
ic
fr
am
ewo
r
ks
,
w
her
e
the
gross
cond
uct
of
a
f
r
a
m
ewo
r
k
ca
n
be
watch
ed
ra
pid
ly
an
d
with
out
the
a
dm
inist
r
at
ions
of
operati
ons i
n
quire a
bout m
ast
er.
I
n
the E
voluti
onary F
uz
zy
Cogniti
ve
Ma
ps
each st
at
e is adv
a
ncin
g i
n
li
gh
t
of
nondet
erm
i
nisti
c
ou
tsi
de
c
ausali
ti
es
con
ti
nuously
.
E
-
FC
M
is
dev
el
ope
d
with
tw
o
pri
nciple
pa
rts:
ideas
an
d
causal
co
nnect
ion
s
.
Co
ncep
t
(C),
wh
ic
h
r
epr
ese
nts
a
v
ariable
of
inte
rest
in
a
real
-
tim
e
s
yst
e
m
a
nd
is
expresse
d
as
a
tup
le
:
=
(
,
,
)
(1)
Wh
e
re,
S
de
note
s
the
sta
te
value
of
the
c
on
ce
pt.
T
is
the
e
vo
l
ving
ti
m
e
fo
r
t
he
c
oncept,
re
pr
ese
nting
a
m
ul
ti
ple o
f
a
f
i
xed tim
e sli
ce t
0
a
nd Ps
is
the
pro
bab
il
it
y of s
el
f
m
utati
on
.
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
Fuzz
y
Co
gn
it
iv
e Ma
ps
Base
d Game
Bal
an
ci
ng S
y
ste
m
i
n
R
eal Time
(Pate
l K
alpana
D
hanji)
337
Ca
us
al
relat
ionship
(R),
wh
ic
h
re
pr
ese
nts
th
e
stren
gth
a
nd
pro
bab
il
it
y
of
the
causal
e
ff
ec
t
fr
om
one
con
ce
pt to
a
no
t
her co
nce
pt.
It
is def
i
ned as a
tup
le
:
=
(
,
,
)
(2)
Wh
e
re
W
is
the
weig
ht
m
a
trix
of
th
e
ca
us
al
relat
io
nsh
ip,
∈
[0
,
1].
S
denotes
whet
her
th
e
cau
s
al
relat
ion
s
hip
is
ei
ther
posit
iv
e
(+)
or
ne
gati
ve
(
−
).
is
the
prob
a
bili
ty
that
the
causal
con
ce
pt
af
fects
the
resu
lt
co
nce
pt
C
.
Fu
zzy
ca
us
a
l
relat
ion
s
hip
s
for
a
syst
em
w
it
h
n
va
riables
can
be
re
pr
ese
nted
as
a
n
×
n
weig
ht
m
at
rix
W
:
=
(
11
12
21
22
⋮
⋮
⋯
1
⋯
2
⋯
⋮
1
⋮
⋮
1
2
⋯
⋯
⋮
⋯
)
(3)
Fo
r
a syste
m
w
it
h
n va
riables,
the m
utu
al
cau
sal
p
r
obabili
ty
can be
represe
nted
a
s a
n x n
m
at
rix
:
=
(
11
12
21
22
⋮
⋮
⋯
1
⋯
2
⋯
⋮
1
⋮
⋮
1
2
⋯
⋯
⋮
⋯
)
(4)
Diff
e
re
nt conc
epts m
igh
t ha
ve
d
if
fer
e
nt e
vo
l
ving ti
m
es. F
or a syst
em
w
it
h
n varia
bles, it c
an be
represe
nted
as a
vector T:
=
(
1
2
⋮
⋮
)
(5)
Be
sides
the
ca
us
al
ef
fects
f
rom
oth
ers
c
once
pts,
eac
h
co
nce
pt
will
al
so
al
te
rn
at
e
it
s
inter
nal
sta
te
rand
om
ly
in
real
tim
e.
Each
co
nce
pt
is
m
od
el
le
d
with
ver
y
sm
al
l
mu
ta
ti
on
pro
babi
li
t
y.
