I
n
d
on
e
s
ian
Jou
r
n
al
of
E
lec
t
r
ical
E
n
gin
e
e
r
in
g
a
n
d
Com
p
u
t
e
r
S
c
ience
Vol.
25
,
No.
3
,
M
a
r
c
h
2022
,
pp.
1344
~
1355
I
S
S
N:
2502
-
4752,
DO
I
:
10
.
11591/i
jee
c
s
.
v
25
.i
3
.
pp
1344
-
1355
1344
Jou
r
n
al
h
omepage
:
ht
tp:
//
ij
e
e
c
s
.
iaes
c
or
e
.
c
om
N
e
w
l
am
b
d
a
t
u
n
i
n
g ap
p
r
oa
c
h
of
si
n
gl
e
i
n
p
u
t
f
u
z
z
y l
ogi
c
u
si
n
g
gr
ad
ie
n
t
d
e
sc
e
n
t
al
gor
it
h
m
a
n
d
p
ar
t
i
c
le
swar
m
op
t
imiz
at
io
n
F
au
z
al
Naim
Z
oh
e
d
i,
M
oh
d
S
h
ah
r
iee
l
M
oh
d
Ar
as
,
Hyre
il
Anu
ar
Kas
d
irin
,
Nurdi
an
a
B
in
t
i
Nor
d
in
U
nde
r
w
a
te
r
T
e
c
hnol
ogy R
e
s
e
a
r
c
h G
r
oup (
U
T
e
R
G
)
, C
e
nt
e
r
f
or
R
obot
ic
s
a
nd I
ndus
tr
ia
l
A
ut
oma
ti
on (
C
E
R
I
A
)
,
F
a
kul
ti
K
e
ju
r
ut
e
r
a
a
n E
le
kt
r
ik
, U
ni
ve
r
s
it
i
T
e
kni
ka
l
M
a
la
ys
ia
M
e
la
ka
, M
e
la
ka
,
M
a
la
y
s
ia
Ar
t
icle
I
n
f
o
AB
S
T
RA
CT
A
r
ti
c
le
h
is
tor
y
:
R
e
c
e
ived
Oc
t
29
,
2021
R
e
vis
e
d
J
a
n
11
,
202
2
Ac
c
e
pted
J
a
n
17
,
202
2
Un
d
erw
a
t
er
remo
t
e
l
y
o
p
era
t
ed
v
e
h
i
c
l
e
(
RO
V
)
i
s
i
mp
o
rt
a
n
t
i
n
u
n
d
er
w
at
er
i
n
d
u
s
t
ri
es
as
w
e
l
l
as
fo
r
s
afet
y
p
u
r
p
o
s
es
.
It
can
d
i
v
e
d
ee
p
er
t
h
a
n
h
u
ma
n
s
an
d
can
rep
l
ace
h
u
man
s
i
n
a
h
azar
d
o
u
s
u
n
d
er
w
at
er
en
v
i
r
o
n
me
n
t
.
RO
V
d
e
p
t
h
co
n
t
ro
l
i
s
d
i
ffi
c
u
l
t
d
u
e
t
o
t
h
e
h
y
d
ro
d
y
n
am
i
c
o
f
t
h
e
RO
V
i
t
s
el
f
a
n
d
t
h
e
u
n
d
erw
a
t
er
en
v
i
r
o
n
me
n
t
.
O
v
ers
h
o
o
t
i
n
t
h
e
d
ep
t
h
c
o
n
t
ro
l
may
cau
s
e
d
ama
g
e
t
o
t
h
e
RO
V
an
d
i
t
s
i
n
v
e
s
t
i
g
a
t
ed
l
o
cat
i
o
n
.
T
h
i
s
p
ap
er
p
re
s
en
t
i
n
g
a
n
e
w
t
u
n
i
n
g
ap
p
r
o
ach
o
f
s
i
n
g
l
e
i
n
p
u
t
fu
zz
y
l
o
g
i
c
c
o
n
t
ro
l
l
er
(
SI
FL
C
)
w
i
t
h
g
ra
d
i
e
n
t
d
es
ce
n
t
a
l
g
o
ri
t
h
m
(
G
D
A
)
a
n
d
p
ar
t
i
c
l
e
s
w
arm
o
p
t
i
m
i
zat
i
o
n
(
PSO
)
i
mp
l
emen
t
at
i
o
n
fo
r
R
O
V
d
ep
t
h
c
o
n
t
ro
l
.
T
h
e
RO
V
w
as
m
o
d
e
l
ed
u
s
i
n
g
s
y
s
t
em
i
d
e
n
t
i
fi
ca
t
i
o
n
t
o
s
i
m
u
l
a
t
e
t
h
e
d
e
p
t
h
s
y
s
t
em.
Pro
p
o
rt
i
o
n
al
i
n
t
eg
ra
l
d
eri
v
at
i
v
e
(
PID
)
co
n
t
r
o
l
l
er
w
as
a
p
p
l
i
e
d
t
o
t
h
e
mo
d
e
l
a
s
a
b
as
i
c
co
n
t
r
o
l
l
er.
SIFL
C
w
as
t
h
en
i
mp
l
eme
n
t
e
d
i
n
t
h
ree
t
u
n
i
n
g
a
p
p
ro
ac
h
es
w
h
i
ch
are
h
eu
r
i
s
t
i
c,
G
D
A
,
an
d
PSO
.
T
h
e
o
u
t
p
u
t
t
ra
n
s
i
en
t
w
as
s
i
m
u
l
a
t
ed
u
s
i
n
g
M
at
l
a
b
/
S
i
mu
l
i
n
k
a
n
d
t
h
e
p
ercen
t
o
v
ers
h
o
o
t
(
O
S),
t
i
me
r
i
s
e
(
T
r),
an
d
s
et
t
l
i
n
g
t
i
me
(T
s
)
o
f
t
h
e
s
y
s
t
ems
w
i
t
h
o
u
t
an
d
w
i
t
h
co
n
t
r
o
l
l
ers
w
ere
co
mp
ared
an
d
an
al
y
zed
.
T
h
e
res
u
l
t
s
h
o
w
s
t
h
at
SIFL
C
G
D
A
o
u
t
p
u
t
h
as
t
h
e
b
e
s
t
t
ra
n
s
i
en
t
res
u
l
t
at
0
.
1
0
2
1
%
(O
S),
0
.
7
9
9
2
s
(T
r),
an
d
0
.
9
7
9
0
s
(T
s
).
K
e
y
w
o
r
d
s
:
Gr
a
dient
de
s
c
e
nt
a
lgor
it
hm
P
a
r
ti
c
le
s
wa
r
m
opti
mi
z
a
ti
on
P
r
opor
ti
ona
l
int
e
gr
a
l
de
r
ivative
c
ontr
oll
e
r
R
e
mot
e
ly
o
pe
r
a
ted
ve
hicle
S
ingl
e
input
f
uz
z
y
logi
c
c
ontr
oll
e
r
Th
i
s
i
s
a
n
o
p
en
a
c
ces
s
a
r
t
i
c
l
e
u
n
d
e
r
t
h
e
CC
B
Y
-
SA
l
i
ce
n
s
e.
C
or
r
e
s
pon
din
g
A
u
th
or
:
F
a
uz
a
l
Na
im
Z
ohe
di
U
nd
e
r
wa
te
r
T
e
c
h
no
lo
gy
R
e
s
e
a
r
c
h
G
r
ou
p
(
U
T
e
R
G
)
,
C
e
n
te
r
f
o
r
R
ob
ot
ics
a
nd
I
nd
us
t
r
ia
l
A
ut
om
a
t
io
n
(
C
E
R
I
A
)
,
F
a
kult
i
Ke
jur
uter
a
a
n
E
lektr
ik
,
Unive
r
s
it
i
T
e
knikal
M
a
lays
ia
M
e
l
a
ka
76100
Dur
ian
T
ungga
l,
M
e
laka
,
M
a
lays
ia
E
mail:
f
a
uz
a
l@ut
e
m.
e
du.
my
1.
I
NT
RODU
C
T
I
ON
I
n
the
unde
r
wa
ter
e
nginee
r
ing
f
ield
,
r
e
mot
e
ly
op
e
r
a
ted
ve
hicle
(
R
OV
)
p
lays
a
n
im
por
tant
r
ole
in
unde
r
wa
ter
obs
e
r
va
ti
on,
inves
ti
ga
ti
on,
a
nd
ins
pe
c
ti
on
[
1]
–
[
3
]
.
E
s
pe
c
ially
in
the
oi
l
a
nd
ga
s
indus
tr
y,
R
OV
is
us
e
d
to
do
unde
r
wa
ter
pipe
ins
pe
c
ti
ons
a
s
we
ll
a
s
r
e
pa
ir
ing
jobs
.
