TELKOM
NIKA
, Vol.12, No
.2, June 20
14
, pp. 343~3
4
8
ISSN: 1693-6
930,
accredited
A
by DIKTI, De
cree No: 58/DIK
T
I/Kep/2013
DOI
:
10.12928/TELKOMNIKA.v12i2.2112
343
Re
cei
v
ed
Jan
uary 13, 201
4
;
Revi
sed
Ap
ril 17, 2014; Accepted Ma
y
2, 2014
Self-learning PID Control for X-Y NC Position Table with
Uncertainty Base on Neural Network
Hu Xiaoping
*
, Wang Ch
ao
, Zhang Wen
hui, Ma Jing
Institute of
T
e
chno
lo
g
y
,
L
ish
u
i Univers
i
t
y
Chin
a -32
3
0
0
0
,
T
e
lp 0578-
22
712
50, F
a
x
05
78-2
271
25
0
*Corres
p
o
ndi
n
g
author, e-ma
i
l
: 7157
03
595
@
qq.com
A
b
st
r
a
ct
An ad
aptiv
e ra
dical
bas
is fun
c
tion (RBF
) n
e
u
ral
net
w
o
rk PID contro
l sche
m
e for X-Y
pos
ition ta
bl
e
is prop
ose
d
b
y
the pap
er. F
i
rstly, X-Y posi
t
ion t
abl
e
mo
d
e
l is esta
blis
h
ed, contro
ller
base
d
on
neut
ral
netw
o
rk is use
d
to learn a
dap
tive and co
mpe
n
sate unc
ertai
n
ty mod
e
l of X-Y posit
io
n tabl
e, neutral n
e
tw
ork
is use
d
to stu
d
y
mo
de
l. PID
neur
al
netw
o
rk control
l
er
bas
e on
au
g
m
e
n
ted var
i
a
b
le
me
thod is
des
ig
n
ed.
PID controll
er is used as assi
stant
directio
n error control
l
er,
neura
l
net
w
o
rk para
m
eters b
a
se on stoch
a
s
t
ic
grad
ient a
l
g
o
rit
h
m c
an
be
adj
ust ada
ptive
on
line. T
h
e
sim
u
lation r
e
sults show that
the pre
s
ented c
ontro
ll
er
has i
m
p
o
rtant eng
ine
e
ri
ng val
ue.
Ke
y
w
ords
:
RBF
neural n
e
tw
ork
;
Self-lear
ni
ng contro
l
;
X-Y NC positi
on ta
ble
;
PID contro
l
1. Introduc
tion
Along
with t
he
comp
uter and
advan
ced
control te
chn
o
logy d
e
velopment, t
he hi
gh
pre
c
isi
on
digi
tal motor d
r
i
v
e tech
nolo
g
y
has be
co
m
e
the
main
st
ream
of the
developm
ent
o
f
nume
r
ical co
ntrol (NC) table. X-Y NC table is a
plane
control
syste
m
in two dimensi
onal spa
c
es.
It can n
o
t o
n
ly com
p
lete
two di
men
s
i
onal pl
ane
p
r
ocessin
g
, al
so
ca
n be
u
s
ed
a
s
a l
a
rge
prototype ma
chin
e for NC table
、
rob
o
ts and othe
r equipm
ent. But the NC table has
ce
rta
i
n
nonlin
ear an
d
couplin
g ch
a
r
acte
rs, so if the hi
ghe
r co
ntrol accu
ra
cy is need, the traditional PID
control te
ch
n
o
logy i
s
difficult to
meet
the
hi
gh
er control accu
racy req
u
ire
m
ents
[1]
-
[2]. T
o
eliminate the
s
e no
nline
a
r
factors of X-Y NC tabl
e, some advan
ced
co
ntrol strategie
s
are
u
s
ed
for the no
nlin
ear
system [
3
]-[5]. For ex
ample: ad
apti
v
e control an
d rob
u
st
cont
rol [6]-[7], fuzzy
control [8]-[13
], neural network
cont
rol, et al [14]-[17].
The
nonlin
ea
r u
n
ce
rtain
problem
s of
sy
stem
are
con
s
ide
r
ed
by p
aper,
ada
ptive contro
l
strategy
can
achi
eve g
ood
co
ntrol
effect
, but the
co
ntrol
strate
gy is
diffic
u
lt to
c
o
py with the no-
lineari
z
atio
n para
m
eter of
system. “Cha
ttering” p
r
obl
em of variabl
e stru
ct
ure control can no
t be
eliminated.
Robu
st cont
rol
stra
tegy ne
e
d
the upp
er
boun
d of un
certain p
a
rts .