If
the
pr
ob
a
bili
ty
is
hig
h,
the
syst
e
m
w
ou
l
d becom
e v
ery
unsta
ble.
F
o
r
a
s
yst
e
m
w
it
h
n v
ariables i
t ca
n be
represe
nted as a
vector
:
=
(
1
2
⋮
⋮
)
(6)
The
c
on
ce
pts
i
n
the
syst
em
up
date
t
heir
sta
te
s
in
their
res
pe
ct
ive
ev
olv
in
g
ti
m
e.
The
sta
te
value
of
c
oncept
is u
pd
at
e
d
acc
ordin
g
t
o
the
foll
ow
i
ng
eq
uat
ion
s:
∆
+
=
(
1
∑
∆
+
2
∆
=
0
)
(7)
+
=
+
∆
+
(8)
Wh
e
re
f
is
the
act
ivati
on
f
unc
ti
on
to
regulat
e
the
sta
te
value
.
is
the
sta
te
va
lue
of
co
nce
pt
at
tim
e
t.
Δ
i
is
the
sta
te
va
lue
cha
nge
of
co
ncep
t
at
tim
e
t.
T
is
the
evo
l
ving
ti
m
e
of
co
nce
pt
to
update
it
s
val
ue.
Diff
e
re
nt
con
c
epts
m
a
y
hav
e
diff
e
ren
t
e
vo
l
ving
tim
es.
The
k
1
an
d
k
2
va
lues
are
tw
o
w
ei
gh
t
co
ns
ta
nts
.
The
su
m
m
a
ti
on
Δ
is
su
bjec
te
d
to
c
onditi
on
al
pro
bab
il
it
y
,
an
d
Δ
is
sub
j
ect
ed
t
o
sel
f
-
m
uta
ti
on
pro
bab
il
it
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.
9
,
No.
2
,
Fe
bruary
2
01
8
:
335
–
341
338
3.
E
X
PERI
MEN
TS A
ND R
E
S
ULTS
So
as
to
te
ntati
vely
appr
ov
e
our
m
od
el
,
we
buil
t
up
the
Tim
e
ov
e
r
div
er
sio
n.
Tim
e
Ov
er
i
s
a
run
ner
so
rt
di
ver
si
on
wh
e
re
a
young
fell
ow
esc
apes
f
ro
m
a
t
wiste
r
to
sp
a
r
e
him
se
lf.
Fig
ur
e
1
outl
ines
so
m
e
screen
shots
of
Ti
m
e
Ov
e
r
a
m
us
e
m
ent.
In
a
pr
e
pa
rato
ry
adap
ta
ti
on,
t
he
am
us
e
m
ent
had
j
us
t
t
wo
se
tt
ing
factors:
score
and
s
pee
d.
T
he
div
e
rsion
com
pu
te
s
the
sco
re
var
ia
ble
as
pe
r
the
qua
ntit
y
of
thi
ngs
that
a
play
er
gathe
rs.
T
he
s
peed
var
ia
ble
has
co
ns
ist
ent
incenti
ve
in
t
he
di
ver
si
on.
Af
te
r
wa
rd,
we
add
e
d
m
or
e
set
ti
ng
factors
to
e
nh
ance
am
us
e
m
ent
play
,
c
onsideri
ng
pe
rsp
ec
ti
ves,
f
or
e
xa
m
ple,
play
er
ti
redness,
t
otall
ing
si
x
factors:
1)
Sta
m
ina: rep
re
sents the
p
la
ye
r’
s
ene
rg
y,
whi
ch
inc
reases
as
the
play
er c
ollec
ts m
or
e it
e
m
s in
t
he gam
e.
2)
Sp
ee
d:
re
prese
nts
the
play
er’s
sp
ee
d,
wh
ic
h
relat
es
to
sta
m
ina.
S
peed
de
creases
over
tim
e
to
sim
ulate
the
play
er c
har
act
e
r’
s ti
re
dness.
3)
O
bs
ta
cl
e
ty
pe
:
there
are
thr
ee
ty
pes
of
obsta
cl
es:
easy
,
def
a
ult,
a
nd
ha
rd.