R
OV
nor
mally
s
uf
f
e
r
e
d
f
r
om
pr
obl
e
ms
that
include
pos
e
r
e
c
ove
r
y
o
r
s
tation
ke
e
ping
,
unde
r
a
c
tuate
d
c
ondit
ions
,
c
oupli
ng
is
s
ue
s
,
a
nd
c
omm
unica
ti
on
tec
hniques
[
4]
.
T
his
r
e
s
e
a
r
c
h
pa
pe
r
wa
s
f
oc
us
ing
on
R
OV
de
pth
c
ontr
ol
or
s
tation
ke
e
ping.
S
tation
ke
e
ping
a
t
a
c
e
r
tain
de
pth
is
ve
r
y
im
po
r
tant
f
or
unde
r
wa
ter
e
xplo
r
a
ti
on
a
nd
ins
pe
c
ti
on
[
5]
–
[
7]
.
How
e
ve
r
,
c
ontr
o
ll
ing
R
OV
is
di
f
f
icult
be
c
a
us
e
of
the
une
xpe
c
ted
a
nd
unpr
e
dicta
ble
un
de
r
wa
ter
e
nvir
onment
[
4
]
,
[
8
]
.
T
his
is
due
to
the
nonli
ne
a
r
hydr
odyna
mi
c
s
e
f
f
e
c
t,
c
oupled
c
ha
r
a
c
ter
s
of
plant
e
qua
ti
ons
,
lac
k
of
pr
e
c
is
e
models
of
unde
r
wa
ter
v
e
hicle
hydr
odyna
mi
c
s
a
nd
unc
e
r
tainty
pa
r
a
me
ter
s
[
9]
,
[
10]
,
a
s
we
ll
a
s
the
pr
e
s
e
nc
e
of
e
nvir
onmenta
l
dis
tur
ba
n
c
e
s
[
1]
,
[
11]
–
[
14
]
.
C
ontr
oll
e
r
de
s
ign,
ba
s
e
d
on
s
im
ple
models
o
f
un
de
r
wa
ter
ve
hicle
mas
s
a
nd
dr
a
g,
ge
ne
r
a
ll
y
yields
una
c
c
e
ptabl
e
pe
r
f
or
manc
e
s
[
15]
.
L
inea
r
(
c
onve
nti
ona
l)
c
ontr
oll
e
r
is
una
ble
to
a
de
qua
tely
c
ontr
ol
the
unmanne
d
unde
r
wa
ter
ve
hicle
(
UUV
)
s
a
ti
s
f
a
c
tor
il
y
[
16]
.
E
ve
n
f
or
a
one
-
a
xis
mot
ion
s
uc
h
a
s
ve
r
ti
c
a
l
mot
ion
or
he
a
ve
mot
ion
,
c
ons
is
tent
pe
r
f
o
r
manc
e
f
or
a
de
s
ir
a
ble
r
a
nge
is
r
e
quir
e
d
.
Ove
r
s
hoot
in
the
s
ys
tem
is
Evaluation Warning : The document was created with Spire.PDF for Python.
I
ndone
s
ian
J
E
lec
E
ng
&
C
omp
S
c
i
I
S
S
N:
2502
-
4752
N
e
w
lambda
tuni
ng
appr
oac
h
of
s
ingl
e
input
fuz
z
y
logi
c
us
ing
gr
a
dient
de
s
c
e
nt
…
(
F
auz
al
N
aim
Z
oh
e
di)
1345
una
c
c
e
ptable
a
s
it
c
a
n
ha
r
m
the
R
OV
or
i
ts
ins
pe
c
ti
on
loca
ti
on
[
14
]
,
[
17]
–
[
20
]
.
I
t
is
be
s
t
to
ha
ve
a
s
m
ini
mum
ove
r
s
hoot
a
s
pos
s
ibl
e
in
the
R
OV
s
ys
tem.
T
he
r
e
a
r
e
many
c
ontr
ol
ler
s
de
s
igned
by
the
r
e
s
e
a
r
c
he
r
to
c
a
ter
to
thi
s
pr
oblem.
T
he
r
e
a
r
e
pr
opor
ti
ona
l,
int
e
gr
a
l,
a
nd
de
r
ivative
(
P
I
D)
ba
s
e
d
c
ontr
oll
e
r
s
a
nd
a
r
ti
f
icia
l
ba
s
e
d
c
ontr
oll
e
r
s
.
P
I
D
is
a
s
im
ple
c
ontr
ol
tec
hnique
that
ha
s
be
e
n
univer
s
a
ll
y
us
e
d
be
c
a
us
e
of
the
s
im
pli
c
it
y
of
im
pleme
ntation
in
a
r
e
a
l
-
ti
me
s
ys
tem.
E
ve
n
f
or
wor
k
c
las
s
R
O
V,
P
I
D
is
us
e
d
a
s
it
s
c
ontr
oll
e
r
.
How
e
ve
r
,
the
li
mi
tation
is
that
i
t
c
a
nnot
dyna
mi
c
a
ll
y
c
ompens
a
te
f
or
un
modelled
ve
hi
c
le’
s
hydr
odyna
mi
c
s
f
or
c
e
s
or
unknown
dis
tur
ba
nc
e
s
.
F
ur
ther
mor
e
,
the
c
ontr
a
dictor
y
pa
r
a
mete
r
c
onf
igur
a
ti
on
s
uc
h
a
s
be
twe
e
n
the
r
is
e
ti
me
a
nd
ove
r
s
hoot
may
a
ls
o
e
xis
t.
P
I
D
c
ontr
oll
e
r
s
ha
ve
be
e
n
im
pleme
nted
in
pr
e
vious
wor
k
f
or
tr
a
c
king
pur
pos
e
s
in
UU
V
[
21]
–
[
25]
a
s
we
ll
a
s
in
R
OV
[
17]
,
[
26]
.
No
r
mall
y,
the
P
I
D
c
on
tr
oll
e
r
wa
s
us
e
d
a
s
a
ba
s
ic
c
ontr
ol
ler
to
be
c
ompar
e
d
with
a
nother
c
ompl
e
x
c
ontr
oll
e
r
s
uc
h
a
s
with
li
ne
a
r
q
ua
dr
a
ti
c
r
e
gulator
(
L
QR
)
[
6
]
,
F
u
zzy
-
ba
s
e
d
P
I
D
[
27]
,
[
28
]
.
T
he
P
I
D
wa
s
ha
r
d
to
be
tuned
to
c
ope
with
the
no
n
-
li
ne
a
r
na
tur
e
of
the
unde
r
wa
ter
e
nvi
r
onmen
t,
in
whic
h
it
typi
c
a
ll
y
pr
oduc
e
s
high
ove
r
s
hoot
a
nd
high
s
te
a
dy
-
s
tate
e
r
r
or
.
Due
to
the
li
mi
tation
of
P
I
D,
a
r
ti
f
icia
l
int
e
ll
igent
(
AI
)
ba
s
e
d
c
ontr
oll
e
r
s
s
uc
h
a
s
f
uz
z
y
lo
gic
c
ontr
oll
e
r
(
F
L
C
)
a
nd
a
r
ti
f
icia
l
ne
ur
a
l
ne
twor
k
(
AN
N)
ha
d
be
e
n
int
r
oduc
e
d
to
c
ontr
o
l
R
OV
.
AN
N
wa
s
us
e
d
to
pr
e
dict
th
e
pe
r
f
or
manc
e
of
the
R
OV
de
pth
s
ys
tem
ba
s
e
d
on
pr
e
vious
input
a
nd
mi
nim
ize
the
c
os
t
f
unc
ti
on
[
29]
.
T
he
n,
the
be
s
t
in
put
is
s
ugge
s
ted.
T
he
AN
N
r
e
s
ult
s
we
r
e
s
upe
r
ior
c
ompar
e
d
to
other
c
ontr
oll
e
r
s
that
we
r
e
e
xpe
r
im
e
nted
with.
AN
N
wa
s
a
ls
o
u
s
e
d
to
tune
the
P
I
D
a
nd
a
da
pt
with
the
de
pth
c
ha
ng
ing
of
R
OV
[
30]
.
I
t
wa
s
a
ls
o
us
e
d
f
o
r
the
s
a
me
pur
po
s
e
in
[
31
]
by
im
pleme
nti
ng
a
r
a
d
ial
ba
s
is
f
unc
ti
on
ne
ur
a
l
ne
twor
k
(
R
B
F
NN
)
f
or
tr
a
jec
tor
y
tr
a
c
king
f
o
r
t
he
a
utonom
ous
unde
r
wa
ter
ve
hicle
(
AU
V)
.