B
ecau
se
neu
ral
netwo
rk
ha
s good l
earni
ng ability, n
eural
network
can a
ppro
a
ch
un
certai
nty model of
the
nonlin
ear sy
stem. Neu
r
al
network co
ntrol strate
gy is used mo
re and mo
re
widely in th
e
nonlin
ear system.
Wei [18], Ma
[19] put forward fuzzy a
daptiv
e co
ntrol scheme, b
u
t there are too many
rules of fuzzy logic to be desig
ned, t
he cal
c
ul
ation with the nu
mber of rul
e
s increase
the
cal
c
ulatio
n burde
n. If rules are too little, t
he fuzzy adaptive cont
rol sche
me can’t ensure t
h
e
control a
c
curacy. Wen [20]
put forwa
r
d
neutral
net
wo
rk
co
ntrol
sch
e
me for the X
-
Y table, b
u
t the
back-propa
ga
tion algo
rith
m need l
a
rg
e amou
nt of ca
l
c
ulatio
n, so it is
difficult in engin
e
e
ring
appli
c
ation.
Wan
g
[21] p
r
opo
sed
a rob
u
st ad
aptive
control meth
o
d
for X-Y ta
bl
e, but this
me
thod
requi
re
s the u
pper b
oun
d o
f
some un
ce
rtain system.
Acco
rdi
ng to
the defe
c
t of
thes
e
control
method
s,
the radial
ba
si
s f
unctio
n
(RBF
) ne
ural
netwo
rk Self-l
earni
ng control strat
egy is
put forwa
r
d for un
certai
nt
y X-Y position table system
by
the pape
r. Fi
rstly, the X-Y position ta
bl
e shaft
sy
ste
m
dynamics
model i
s
esta
blish
ed, and
the
neural network PID cont
roll
er ba
se on a
ugmente
d
variable valu
e method is d
e
s
ign
ed, be
ca
use
of good app
roximation abi
lity of neural netwo
rk, RB
F neural network is u
s
e
d
to achieve self-
learni
ng control. Adaptive adjustme
n
t law of t
he network wei
ghts a
n
d
hidden laye
r para
m
eters i
s
desi
gne
d by
the improve
m
ent sto
c
h
a
s
tic g
r
a
d
i
ent
algorith
m
; the improveme
n
t algo
rithm
can
improve
the
learni
ng
sp
e
ed. PID
co
ntrolle
r i
s
mai
n
controll
er i
n
the
ea
rly
stage
of
con
t
rol.
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 16
93-6
930
TELKOM
NIKA
Vol. 12, No. 2, June 20
14: 343 – 34
8
344
Grad
ually, ne
ural net
work
become
s
mai
n
controller
b
y
learning. At last, simulati
on re
sults
sh
ow
that the prop
ose
d
schem
e
is effectiv
e and ha
s the hi
gher
cont
rol a
c
cura
cy.
2. D
y
namic
Equation o
f
X-Y Position
Platform
X-Y NC po
sitioning ta
ble
system i
s
sh
own i
n
figure
1. The
syste
m
con
s
i
s
t of
the ball
scre
w, se
rvo
driver, servo
motors, indu
strial
co
ntrol e
quipme
n
ts, m
o
tion co
ntrol
cards
and th
e
cod
e
disk sy
stem, et al.
Figure 1. X-Y Position platform
Dynami
c
mo
del of moto
r is ign
o
red,
inertial force and fri
c
tio
n
force and
other
disturban
ce a
r
e con
s
ide
r
ed
by the paper in the
operation process of
the
table, system dynam
i
cs
model can be
written a
s
[21]:
Dx
C
x
F
(1)
Whe
r
e,
x
,
x
are defined a
s
the system mo
vement
spe
e
d
and accel
e
ration sep
a
rately.
D
is defined a
s
positioni
ng table qu
ality.
x
is define
d
as
scre
w with
sli
d
ing blo
c
k displacement;
Cx
is d
e
fined
as visco
us fri
c
tion;
C
is define
d
as viscou
s
friction
co
efficient,
F
is
de
fin
e
d
as
static frictio
n
force a
nd co
ulomb fri
c
ti
on
force an
d the friction Stri
beck effect;
is defined a
s
motor outp
u
t torque.
3. Designed
of PID Con
t
r
o
ller base o
n
Neur
al Netw
o
r
k
Set
H
Cx
F
, then the dynamic m
o
d
e
l can b
e
writ
ten as
Dx
H
(2)
System (2)
doe
s not ex
ist unmo
dele
d
dy
nami
cs
and fri
c
tion
force; the fo
llowing
desi
gne
d co
n
t
roller (3)
can
guara
n
tee th
e stability of the syste
m
.