T
hese
ty
pe
s
re
pr
e
sent
how
diff
ic
ult t
he o
bst
acl
es are.
4)
Ob
sta
cl
e
pe
rio
d:
re
presents
t
he
per
i
od
(tim
e
interval)
that
the
ga
m
e
us
es
to
insert
ob
sta
cl
es
in
the
gam
e
scene.
5)
Item
t
ype:
there
are
tw
o
ty
pe
s
o
f
co
ll
ect
ible i
tem
s in
the ga
m
e:
w
at
er bottl
e an
d
see
ds
.
Both
it
em
s incr
ease
play
er s
ta
m
ina, but w
at
e
r bo
tt
le
s provide
m
or
e stam
in
a than
seeds.
6)
Item
per
iod
:
re
pr
ese
nts
the
pe
rio
d
(ti
m
e
interval)
that
the
ga
m
e
us
es
to
insert
colle
ct
ible
ite
m
s
in
the
ga
m
e
scene.
Ever
y
set
ti
ng
var
ia
ble
is
a
f
luff
y
est
eem
,
sta
nd
a
rd
iz
e
d
t
o
the
sco
pe
of
[0,1
]
.
The
m
ean
of
e
ve
ry
var
ia
ble este
em
r
elies o
n u
p
on p
a
rtic
ular di
ver
si
on p
la
ns
. For ef
fortl
essn
ess w
e c
ha
racteri
zed im
ped
im
ent so
r
t
as
m
app
ing
t
he
real
est
im
a
tio
n
of
obstruct
ion
s
ort
to
t
he
or
et
ic
al
"si
m
ple",
"
def
a
ult",
an
d
"
hard"
t
rou
ble
hindra
nce
le
ve
ls.
The
"si
m
ple"
tro
ub
le
le
vel
m
aps
to
the
sc
op
e
of
[
0,0.33]
,
the
"de
fa
ult"
tro
uble
le
vel
m
aps
to
[0.34,
0.6
6]
an
d
the
"ha
r
d"
le
vel
m
aps
to
r
un
[
0.6
6,1].
T
he
Item
so
rt
as
m
app
in
g
the
re
al
est
i
m
ation
of
thi
ng
so
rt
to
cal
c
ulate
d
"water"
a
nd
"seeds".
T
he
water
thi
ng
a
ppears
t
o
the
ra
ng
e [0,0.5].
T
he
seed
thin
g
a
ppears
t
o
the r
a
nge [
0.6,1]
. We relat
e e
ver
y set
ti
ng
va
riable to
the ac
com
pan
yi
ng
i
de
as:
C1: Stam
ina.
C2: S
peed.
C3: O
bs
ta
cl
e t
ype.
C4: O
bs
ta
cl
e pe
rio
d.
C5: Item
ty
pe.
C6: Item
p
erio
d.
Fig
ure
2.
The
E
-
FCM
Mo
del
f
or
Tim
e O
ver Gam
e
Figure
3
sho
w
s
the
la
st
Ti
m
e
Ov
e
r'
s
E
-
FC
M
m
od
el
,
,
i
∈
[1,6
]
s
peaks
to
eve
ry
set
ti
ng
va
riable,
m
ark
ed
bolt
s
sp
eak
to
ca
us
al
connecti
ons
be
tween
set
ti
ng
facto
rs.
A
posit
ive
sign,
im
plies
posit
ive
causal
relat
ion
s
hip
a
nd
neg
at
ive
sig
n im
plie
s n
egati
ve
relat
io
nship
. Tab
le
1
outl
in
es the
pro
bab
il
ist
ic
w
ei
gh
t
net
work
W
of
ca
us
al
connecti
ons,
w
hich
are
res
olve
d
ei
ther
from
a
sp
eci
al
ist
inform
ation
or
le
arn
t
from
a
l
earni
ng
base; as t
he
m
od
el
inten
de
d for this am
us
em
e
nt is b
a
sic
, th
e
weig
hts w
e
re
gi
ven
by the
d
i
ve
rsion
plan
ner.
The
fr
am
ewo
r
k
is a o
nes gr
i
d
si
nc
e w
e c
onside
r
t
he
li
kelih
ood t
hat an i
dea
influ
enci
ng anot
he
r
idea
is on
e
.