B
oth
s
howe
d
good
r
e
s
ult
s
.
How
e
ve
r
,
the
downs
ide
of
AN
N
wa
s
the
long
c
omput
a
ti
ona
l
ti
me
that
may
lea
d
to
a
lagging
pr
oblem
.
Anothe
r
AI
-
ba
s
e
d
c
ontr
oll
e
r
f
or
the
R
OV
s
ys
tem
is
the
F
L
C
whic
h
wa
s
a
ppli
e
d
in
R
OV
[
21]
,
[
23]
,
a
nd
AU
V
[
32
]
.
T
he
F
L
C
c
ontr
oll
e
r
c
a
n
c
ope
with
a
n
unknown
mathe
matica
l
modeling
s
ys
tem.
I
mpl
e
menta
ti
on
of
F
L
C
e
a
s
e
s
the
ne
e
d
f
or
pr
e
c
is
e
a
nd
c
ompl
e
x
hydr
odyna
mi
c
modeling
of
the
ve
hi
c
le.
F
or
e
xa
mpl
e
,
F
L
C
wa
s
s
uc
c
e
s
s
f
ull
y
us
e
d
to
tune
the
P
I
D
c
ontr
oll
e
r
f
or
unde
r
wa
ter
ve
hicle
[
33]
.
How
e
ve
r
,
e
ve
n
with
the
a
da
ptabili
ty
a
dva
ntage
,
F
L
C
ha
s
a
c
ha
ll
e
nging
leve
l
of
c
ompl
e
xit
y
.
He
nc
e
,
a
s
im
p
li
f
ied
s
ingl
e
input
f
uz
z
y
logi
c
c
ontr
oll
e
r
(
S
I
F
L
C
)
is
pr
opos
e
d
to
c
ontr
ol
the
de
pth
of
R
OV
.
P
r
e
vious
wor
ks
ha
ve
s
hown
that
S
I
F
L
C
ha
s
e
xc
e
ll
e
nt
p
e
r
f
or
manc
e
,
a
nd
it
e
xa
c
tl
y
r
e
s
e
mbl
e
d
c
onve
nti
o
na
l
F
L
C
tr
a
ns
ient
r
e
s
pons
e
[
34]
,
[
35]
.
S
I
F
L
C
r
e
duc
e
s
the
input
of
c
onve
nti
ona
l
F
L
C
int
o
s
ingl
e
input
s
ingl
e
output
(
S
I
S
O)
s
ys
tem.
Nor
mally,
a
tr
ial
-
a
nd
-
e
r
r
or
(
he
ur
is
ti
c
)
method
wa
s
us
e
d
to
f
ind
the
opti
mum
pa
r
a
m
e
ter
.
C
on
s
e
que
ntl
y,
it
take
s
mor
e
ti
me
e
xe
c
uti
on
to
f
ind
the
opti
mum
pa
r
a
mete
r
s
.
T
his
pa
pe
r
pr
e
s
e
nts
a
ne
w
tuni
ng
a
ppr
oa
c
h
of
S
I
F
L
C
with
gr
a
dient
de
s
c
e
nt
a
lgor
it
hm
(
GD
A)
a
nd
pa
r
ti
c
le
s
wa
r
m
opti
m
iza
ti
on
(
P
S
O)
im
pleme
ntati
on
f
or
R
OV
de
pth
c
ontr
ol
.
T
he
R
OV
w
a
s
model
e
d
us
ing
s
ys
tem
identif
ica
ti
on
to
s
im
ulate
the
de
pth
s
ys
tem.
P
I
D
c
ontr
oll
e
r
wa
s
a
ppli
e
d
to
the
model
a
s
a
ba
s
ic
c
ontr
oll
e
r
.
S
I
F
L
C
wa
s
then
im
pleme
nted
in
thr
e
e
dif
f
e
r
e
nt
tuni
ng
a
ppr
oa
c
he
s
whic
h
we
r
e
tr
ial
-
a
nd
-
e
r
r
or
(
he
ur
is
ti
c
)
,
GD
A,
a
nd
P
S
O
.
T
he
ou
tput
tr
a
n
s
ien
t
wa
s
s
im
ulate
d
us
ing
M
a
tl
a
b/
S
im
uli
nk
a
nd
the
pe
r
c
e
nt
ove
r
s
hoot
(
OS)
,
r
is
e
ti
me
(
T
r
)
,
a
nd
s
e
tt
li
ng
ti
m
e
(
T
s
)
of
the
s
ys
tems
without
a
nd
with
c
ont
r
oll
e
r
s
we
r
e
c
ompar
e
d
a
nd
a
na
lyze
d.
I
n
ter
ms
of
de
pth
c
ont
r
ol,
the
ove
r
s
hoot
(
%
OS)
is
a
n
im
po
r
tant
pa
r
a
mete
r
to
obs
e
r
ve
a
s
a
high
va
lue
may
da
mage
the
R
OV
or
it
s
inv
e
s
ti
ga
ti
on
plac
e
[
14
]
,
[
18]
–
[
20
]
,
[
36]
.
T
he
r
is
e
ti
me
(
T
r
)
s
hows
the
ti
me
take
n
to
ge
t
to
the
de
s
ir
e
d
point
w
hil
e
the
s
e
tt
li
ng
ti
me
is
the
ti
me
R
OV
s
tabili
z
e
s
a
t
a
s
te
a
dy
s
tate
.
2.
S
YST
E
M
M
ODE
L
L
I
NG
I
n
thi
s
pa
pe
r
,
the
R
OV
wa
s
modele
d
us
ing
the
s
ys
tem
identif
ica
ti
on
(
S
I
)
method
in
a
M
a
tl
a
b
c
omput
ing
e
nvir
onment
.
F
or
s
ys
tem
identif
ica
ti
on
,
the
he
a
ve
or
ve
r
ti
c
a
l
moveme
nt
of
R
OV
is
be
in
g
tes
ted
e
xpe
r
im
e
ntally.
R
e
a
l
-
ti
me
inpu
t
-
output
e
xpe
r
im
e
ntal
da
ta
wa
s
ga
ther
e
d.
5
s
tep
s
ne
e
d
to
be
c
on
s
ider
e
d
in
im
pleme
nti
ng
s
ys
tem
identif
ica
ti
on.
F
igur
e
1
s
how
s
the
5
s
teps
f
or
the
S
I
a
ppr
oa
c
h.
T
he
s
teps
a
r
e
obs
e
r
va
ti
on
a
nd
da
ta
ga
ther
ing,
model
s
tr
uc
tur
e
s
e
lec
ti
on,
m
ode
l
e
s
ti
mation
,
model
v
a
li
da
ti
on
,
a
nd
model
a
p
pli
c
a
ti
on
[
37]
.
T
he
r
e
a
r
e
two
s
e
ts
of
da
ta
ne
e
de
d
f
or
thi
s
e
xp
e
r
im
e
nt:
tr
a
ini
ng
a
nd
va
li
da
ti
on
da
ta
.
T
he
mul
ti
-
s
ine
s
ignal
wa
s
us
e
d
to
ge
t
the
e
xpe
r
im
e
ntal
da
ta
f
o
r
t
r
a
ini
ng
a
nd
va
li
da
ti
on
.
T
he
input
a
nd
ou
tput
da
ta
we
r
e
r
e
c
or
de
d,
a
nd
the
tr
a
ns
f
e
r
f
unc
ti
on
o
f
the
s
ys
tem
wa
s
e
s
ti
mate
d
us
ing
the
M
a
tl
a
b
tool
box
.
T
he
input
given
to
t
he
R
OV
s
ys
tem
c
a
n
be
a
puls
e
,
s
teps
,
r
a
ndom
binar
y
s
e
que
nc
e
(
R
B
S
)
,
ps
e
udo
-
r
a
ndom
binar
y
(
P
R
B
S
)
,
m
-
leve
l
ps
e
udo
-
r
a
ndom
(m
-
P
R
S
)
,
a
nd
mul
ti
-
s
ine
[
37]
.
I
n
the
M
a
tl
a
b
c
omput
ing
e
nvir
onment
,
mul
ti
-
s
in
e
input
wa
s
given
to
t
he
s
ys
tem
a
nd
the
ins
tr
ument
va
r
iable
(
I
V
)
a
ppr
oa
c
h
wa
s
s
e
lec
ted.
T
he
n,
the
s
e
lec
ted
model
s
tr
uc
tur
e
wa
s
im
pleme
nted
f
o
r
model
e
s
ti
mation
a
nd
model
va
li
da
ti
on
to
ge
ne
r
a
te
a
n
R
OV
model.
F
inally,
the
model
ge
ne
r
a
ted
is
us
e
d
t
o
de
s
ign
th
e
R
OV
c
ontr
oll
e
r
.