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
Fuzz
y
Co
gn
it
iv
e Ma
ps
Base
d Game
Bal
an
ci
ng S
y
ste
m
i
n
R
eal Time
(Pate
l K
alpana
D
hanji)
339
Table
1
. Pr
ob
a
bili
sti
c
W
ei
gh
t
Mat
rix
W of
the
Ca
s
ual Rel
at
ion
s
hip
W
ij
C
1
C
2
C
3
C
4
C
5
C
6
C
1
0
1
0
0
-
1
0
.1
C
2
-
0
.1
0
1
-
1
0
0
.06
C
3
0
0
0
0
0
0
C
4
0
0
0
0
0
0
C
5
0
0
0
0
0
0
C
6
0
0
0
0
0
0
The
e
nactm
ent
w
ork
c
hosen
f
or
t
he
e
xam
inati
on
s
was
the
strat
e
gic
ca
pacit
y,
on
acc
ount
of
t
he
delic
at
e
lim
it
.
This
im
plies
the
co
ns
e
qu
e
nc
e
of
st
rategic
relapse
ca
n
be
decip
her
e
d
as
the
li
kelihood
of
watchin
g
certa
in
reacti
on
a
nd
li
kelihoo
d
ought
to
be
a
num
ber
in
the
vicinit
y
of
0
a
nd
1,
c
om
pr
eh
ensive
.
Keep
i
ng
in
m
i
nd
the
e
nd
goa
l
to
m
od
el
play
er
associat
io
ns
with
t
he
E
-
F
CM
,
we
a
dd
e
d
two
bolt
s
to
t
he
E
-
FCM
sho
w.
Th
e
1
bo
lt
,
s
peak
s
to
the
sta
m
ina
that
the
play
er
earn
e
d
by
gath
erin
g
thi
ngs.
T
he
2
bolt
s
pea
ks
to
sta
m
ina
m
is
fortu
ne.
T
he
st
a
m
ina
est
ee
m
dim
inishes
al
w
ay
s.
W
e
util
iz
e
the
"dive
rsion
ou
tl
ine"
as
t
he
tim
e
un
it
.
I
n
su
c
h
m
ann
er
,
we
con
si
der
that
ti
m
e
dev
el
ops
as
the
div
ersi
on
outl
ine
su
cc
ession
a
dv
a
nc
es.
W
e
refresh
the
six
set
ti
ng
f
act
ors
each e
dg
e
,
as
per the
d
e
velo
pi
ng tim
e T
.
Fig
ure
3.
E
-
FCM
Sim
ulatio
n 1 in Ti
m
eOv
er
Gam
e
Fig
ure
4. E
-
FCM
Sim
ulatio
n 2 in Ti
m
e
Ov
er
Gam
e
T = (1
1 1 1
1 1)
The
qual
it
ie
s
in
T
sig
nify
t
he
ti
m
e
interi
m
in
wh
ic
h
a
va
riable
is
r
efr
es
he
d.
For
instance,
a
n
est
i
m
ation
of
1
im
plies
that
a
var
ia
ble
is
re
fr
es
he
d
each
e
dg
e
.
A
n
est
im
at
ion
of
2
im
plies
that
a
var
i
able
is
refreshe
d
eac
h
two
e
dges,
et
c
et
era.
F
or
Tim
e
O
ver
di
ver
si
on,
because
of
it
s
strai
gh
tf
orw
ardness
,
w
e
dole
out
the
est
im
a
ti
on
of
one
t
o
al
l
s
et
ti
ng
fact
or
s
in
T
.
I
n
dif
fer
e
nt
set
ti
ng
s
,
when
it
is
re
quire
d
that
dive
rse
set
ti
ng
factors
are
re
freshe
d
non
c
oncurrently
,
eve
r
y
set
ti
ng
var
ia
ble
m
us
t
hav
e
it
s
par
ti
cular
adv
a
ncin
g
ti
m
e.
For
instance,
t
o
de
m
on
strat
e
the
consi
der
of
rai
n
an
e
nvir
on
m
ent,
the
a
dv
a
nc
ing
ti
m
e
of
the
rain
c
ou
l
d
be
10
on
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.