T
he
e
xpe
r
im
e
nt
wa
s
c
onduc
ted
in
a
c
ontr
ol
led
e
nvir
onment
whe
r
e
the
dis
tur
ba
n
c
e
s
we
r
e
not
c
ons
ider
e
d.
I
n
the
ins
tr
ument
va
r
iable
a
ppr
oa
c
h,
the
I
V
model
wa
s
c
ombi
ne
d
with
3
poles
a
nd
2
z
e
r
os
f
or
the
tr
a
ns
f
e
r
f
unc
ti
on.
T
he
be
s
t
-
f
it
ti
ng
matc
h
wa
s
9
6.
43%
.
T
he
tr
a
ns
f
e
r
f
unc
ti
on
ge
ne
r
a
ted
is
s
hown
in
(
1)
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
S
S
N
:
2502
-
4752
I
ndone
s
ian
J
E
lec
E
ng
&
C
omp
S
c
i
,
Vol.
25
,
No.
3
,
M
a
r
c
h
20
22
:
1344
-
1355
1346
(
)
=
0
.
02332
2
+
0
.
04058
+
0
.
01126
3
+
0
.
7114
2
+
0
.
1861
+
0
.
01398
(
1)
T
he
ge
ne
r
a
ted
output
tr
a
ns
ient
r
e
s
pons
e
is
s
hown
in
F
igur
e
2
.
T
he
output
r
e
s
ult
ha
d
no
ove
r
s
hoot,
18.
18
s
of
T
r
,
33.
21
s
T
s
,
a
nd
0.
1947
o
f
s
tea
dy
s
tate
e
r
r
or
(
S
S
E
)
.
Although
the
R
OV
did
not
ove
r
s
ho
ot,
thes
e
r
e
s
ult
s
s
howe
d
that
the
R
OV
model
ha
d
a
19
.
47
%
tar
ge
t
e
r
r
or
a
nd
i
t
took
a
ppr
oxim
a
tely
ha
lf
a
mi
nute
to
s
tabili
z
e
.
T
his
ge
ne
r
a
ted
model
wa
s
then
s
im
ulate
d
in
M
a
tl
a
b/
S
im
uli
nk
a
s
a
c
los
e
d
-
loop
s
ys
tem
s
ho
wn
in
the
block
diagr
a
m
in
F
igur
e
3.
T
his
ge
ne
r
a
ted
model
wa
s
then
s
im
ulate
d
in
M
a
tl
a
b/
S
im
uli
nk
a
s
a
c
los
e
d
-
loop
s
ys
tem
s
hown
in
the
block
diagr
a
m
in
F
igur
e
3.
F
r
om
F
igu
r
e
4,
the
c
l
os
e
d
-
loop
model
ha
d
a
f
a
s
ter
T
r
(
9.
07
s
)
a
nd
T
s
(
14.
76
s
)
c
ompar
e
d
to
the
ope
n
-
loop
r
e
s
ult
but
the
s
tea
dy
-
s
tate
e
r
r
or
s
hoot
up
to
55.
55
%
f
r
om
the
input
give
n
to
the
s
ys
tem.
F
r
om
the
output
r
e
s
ult
,
the
c
ontr
oll
e
r
ne
e
ds
to
be
a
ppli
e
d
to
ge
t
a
be
tt
e
r
outpu
t
r
e
s
pons
e
.
F
igur
e
1
.
S
ys
tem
identi
f
ica
ti
on
a
ppr
oa
c
h
f
or
model
li
ng
of
R
OV
F
igur
e
2
.
T
r
a
ns
ien
t
r
e
s
pons
e
of
the
R
OV
model
F
igur
e
3
.
M
a
tl
a
b/
S
im
uli
nk
c
los
e
d
-
loop
block
diagr
a
m
Evaluation Warning : The document was created with Spire.PDF for Python.
I
ndone
s
ian
J
E
lec
E
ng
&
C
omp
S
c
i
I
S
S
N:
2502
-
4752
N
e
w
lambda
tuni
ng
appr
oac
h
of
s
ingl
e
input
fuz
z
y
logi
c
us
ing
gr
a
dient
de
s
c
e
nt
…
(
F
auz
al
N
aim
Z
oh
e
di)
1347
F
igur
e
4
.
Ope
n
-
loop
a
nd
c
los
e
loop
t
r
a
ns
ient
outpu
t
c
ompar
is
on
3.
CONT
ROL
L
E
R
DE
S
I
GN
I
n
thi
s
pa
pe
r
,
the
P
I
D
c
ontr
oll
e
r
wa
s
de
s
igned
us
i
ng
a
uto
-
tuni
ng
pr
ovided
by
M
a
tl
a
b/
S
im
uli
nk
.
T
he
S
I
F
L
C
c
ontr
oll
e
r
wa
s
de
s
igned
a
nd
tuned
us
ing
a
he
ur
is
ti
c
method,
GD
A,
a
nd
P
S
O
.
P
I
D
wa
s
us
e
d
a
s
a
ba
s
ic
c
ontr
oll
e
r
to
be
c
ompar
e
d
wi
th
the
S
I
F
L
C
c
ontr
oll
e
r
de
s
igned.
3.
1.
P
I
D
c
on
t
r
oll
e
r
P
I
D
c
ontr
oll
e
r
is
the
ba
s
ic
c
ontr
oll
e
r
a
ppli
e
d
to
the
R
OV
s
ys
tem.
T
he
P
,
I
,
a
nd
D
blocks
we
r
e
put
in
pa
r
a
ll
e
l
in
f
r
ont
o
f
the
plant
to
c
ontr
ol
the
s
ys
tem.
T
he
P
-
block
c
ounter
s
the
dir
e
c
t
e
r
r
or
,
the
I
-
block
r
e
c
ti
f
ies
the
tot
a
l
e
r
r
or
s
in
the
s
ys
tem
while
the
D
-
block
mi
nim
ize
s
the
s
pe
e
d
of
the
e
r
r
o
r
s
.
T
he
P
-
c
ont
r
oll
e
r
w
il
l
make
the
r
e
s
pons
e
f
a
s
ter
but
int
e
nds
to
pr
oduc
e
ove
r
s
ho
ot.
T
he
I
-
c
ontr
oll
e
r
e
li
mi
na
tes
S
S
E
while
the
D
c
ontr
oll
e
r
de
c
r
e
a
s
e
ove
r
s
hoot.
T
he
P
I
D
c
ontr
oll
e
r
block
di
a
gr
a
m
is
s
hown
in
F
igu
r
e
5.
T
he
P
I
D
wa
s
tune
d
us
ing
a
utom
a
ti
c
tuni
ng
in
M
a
tl
a
b/
S
im
u
li
nk
[
38]
.
F
igur
e
5
.
P
I
D
c
ontr
oll
e
r
block
d
iagr
a
m
3.
2.
F
L
C
c
on
t
r
oll
e
r
F
uz
z
y
logi
c
c
ontr
o
ll
e
r
(
F
L
C
)
is
a
human
de
c
is
ion
ba
s
e
d
c
ontr
oll
e
r
int
r
oduc
e
d
by
L
otf
i
A
.
Z
a
de
h
.
I
t
wa
s
int
r
oduc
e
d
in
1965.
W
ha
t
do
be
done
or
r
e
a
c
ti
on
of
the
s
ys
tem
is
ba
s
e
d
on
human
pe
r
s
pe
c
ti
ve
of
the
th
ig
it
s
e
lf
.
In
F
L
C
a
lgor
it
hm,
ther
e
a
r
e
f
our
(
4)
ba
s
ic
c
omponents
;
f
uz
z
if
ica
ti
on,
knowle
dge
ba
s
e
d
(
r
ules
)
,
inf
e
r
e
nc
e
e
ngine
a
nd
de
f
uz
z
if
ica
ti
on.
T
he
f
uz
z
if
i
c
a
ti
on
c
ha
nge
the
r
a
w
da
ta
int
o
membe
r
s
hip
f
unc
ti
one
d,
knowle
dge
ba
s
e
d
c
r
e
a
te
r
ules
f
or
de
c
is
ion
making,
inf
e
r
e
nc
e
e
ngine
whic
h
a
c
t
a
s
int
e
ll
igent
s
ys
tem
a
nd
de
f
uz
z
if
ica
ti
on
c
ha
nge
the
f
uz
z
y
de
c
is
ion
da
ta
int
o
r
e
a
l
output
r
a
w
da
ta.
F
igur
e
6
s
hows
t
he
ba
s
ic
c
onf
igur
a
ti
on
of
F
L
C
c
omponent
in
a
block
d
iagr
a
m
[
39]
.