9
,
No.
2
,
Fe
bruary
2
01
8
:
335
–
341
340
the
off
cha
nc
e
that
we
nee
d
to
in
dicat
e
that
the
rain
com
po
ne
nt
is
refreshe
d
eac
h
10
to
outl
ine.
The
unde
rly
ing
esti
m
at
ion
s of the
six sett
ing fact
or
s
are:
S_0= (1
0.5 0
.
1 0.4
0.09)
Figures
3,
4,
5,
6
an
d
7,
re
pr
esent
the
after
eff
ect
s
of
ongoin
g
reprod
uct
ion
s
that
we
pl
ann
e
d
a
nd
directed
t
o
te
st
the
co
nduct
of
our
E
-
FCM
disp
la
y.
Eac
h
f
igure
outl
ines
the
set
ti
ng
factor
s
i
n
each
dive
rsi
on
ou
tl
ine.
Give
n
the
under
ly
in
g
set
up
S
_0,
we
ex
pected
that
in
the
re
pro
duct
ion
s,
the
set
ti
ng
fact
or
s
c
ha
ng
e
as
the
play
er
c
ollaborat
e
al
on
g
the
div
er
sio
n.
Each
fig
ur
e
ou
tl
ines
the
set
ti
ng
facto
rs
i
n
e
ach
div
er
sio
n
ou
tl
ine
,
exh
i
biti
ng
t
hat
a
ll
r
ecreat
io
ns
carried
on a
s
w
e antic
ipate
d.
Fig
ure
5.
E
-
FCM
Sim
ulatio
n 3 of
Tim
e o
ver Gam
e
Fig
ure
6.
E
-
FCM
Sim
ulatio
n 4 of
Tim
e o
ver Gam
e
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
Fuzz
y
Co
gn
it
iv
e Ma
ps
Base
d Game
Bal
an
ci
ng S
y
ste
m
i
n
R
eal Time
(Pate
l K
alpana
D
hanji)
341
Fig
ure
7.
Sim
ulati
on
5 of
Tim
e over
G
am
e
4.
RESU
LT
A
N
D DIS
CUSSI
ON
The
play
er
act
ion
s
of
eat
in
g
m
or
e
or
fe
wer
it
e
m
s
are
ref
le
ct
ed
in
the
i
nc
rease
an
d
dec
rease
of
th
e
stam
ina
value.
The
it
em
s
per
io
d
is
pro
port
ion
al
to
th
e
stam
ina,
bu
t
it
s
curve
is
s
of
te
r
since
the
re
i
s
le
ss
stam
ina
and
th
e
it
e
m
s
per
iod
is
sh
ort
er,
e
nsu
rin
g
that
the
pl
ay
er
will
hav
e it
e
m
s
to
e
at
,
in
order
t
o
increa
se
his
stam
ina
value
and,
there
fore,
increase
his
s
peed
value.
T
he
it
e
m
ty
pe
is
i
nv
e
rsely
pro
portion
al
relat
ed
to
the
stam
ina
value
because
of
the
i
m
pact
of
it
e
m
s
wh
e
n
t
he
val
ue
of
sta
m
ina
i
s
low:
it
m
us
t
be
hi
gh
e
r
s
o
th
at
the
stam
ina
value
can
be
increa
s
ed.
Du
e
t
o
thes
e
changes
,
w
hi
ch
af
fect
dire
ct
ly
to
the
act
ions
of
eat
i
ng
or
not
the
it
e
m
s,
the conte
xt v
a
riables
tend to
present
pe
ak.
5.
CONCL
US
I
O
N
We
watche
d
that
al
te
ring
the
E
-
FCM
de
li
ver
ed
the
c
ov
et
e
d
res
ult,
as
the
play
e
r
play
s
th
e
a
m
us
em
ent;
ou
r
te
ch
nique
c
ould
c
ha
ng
e
the
tro
ub
le
le
vels
powe
rfull
y
util
iz
ing
the
set
ti
ng
fa
ct
ors
a
nd
pl
ay
er
connecti
on as
data sou
rces. S
ub
s
eq
ue
ntly
, we infe
r
that t
he pr
opos
e
d
te
c
hniqu
e is
pro
fici
ent and is
ve
rsati
le
to
the p
la
ye
r
n
ee
ds p
rogressi
vely
,
en
ha
ncin
g
t
he
am
us
e
m
ent p
la
y i
nvolv
em
ent.