T
he
c
omm
on
7
X
7
F
L
C
table
is
s
hown
in
T
a
ble
1.
T
he
Z
in
the
table
s
tand
f
or
z
e
r
o
whic
h
indi
c
a
te
the
c
e
nt
e
r
of
the
de
c
is
ion
table
.
P
L
s
tand
f
o
r
pos
it
ive
lar
ge
,
P
M
s
tand
f
or
pos
it
ive
medium
a
nd
P
S
s
ta
nd
f
or
pos
it
ive
s
mall.
T
he
NL
,
NM
,
a
nd
NS
a
r
e
the
oppos
it
e
of
P
L
,
P
M
a
nd
P
S
whe
r
e
they
indi
c
a
te
the
ne
ga
t
ive
s
ide
of
the
table
.
T
he
output
r
e
s
ult
f
o
r
F
L
C
is
s
e
lec
ted
b
a
s
e
d
on
the
table
.
F
igur
e
6
.
B
a
s
ic
c
onf
igur
a
ti
on
of
F
L
C
c
omponent
Evaluation Warning : The document was created with Spire.PDF for Python.
I
S
S
N
:
2502
-
4752
I
ndone
s
ian
J
E
lec
E
ng
&
C
omp
S
c
i
,
Vol.
25
,
No.
3
,
M
a
r
c
h
20
22
:
1344
-
1355
1348
T
a
ble
1
.
7
X
7
F
L
C
table
E
r
r
vs
du/
dt
or
1/
s
PL
PM
PS
Z
NS
NM
NL
NL
Z
NS
NM
NL
NL
NL
NL
NM
PS
Z
NS
NM
NL
NL
NL
NS
PM
PS
Z
NS
NM
NL
NL
Z
PL
PM
PS
Z
NS
NM
NL
PS
PL
PL
PM
PS
Z
NS
NM
PM
PL
PL
PL
PM
PS
Z
NS
PL
PL
PL
PL
PL
PM
PS
Z
3.
3.
S
I
F
L
C
c
on
t
r
oll
e
r
S
I
F
L
C
c
ont
r
oll
e
r
is
de
s
igned
ba
s
e
d
on
c
onve
nti
ona
l
F
L
C
de
s
igned.
T
he
c
onve
nti
ona
l
F
L
C
table
,
T
a
ble
1
is
manipulate
d
us
ing
s
igned
dis
tanc
e
method
(
S
DM
)
whic
h
r
e
duc
e
d
the
r
ules
table
t
o
a
one
-
dim
e
ns
ional
a
r
r
a
y
[
40]
,
[
41]
.
F
r
om
T
a
ble
1,
it
c
a
n
be
s
e
e
n
ther
e
is
a
c
ons
is
t
e
nt
pa
tt
e
r
n
in
the
de
c
is
ion
-
making
of
the
F
L
C
output
.
F
r
om
T
a
ble
s
1
a
nd
2
diagona
l
l
ines
we
r
e
c
r
e
a
ted
whic
h
a
r
e
na
med
L
Z
a
nd
L
NS.
‘
d’
is
the
dis
tanc
e
be
twe
e
n
L
Z
a
nd
L
NS
given
by
(
2)
.
T
he
la
mbda
(
λ
)
e
qua
ti
on
is
s
hown
in
(
3
)
a
nd
(
4
)
.
=
+
√
1
+
2
=
√
1
+
2
+
√
1
+
2
(
2)
̇
+
=
0
(
3)
∴
=
−
̇
(
4)
F
igur
e
7
s
hows
the
de
r
ivation
of
d,
whic
h
is
th
e
dis
tanc
e
be
twe
e
n
point
,
Q
,
a
nd
point
,
P
.
T
he
c
onve
nti
o
na
l
F
L
C
table
is
now
r
e
duc
e
d
to
T
a
ble
2
whe
r
e
the
diagona
l
li
ne
wa
s
r
e
pr
e
s
e
nted
by
L
N
L
,
L
NM
,
L
NS,
L
Z
,
L
P
S
,
L
P
M
,
a
nd
L
P
L
while
NL
,
NM
,
N
S
,
Z
,
P
S
,
P
M
,
a
nd
P
L
r
e
pr
e
s
e
nt
the
outpu
t
of
c
or
r
e
s
ponding
diagona
l
li
ne
s
.
F
igur
e
7
.
De
r
ivation
of
d
,
the
dis
tanc
e
be
twe
e
n
poi
nt
Q
a
nd
P
T
his
input
-
output
of
S
I
F
L
C
c
a
n
be
r
e
plac
e
d
by
a
lookup
table
.
S
I
F
L
C
wa
s
then
tuned
us
ing
the
pr
opos
e
d
lambda
(
λ
)
tuni
ng
method.
T
he
va
lue
of
(
λ
)
wa
s
he
ur
is
ti
c
a
ll
y
c
ha
nge
d
to
obtain
the
be
s
t
output
.
T
he
(
λ
)
wa
s
li
nke
d
to
the
F
L
C
by
t
he
input
of
the
F
L
C
.
T
he
r
a
nge
of
e
r
r
or
a
nd
int
e
gr
a
l
e
r
r
o
r
wa
s
pl
ott
e
d
in
a
gr
a
ph
s
hown
in
F
igu
r
e
8
.
T
a
ble
2
.
R
e
duc
e
d
F
L
C
table
us
ing
S
DM
d
L
N
L
L
N
M
L
N
S
LZ
L
P
S
L
P
M
L
P
L
out
NL
NM
NS
Z
PS
PM
PL
LZ
L
NS
Evaluation Warning : The document was created with Spire.PDF for Python.
I
ndone
s
ian
J
E
lec
E
ng
&
C
omp
S
c
i
I
S
S
N:
2502
-
4752
N
e
w
lambda
tuni
ng
appr
oac
h
of
s
ingl
e
input
fuz
z
y
logi
c
us
ing
gr
a
dient
de
s
c
e
nt
…
(
F
auz
al
N
aim
Z
oh
e
di)
1349
F
igur
e
8
.
P
lot
ted
gr
a
ph
o
f
inpu
t
2
ve
r
s
us
input
1
F
L
C
3.
4.
S
I
F
L
C
h
e
u
r
is
t
ic
t
u
n
i
n
g
m
e
t
h
od
T
he
gr
a
dient
o
f
the
li
ne
is
lambda
(
λ
)
.
T
he
lamb
da
(
λ
)
wa
s
va
r
ied
up
a
nd
down
he
ur
is
ti
c
a
ll
y.
T
he
va
r
iation
o
f
(
λ
)
-
tuned
S
I
F
L
C
r
e
s
ult
wa
s
then
a
na
lyze
d,
a
nd
the
be
s
t
r
e
s
ult
w
a
s
s
e
lec
ted.
F
igur
e
9
s
hows
the
f
low
diagr
a
m
of
the
he
ur
is
ti
c
tuni
ng
pr
oc
e
s
s
.
F
igur
e
9
.
F
low
diagr
a
m
f
or
S
I
F
L
C
he
ur
is
ti
c
tuni
ng
3.
5.
S
I
F
L
C
GDA
t
u
n
i
n
g
m
e
t
h
od
GD
A
is
a
n
a
lgor
it
hm
that
it
e
r
a
ti
ve
ly
r
uns
unti
l
i
t
r
e
a
c
he
s
the
mi
nim
um
va
lue
of
a
f
unc
ti
on.
T
he
GD
A
is
us
e
d
to
r
e
plac
e
the
he
ur
is
ti
c
lambd
a
(
λ
)
tuni
ng
f
or
S
I
F
L
C
.
T
he
objec
ti
ve
f
unc
ti
on
wa
s
obtaine
d
f
r
om
the
pr
e
dicte
d
output
whe
n
c
ompar
e
d
to
the
input
given.
I
t
is
a
s
im
ple
mathe
matica
l
method
that
is
ba
s
e
d
on
the
dif
f
e
r
e
nti
a
l
e
qua
ti
on
w
he
r
e
the
ini
t
ial
point
ou
tput
wa
s
moved
towa
r
ds
the
tar
ge
ted
output
by
c
a
lcula
ti
ng
the
e
r
r
o
r
s.
De
r
ivative
o
f
a
n
objec
ti
ve
f
unc
ti
on
will
de
ter
mi
ne
the
we
ight
o
f
the
objec
ti
ve
f
unc
ti
on
f
o
r
the
ne
xt
point
.
T
wo
(
2)
i
mpor
tant
pa
r
a
mete
r
s
a
r
e
c
ons
ider
e
d
whic
h
a
r
e
the
dir
e
c
ti
on
of
moveme
nt
a
nd
the
s
ize
of
the
s
tep.
T
he
dir
e
c
ti
on
of
moveme
nt
is
de
f
ined
by
th
e
tange
nt
of
the
ini
ti
a
l
point
.
T
he
s
tee
pne
s
s
of
the
tange
nt
li
ne
a
ls
o
s
hows
how
ne
a
r
the
point
is
to
the
mi
n
i
mum
point
a
nd
inf
luenc
e
s
the
c
hoice
of
the
lea
r
n
ing
r
a
te.
F
igur
e
10
s
hows
the
f
low
diagr
a
m
of
the
gr
a
dient
de
s
c
e
nt
a
lgor
it
hm
[
42]
.
T
he
ba
s
ic
e
q
ua
ti
on
is
s
how
n
in
(
5
)
.
1
=
0
−
(
∗
)
(
5)
W
he
r
e
,
0
=
c
ur
r
e
nt
pos
it
ion
1
=
ne
xt
pos
it
ion
E
r
r
=
e
r
r
o
r
(
0
−
0
)
L
=
L
e
a
r
ning
r
a
te
Evaluation Warning : The document was created with Spire.PDF for Python.
I
S
S
N
:
2502
-
4752
I
ndone
s
ian
J
E
lec
E
ng
&
C
omp
S
c
i
,
Vol.
25
,
No.
3
,
M
a
r
c
h
20
22
:
1344
-
1355
1350
F
igur
e
10
.
F
low
diagr
a
m
o
f
g
r
a
dient
de
s
c
e
nt
a
lgor
i
thm
3.
6.
S
I
F
L
C
P
S
O
t
u
n
i
n
g
m
e
t
h
od
P
S
O
wa
s
p
r
opos
e
d
by
[
43
]
in
1995.
I
t
is
ins
pir
e
d
by
the
be
ha
vior
s
o
f
f
is
he
s
s
c
hooli
ng
a
nd
bi
r
ds
f
locking
to
s
e
a
r
c
h
f
or
f
oods
tuf
f
a
t
a
c
e
r
tain
s
pe
e
d
a
nd
pos
it
ion.
T
he
li
ke
ne
s
s
is
r
e
c
ognize
d
be
twe
e
n
a
pa
r
ti
c
le
a
nd
a
s
wa
r
m
e
leme
nt
[
34]
,
[
44]
.
T
he
pa
r
ti
c
le
mov
e
ment
is
c
a
tegor
ize
d
by
two
f
a
c
tor
s
:
it
s
c
ur
r
e
nt
p
os
it
ion
x
a
nd
ve
locity
v
,
r
e
s
pe
c
ti
ve
ly.
I
t
ha
s
be
e
n
us
e
f
ul
e
f
f
e
c
ti
ve
ly
to
a
va
r
iety
of
opti
mi
z
a
ti
on
pr
oblems
[
4
5]
–
[
47]
.
T
he
pa
r
ti
c
le
s
wa
r
m
opti
mi
z
a
ti
on
a
lgor
it
hm
is
a
na
lyze
d
by
us
ing
s
tanda
r
d
r
e
s
ult
s
f
r
om
the
dyna
mi
c
theor
y
[
48]
.
T
he
P
S
O
a
lgor
it
h
m
be
gins
by
ini
ti
a
li
z
ing
t
he
s
wa
r
m
r
a
ndoml
y
in
the
s
e
a
r
c
h
s
pa
c
e
.
T
wo
c
o
ns
e
c
uti
ve
it
e
r
a
ti
ons
,
t
a
nd
t
+
1
c
or
r
e
s
pond
to
the
pos
it
ion
x
of
e
a
c
h
pa
r
ti
c
le
that
c
ha
nge
s
dur
ing
the
i
ter
a
ti
ons
b
y
a
dding
a
ne
w
ve
locity
v.
T
he
ne
w
ve
locity
is
e
s
ti
mat
e
d
by
s
umm
ing
a
n
incr
e
ment
to
the
pr
e
vious
ve
locity
va
lue.
T
he
incr
e
ment
is
a
f
unc
ti
on
o
f
two
c
omponents
r
e
pr
e
s
e
nti
ng
c
ognit
ive
a
nd
s
oc
ial
knowle
dge
[
49]
.
T
he
c
ognit
ive
knowle
dge
of
e
a
c
h
pa
r
ti
c
le
is
include
d
by
e
va
lua
ti
ng
the
dif
f
e
r
e
nc
e
be
twe
e
n
the
c
ur
r
e
nt
po
s
it
ion
x
a
nd
it
s
be
s
t
pos
it
ion,
pbe
s
t
.
T
he
s
oc
ial
knowle
dge
of
e
a
c
h
pa
r
ti
c
le
is
incor
por
a
ted
thr
ough
the
dif
f
e
r
e
nc
e
be
twe
e
n
it
s
c
ur
r
e
nt
pos
it
ion
x
a
nd
the
be
s
t
s
wa
r
m
global
pos
it
ion
a
c
hieve
d,
gbe
s
t
.
He
c
ognit
ive
a
nd
s
oc
ial
kn
owle
dge
f
a
c
tor
s
a
r
e
mul
ti
pli
e
d
by
r
a
ndoml
y
uni
f
or
m
ge
ne
r
a
ted
ter
ms
r
e
s
pe
c
ti
ve
ly
[
49]
.
E
qua
ti
on
(
6
)
s
hows
the
pos
it
ion
ve
c
tor
while
(
7
)
s
hows
the
ve
locity
ve
c
tor
.
P
in
the
e
qua
ti
on
is
pbe
s
t
while
G
is
gbe
s
t
.
+
1
⃗
⃗
⃗
⃗
⃗
⃗
⃗
⃗
⃗
=
⃗
⃗
⃗
⃗
+
+
1
⃗
⃗
⃗
⃗
⃗
⃗
⃗
⃗
⃗
(
6)
+
1
⃗
⃗
⃗
⃗
⃗
⃗
⃗
⃗
⃗
=
⃗
⃗
⃗
⃗
+
1
1
(
⃗
⃗
⃗
⃗
−
⃗
⃗
⃗
⃗
)
+
2
2
(
⃗
⃗
⃗
⃗
−
⃗
⃗
⃗
⃗
)
(
7)
whe
r
e
,
+
1
⃗
⃗
⃗
⃗
⃗
⃗
⃗
⃗
⃗
=
ne
xt
pos
it
ion
⃗
⃗
⃗
⃗
=
c
ur
r
e
nt
pos
it
ion
+
1
⃗
⃗
⃗
⃗
⃗
⃗
⃗
⃗
⃗
=
ve
locity
⃗
⃗
⃗
⃗
=
e
ne
r
ti
a
(
maintain
c
ur
r
e
nt
moveme
nt
dir
e
c
ti
on)
w
=
we
ight
(
incr
e
a
s
e
a
nd
de
c
r
e
a
s
e
e
xploi
tation
a
nd
e
xplor
a
ti
on)
⃗
⃗
⃗
⃗
=
pe
r
s
ona
l
be
s
t
pos
it
ion
⃗
⃗
⃗
⃗
=
g
r
oup
be
s
t
pos
it
ion
1
&
2
=
the
im
pa
c
t
f
a
c
tor
s
1
&
2
=
the
r
a
ndom
va
lue
(
0
–
1)
t
=
number
of
e
ter
a
ti
on
4.
RE
S
UL
T
S
T
he
output
f
or
a
ll
c
ontr
oll
e
r
s
de
s
igned
wa
s
c
ombi
ne
d
int
o
one
block
diagr
a
m
to
c
ompar
e
the
r
e
s
ult
.
T
he
r
e
a
r
e
6
s
ignals
a
na
lyze
d
whic
h
a
r
e
s
tep
inp
ut,
ope
n
-
loop,
c
los
e
d
-
l
oop,
P
I
D
,
S
I
F
L
C
he
ur
is
ti
c
,
S
I
F
L
C
GD
A,
a
nd
S
I
F
L
C
P
S
O.
F
igu
r
e
1
1
s
hows
the
block
diagr
a
m
f
or
the
6
s
ignals
inves
ti
ga
ted.
F
r
om
t
he
block
diagr
a
m,
S
c
ope
1
s
hows
the
6
s
ignals
whi
le
s
c
ope
2
is
us
e
d
to
c
ompar
e
be
twe
e
n
P
S
O
r
e
s
ult
whe
n
the
Evaluation Warning : The document was created with Spire.PDF for Python.
I
ndone
s
ian
J
E
lec
E
ng
&
C
omp
S
c
i
I
S
S
N:
2502
-
4752
N
e
w
lambda
tuni
ng
appr
oac
h
of
s
ingl
e
input
fuz
z
y
logi
c
us
ing
gr
a
dient
de
s
c
e
nt
…
(
F
auz
al
N
aim
Z
oh
e
di)
1351
e
qua
ti
on
wa
s
us
e
d
(
S
I
F
L
C
P
S
O)
a
nd
P
S
O
r
e
s
ult
whe
n
the
lookup
table
wa
s
us
e
d
(
S
I
F
L
C
P
S
O1)
.