REFERE
NCE
S
[1]
T.
Ti
js
,
e
t
a
l.
,
“
D
y
namic
game bala
n
ci
ng
b
y
re
c
ogniz
ing
aff
ec
t
.
Fun a
nd
games
,
”
pp.
88
-
93
,
2008
.
[2]
R.
Hunicke,
“
The
ca
se
for
d
y
n
amic
diffi
cu
lty
adj
ustm
ent
in
g
ames
,”
i
n
Proceedi
ngs
of
the
20
05
ACM
SIGCH
I
Inte
rnational
Co
nfe
renc
e
on
Advance
s in
comput
er
entertainment t
e
chnol
ogy
,
pp
.
429
-
433
,
20
0
5
.
[3]
R.
J.
V.
de
Mede
iros
dan
T.
F.
V.
de
Mede
iro
s,
“
Proce
dura
l
Le
ve
l
Bal
an
ci
ng
in
Runner
Gam
es
,”
i
n
Compu
te
r
Gam
es
and
Digi
tal
Ent
erta
inmen
t
(
SBGAM
ES)
,
I
EE
E
,
2014
Brazi
li
an
S
ymposium on
,
pp
.
109
-
114
,
2014
.
[4]
K.
Sale
n
and
E
.
Zi
m
m
erman,
“
The
game
design
r
ea
der
:
A ru
le
s of
play
an
thol
og
y
,”
MIT
p
ress,
20
06.
[5]
Y.
C
ai,
e
t
al.
,
“
Conte
xt
m
odel
ing
with
evol
ut
iona
r
y
fuz
z
y
cognitiv
e
m
ap
in
int
eractive
stor
y
telli
ng
,”
i
n
IEEE
World
Congress
on
Computati
onal
Intelli
gen
ce
)
.
I
EE
E
I
nte
rnational
Co
nfe
renc
e
on
,
pp
.
2320
-
2325
,
200
8
.
[6]
Y.
Cai
,
et
a
l.
,
“
Crea
ti
ng
an
imm
ersive
game
world
with
evolutiona
r
y
fu
zzy
cogni
ti
v
e
m
aps
,”
IE
EE
computer
graphics
and
ap
pli
cations
,
vo
l/
is
sue:
30
(2),
pp
.
58
-
70
,
2010
.
[7]
A.
S.
Andreou
,
et
al
.
,
“
Evol
ut
io
nar
y
fuz
z
y
cogn
it
ive
m
aps:
A
h
y
brid
s
y
st
em
for
cri
sis
m
ana
g
ement
and
pol
it
i
cal
dec
ision
m
aki
ng
,
”
2003
.
[8]
A.
M.
Kum
ar,
“
An
outc
om
e
of
per
iodized
sm
al
l
side
games
wit
h
and
without
m
ent
al
image
r
y
on
play
ing
abi
l
i
t
y
among i
nte
r
col
l
e
gia
t
e
l
evel
socc
e
r
play
e
rs
,”
India
n
Journal
o
f
S
ci
e
nce
and
Techno
l
ogy
,
vo
l/
issue:
8
(36)
,
2015
.
[9]
B.
Th
angala
ksh
m
i
and
G.
T.
Bh
ara
th
y
,
“
Review
of
cognitive
rad
i
o
Network,
”
I
JM
SR
,
2015
.
[10]
L.
Lu
lu
and
A.
El
nag
ar,
“
A
co
m
par
at
ive
stud
y
bet
wee
n
v
isibi
l
ity
-
bas
ed
roa
dm
a
p
pat
h
pl
anni
ng
al
gorit
hm
s,”
In
tell.
Robot
.
Syst
.
,
200
5.
Evaluation Warning : The document was created with Spire.PDF for Python.