T
his
s
c
ope
o
utput
s
hows
a
n
identica
l
r
e
s
ult
(
F
igur
e
1
2
)
a
s
e
xpe
c
ted.
F
igur
e
12
s
hows
the
output
r
e
s
ult
f
or
s
c
ope
1.
I
n
F
igur
e
1
3
,
S
I
F
L
C
GD
A
s
howe
d
the
mos
t
identica
l
r
e
s
ult
to
the
s
tep
input
given.
I
t
wa
s
then
f
oll
owe
d
by
the
S
I
F
L
C
he
ur
is
ti
c
.
T
he
S
I
F
L
C
P
S
O
s
howe
d
im
pr
ove
ment
in
the
T
r
but
with
a
s
li
ght
s
tea
dy
-
s
tate
e
r
r
or
.
T
he
P
I
D
s
howe
d
s
ome
ove
r
s
hoot
but
no
s
tea
dy
-
s
tate
e
r
r
or
.
T
he
output
r
e
s
ult
is
tabula
ted
in
T
a
ble
3.
F
igur
e
11
.
B
lock
diagr
a
m
f
or
the
6
s
ignals
inves
ti
ga
ted
F
igur
e
12
.
C
ompar
is
on
r
e
s
ult
be
twe
e
n
P
S
O
r
e
s
ult
us
ing
c
omm
a
nd
windows
(
S
I
F
L
C
P
S
O
)
a
nd
S
im
ul
ink
(
S
I
F
L
C
P
S
O1)
Evaluation Warning : The document was created with Spire.PDF for Python.
I
S
S
N
:
2502
-
4752
I
ndone
s
ian
J
E
lec
E
ng
&
C
omp
S
c
i
,
Vol.
25
,
No.
3
,
M
a
r
c
h
20
22
:
1344
-
1355
1352
F
igur
e
13
.
T
he
output
r
e
s
ult
of
s
c
ope
1
T
a
ble
3
.
Output
r
e
s
ult
of
the
c
ontr
oll
e
r
s
'
im
pleme
n
tation
to
R
OV
s
ys
tem
P
I
D
S
I
F
L
C
H
e
ur
is
ti
c
S
I
F
L
C
G
D
A
S
I
F
L
C
P
S
O
T
r
(
s
)
7.0665
7.2529
0.7992
2.3686
T
s
(
s
)
24.6687
10.9736
0.9790
12.2348
%
O
S
7.3613
0.7988
0.1021
16.2368
SSE
0
0
0
0.1
F
r
om
the
ba
r
c
ha
r
t
in
F
igu
r
e
1
4
,
it
is
obvious
that
the
S
I
F
L
C
GD
A
c
ontr
o
l
method
s
howe
d
the
be
s
t
r
e
s
ult
a
s
it
mana
ge
d
to
ge
t
the
lowe
s
t
va
lues
f
or
a
l
l
the
pe
r
f
or
manc
e
pa
r
a
mete
r
s
.
I
n
ter
ms
of
the
r
is
e
ti
me,
T
r
,
the
S
I
F
L
C
GD
A
r
e
c
or
de
d
0
.
7992
s
while
the
S
I
F
L
C
he
ur
is
ti
c
wa
s
the
wor
s
t
a
t
7
.
2592
s
,
ne
a
r
ly
ten
ti
mes
the
S
I
F
L
C
GD
A
r
e
c
o
r
d.
T
he
he
ur
is
ti
c
a
pp
r
oa
c
h
a
n
d
P
I
D
a
ppr
oa
c
h
ha
d
a
lm
os
t
s
im
il
a
r
v
a
lues
a
t
7
s
.
T
he
s
e
r
e
c
or
ds
s
howe
d
that
S
I
F
L
C
GD
A
ha
d
the
f
a
s
tes
t
r
e
s
pons
e
ti
me
to
the
c
ha
nge
of
input
leve
l.
F
or
s
e
tt
l
ing
ti
me,
T
s
,
the
S
I
F
L
C
GD
A
method
r
e
c
or
de
d
0.
9790
s
a
n
d
P
I
D
r
e
c
or
de
d
the
w
or
s
t
a
t
24.
6687
s
.
Othe
r
than
S
I
F
L
C
GD
A,
other
methods
to
s
tabili
z
e
the
R
OV
took
te
n
to
twe
nty
ti
mes
longer
.
F
o
r
the
pe
r
c
e
ntage
of
o
ve
r
s
hoot,
the
S
I
F
L
C
GD
A
s
howe
d
the
be
s
t
r
e
s
ult
whic
h
is
a
t
0.
1021%
.
I
t
is
then
f
oll
owe
d
by
the
S
I
F
L
C
he
ur
is
ti
c
(
0.
7988%
)
,
P
I
D
(
7.
3613
%
)
,
a
nd
S
I
F
L
C
P
S
O
(
16
.
2368%
)
.
T
he
s
e
r
e
s
ult
s
s
howe
d
that
whe
n
the
R
OV
wa
s
c
ontr
oll
e
d
us
ing
the
S
I
F
L
C
GD
A
method,
the
e
r
r
or
to
a
tt
a
in
the
tar
ge
t
wa
s
ve
r
y
s
mall.
F
igur
e
1
5
s
hows
the
c
ompar
is
on
of
S
I
F
L
C
GD
A
with
s
tep
s
ignal.
T
he
S
I
F
L
C
GD
A
looks
ne
a
r
ly
identica
l
to
the
given
s
tep
input
.
F
igur
e
14
.
B
a
r
c
ha
r
t
of
the
pe
r
f
or
manc
e
pa
r
a
mete
r
s
Evaluation Warning : The document was created with Spire.PDF for Python.
I
ndone
s
ian
J
E
lec
E
ng
&
C
omp
S
c
i
I
S
S
N:
2502
-
4752
N
e
w
lambda
tuni
ng
appr
oac
h
of
s
ingl
e
input
fuz
z
y
logi
c
us
ing
gr
a
dient
de
s
c
e
nt
…
(
F
auz
al
N
aim
Z
oh
e
di)
1353
F
igur
e
15
.
C
ompar
is
on
of
S
I
F
L
C
GD
A
with
s
tep
s
ignal
5.
CONC
L
USI
ON
T
wo
c
ontr
oll
e
r
s
with
dif
f
e
r
e
nt
tuni
ng
ha
d
be
e
n
a
ppli
e
d
to
the
R
OV
de
pth
s
ys
tem.
A
ne
w
tuni
ng
a
pp
r
oa
c
h
of
S
I
F
L
C
c
ontr
oll
e
r
ba
s
e
d
on
lambda
(
λ
)
is
pr
opos
e
d
a
nd
wa
s
c
ompar
e
d
with
the
b
a
s
ic
P
I
D
c
ontr
oll
e
r
.
T
he
S
I
F
L
C
GD
A
s
howe
d
the
be
s
t
r
e
s
ult
a
s
it
ha
d
the
lowe
s
t
va
lues
in
a
ll
pe
r
f
or
manc
e
pa
r
a
mete
r
s
inves
ti
ga
ted
whic
h
we
r
e
0.
1021%
(
OS
)
,
0.
7992
s
(
T
r
)
a
nd
0.
9790
s
(
T
s
)
.
T
he
S
I
F
L
C
P
S
O
s
uf
f
e
r
e
d
f
r
om
high
o
ve
r
s
hoot
a
nd
s
ome
s
tea
dy
-
s
tate
e
r
r
or
.
T
he
ba
s
ic
P
I
D
c
ontr
ol
ler
a
ls
o
s
uf
f
e
r
e
d
f
r
om
s
ome
ove
r
s
hoot
a
nd
ha
d
a
long
s
e
tt
li
ng
ti
me.
T
he
S
I
F
L
C
He
ur
is
ti
c
ha
d
a
be
tt
e
r
r
e
s
ult
c
ompar
e
d
to
the
P
I
D
c
ontr
o
ll
e
r
in
T
s
a
nd
%
OS
howe
ve
r
it
wa
s
di
f
f
icult
to
tune
a
nd
r
e
qui
r
e
d
e
x
pe
r
ienc
e
a
nd
ti
me.
T
he
S
I
F
L
C
GD
A
wa
s
a
ble
t
o
obtain
plaus
ibl
e
r
e
s
ult
s
be
c
a
us
e
it
us
e
s
s
pe
c
if
ic
tuni
ng
o
f
the
objec
ti
ve
f
unc
ti
on
whic
h
is
ba
s
e
d
on
a
ll
pa
r
a
mete
r
s
;
%
OS,
T
r
a
nd
T
s
.
On
the
other
ha
nd,
the
S
I
F
L
C
P
S
O
ha
d
a
lar
ge
r
e
r
r
o
r
c
ompar
e
d
to
S
I
F
L
C
GD
A
be
c
a
us
e
the
objec
ti
ve
f
unc
ti
on
us
e
d
wa
s
the
a
b
s
olut
e
mea
n
e
r
r
or
va
lue.
I
t
doe
s
n’
t
s
pe
c
if
ica
ll
y
tune
ba
s
e
d
on
e
a
c
h
pa
r
a
mete
r
.
F
r
om
a
ll
r
e
s
ult
s
,
it
is
pr
ove
n
that
S
I
F
L
C
lambda
(
λ
)
tuni
ng
a
ppr
oa
c
h
s
uc
c
e
s
s
f
ull
y
pr
oduc
e
d
good
output
r
e
s
ult
.
W
it
h
the
im
pleme
ntat
ion
o
f
a
n
opt
i
mi
z
a
ti
on
a
ppr
oa
c
h
s
uc
h
a
s
GD
A
a
nd
P
S
O
,
a
be
tt
e
r
output
r
e
s
ult
c
a
n
be
obtaine
d.
T
his
ha
ppe
n
be
c
a
us
e
it
o
pti
mi
z
e
s
the
va
lue
of
lambda
(
λ
)
to
s
uit
the
inpu
t
output
r
e
s
ult
.
T
he
ob
jec
ti
ve
f
unc
ti
on
s
e
lec
ted
in
r
unning
the
opti
mi
z
a
ti
on
a
ppr
oa
c
h
a
ls
o
plays
a
n
im
po
r
tan
t
r
ole
in
ge
tt
ing
a
good
r
e
s
ult
.
F
o
r
f
utu
r
e
i
mpl
e
menta
ti
on,
va
r
ieties
of
ob
jec
ti
ve
f
unc
ti
ons
im
pleme
ntation
c
a
n
be
s
tudi
e
d
a
nd
pr
opos
e
d
to
the
s
ys
tem.
AC
KNOWL
E
DGE
M
E
NT
S
W
e
wis
h
to
e
xpr
e
s
s
our
gr
a
ti
tude
to
hono
r
a
ble
Unive
r
s
it
y,
Unive
r
s
it
i
T
e
knikal
M
a
lays
ia
M
e
laka
(
UT
e
M
)
.
S
pe
c
ial
a
ppr
e
c
iation
a
nd
gr
a
ti
tude
to
e
s
pe
c
ially
f
or
Unde
r
wa
ter
T
e
c
hnology
R
e
s
e
a
r
c
h
Gr
oup
(
UT
e
R
G)
,
C
e
ntr
e
o
f
R
e
s
e
a
r
c
h
a
nd
I
nnova
ti
on
M
a
na
ge
ment
(
C
R
I
M
)
,
C
e
nter
f
or
R
oboti
c
s
a
nd
I
ndus
tr
ial
Automation
(
C
e
R
I
A)
a
nd
F
a
c
ult
y
of
E
lec
t
r
ica
l
E
n
ginee
r
ing
f
r
om
UT
e
M
f
or
s
uppor
ti
ng
thi
s
r
e
s
e
a
r
c
h.
RE
F
E
RE
NC
E
S
[
1]
W
.
C
he
n,
Y
.
W
e
i,
H
.
L
iu
,
a
nd
H
.
Z
ha
ng,
“
B
io
-
in
s
pi
r
e
d
s
li
di
ng
mode
c
ont
r
ol
le
r
f
or
R
O
V
w
it
h
di
s
tu
r
ba
nc
e
obs
e
r
ve
r
,”
2016
I
E
E
E
I
nt
e
r
nat
io
nal
C
onf
e
r
e
nc
e
on
M
e
c
hat
r
oni
c
s
and
A
ut
om
at
io
n,
I
E
E
E
I
C
M
A
2016
,
pp.
599
–
604,
2016,
doi
:
10.1109/I
C
M
A
.2016.7558631.
[
2]
M
.
S
a
na
p,
S
.
C
ha
udha
r
i,
C
.
V
a
r
ta
k,
a
nd
P
.
C
hi
mur
ka
r
,
“
H
Y
D
R
O
B
O
T
:
A
n
unde
r
w
a
te
r
s
ur
ve
il
la
nc
e
s
w
im
mi
ng
r
obot
,”
P
r
oc
e
e
di
ngs
-
2018
I
nt
e
r
nat
io
nal
C
onf
e
r
e
nc
e
on
C
om
m
uni
c
at
io
n,
I
nf
or
m
at
io
n
and
C
om
put
in
g
T
e
c
hnol
ogy
,
I
C
C
I
C
T
2018
,
vol
.
2018
-
J
a
nua
, pp. 1
–
7, 2018, doi:
10.1109/I
C
C
I
C
T
.2018.832587
2.
[
3]
F
.
H
of
f
ma
nn
a
nd
A
.
B
.
K
e
s
e
l,
“
B
io
lo
gi
c
a
ll
y
in
s
pi
r
e
d
opt
im
iz
a
ti
on
of
unde
r
w
a
te
r
ve
hi
c
le
s
hul
l
ge
ome
tr
ie
s
a
nd
f
in
pr
opul
s
i
on,”
O
C
E
A
N
S 2019
-
M
ar
s
e
il
le
, pp. 1
–
4, 2019, doi:
10.1109/oce
a
ns
e
.2019
.8867134.
[
4]
R
. D
. C
hr
is
t
a
nd R
. L
. W
e
r
nl
i,
T
he
R
O
V
m
anual:
a us
e
r
’
s
gui
de
t
o r
e
m
ot
e
ly
ope
r
at
e
d v
e
hi
c
l
e
s
. 2013.
[
5]
H
.
Y
u,
C
.
G
uo,
a
nd
Z
.
Y
a
n,
“
G
lo
ba
ll
y
f
in
it
e
-
ti
me
s
ta
bl
e
th
r
e
e
-
di
me
ns
io
na
l
tr
a
je
c
to
r
y
-
tr
a
c
ki
ng
c
ont
r
ol
of
unde
r
a
c
tu
a
te
d
U
U
V
s
,”
O
c
e
an E
ngi
ne
e
r
in
g
, vol
. 189, no. M
a
r
c
h, p. 106329, 2019, doi
:
10.1016/j
.oc
e
a
ne
ng.2019.106329.
[
6]
C
.
M
a
i,
S
.
P
e
de
r
s
e
n,
L
.
H
a
ns
e
n,
K
.
J
e
ps
e
n,
a
nd
Z
.
Y
a
ng,
“
M
ode
li
ng
a
nd
C
ont
r
ol
of
I
ndus
tr
ia
l
R
O
V
’
s
f
or
S
e
mi
-
A
ut
ono
mous
S
ubs
e
a
M
a
in
te
n
a
nc
e
S
e
r
vi
c
e
s
,”
I
F
A
C
-
P
ape
r
s
O
nL
in
e
, vol
.
50, no. 1, pp. 13686
–
13691, 2017, doi:
10.1016/j
.i
f
a
c
ol
.2017.08.253
5.
[
7]
A
.
K
ha
dhr
a
oui
,
L
.
B
e
ji
,
S
.
O
tm
a
ne
,
a
nd
A
.
A
bi
c
hou,
“
S
ta
bi
l
iz
in
g
c
ont
r
ol
a
nd
huma
n
s
c
a
le
s
im
ul
a
ti
on
of
a
s
ubma
r
in
e
R
O
V
na
vi
ga
ti
on,”
O
c
e
an E
ngi
ne
e
r
in
g
, vol
. 114, pp. 66
–
78, 2016, doi
:
10.1016/j
.oc
e
a
ne
ng.2015.12.054.
[
8]
Z
.
C
hu,
X
.
X
ia
ng,
D
.
Z
hu,
C
.
L
uo,
a
nd
D
.
X
ie
,
“
A
da
pt
iv
e
tr
a
j
e
c
to
r
y
tr
a
c
ki
ng
c
on
tr
ol
f
o
r
r
e
mot
e
ly
ope
r
a
te
d
ve
hi
c
le
s
c
ons
id
e
r
in
g
th
r
us
te
r
dyna
mi
c
s
a
nd s
a
tu
r
a
ti
on c
on
s
tr
a
in
ts
,”
I
SA
T
r
ans
ac
ti
ons
, vol
. 100, pp. 28
–
37, 2020,
doi
:
10.1016/j
.i
s
a
tr
a
.2019.11.032.
Evaluation Warning : The document was created with Spire.PDF for Python.