()
pd
D
xK
e
K
e
H
(3)
Whe
r
e,
d
ex
x
is d
e
fined a
s
po
sition e
r
ror v
e
ctor,
d
x
is defi
ned a
s
the
desi
r
e
d
positio
n,
p
K
and
d
K
are defin
ed a
s
the feedb
ack gain mat
r
ix.
The stability of
the syste
m
base
d
on t
he Lyapuno
v theory can
be guara
n
te
ed by the
contr
o
lle
r
(3).
Ho
wev
e
r, a
c
c
u
rate
mo
d
e
l of X
-
Y N
C
table
sy
ste
m
is difficult
to get i
n
p
r
a
c
tice
engin
eeri
ng; t
he id
eal m
o
d
e
l can
only b
e
built. If the
system
e
s
timated m
odel
a
r
e
defined
a
s
ˆ
D
and
ˆ
H
.
The co
ntroll
er of the estima
ted model is
desi
gne
d as f
o
llows:
ˆ
ˆ
()
pd
D
xK
e
K
e
H
(4)
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
1693-6
930
Self-Learning PID Control f
o
r X-Y NC Posi
tion Tabl
e with Uncertaint
y .... (Hu Xiaoping)
345
The co
ntrol la
w (4
) and
con
t
rol law (5), th
at is
1
ˆ
[]
dp
eK
e
K
e
D
D
x
H
(5)
Whe
r
e
ˆ
DD
D
,
ˆ
H
HH
.Unce
r
tainty modeli
ng of syste
m
will de
cline
control
perfo
rman
ce.
To solve non
linear
effect
of the X-Y ta
ble,
goo
d no
nlinea
r ap
pro
x
imation abili
ty of the
neural net
wo
rk i
s
con
s
ide
r
ed by the
p
aper. F
u
rthe
r, beca
u
se RBF neural n
e
twork i
s
lo
cal
gene
rali
zatio
n
net
wo
rk, so
controller base
o
n
RB
F
ne
ural
n
e
twork
can gre
a
tly speed u
p
the
learni
ng spee
d and avoid l
o
cal mini
mu
m probl
em.
The X-Y NC table of the no
nlinea
r dyna
mic mod
e
l (2
) can be
writte
n as:
(,
,
)
D
xH
f
x
x
x
(6)
Where, the
total c
o
ntrol
consi
s
t of PI
D fee
dba
ck
controlle
r
P
D
and RBF
ne
ural
netwo
rk cont
roller
NN
.
PID feedba
ck controll
er i
s
desi
gne
d
P
Dd
p
Ke
K
e
(7)
RBF neu
ral n
e
twork
contro
ller is de
sig
n
e
d
(,
,
,
)
NN
d
d
M
xx
x
o
(8)
The total cont
rolle
r is de
sig
ned
P
DN
N
(9)
The co
ntrol
system stru
ctu
r
e is de
sig
n
e
d
as follo
w:
Figure 2. Neu
r
al network Self-lea
rning
control sy
stem
Whe
r
e, PID
feedba
ck co
ntrolle
r plays main
co
ntro
l function in
begin
stage
, neura
l
netwo
rk b
e
lo
ngs to le
arn
stage. at thi
s
ti
me, ne
ur
al
ne
twork
co
ntroll
er
ca
n n
o
t co
mplete le
arni
ng,
error
sho
u
ld
be big
ger. B
u
t PID feedb
ack controll
e
r
pa
rticip
ate in com
pen
sat
i
on control, t
he
combi
nation
controlle
r e
n
sure
that the
system i
s
stabl
e. So, a
s
NN
lea
r
ning,
control
function
of
P
D
becom
e more and mo
re small.
Whe
r
e,
RBF
netwo
rk lo
cal
gen
erali
z
atio
n ba
se
d o
n
stoch
a
sti
c
g
r
ad
ient meth
od i
s
u
s
e
d
.
Gau
ssi
an fun
c
tion i
s
u
s
e
d
as th
e mem
b
ership fu
nctio
n
of ne
ural
n
e
twork
hidd
e
n
layer, the
n
t
h
e
output of hidd
en nod
es i
s
[22]
2
(1
)
2
||
|
|
()
e
x
p
(
)
j
j
j
Xc
ok
(10)
Whe
r
e,
c
is the cente
r
of the basi
s
fun
c
tion,
is the widt
h of the basi
s
function.
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 16
93-6
930
TELKOM
NIKA
Vol. 12, No. 2, June 20
14: 343 – 34
8
346
The output of
the output layer is
(2
)
(
1
)
1
()
()
(
)
m
ii
j
j
j
ok
w
k
o
k
(
1,
2
i
)
(11)
Whe
r
e,
()
ij
wk
is we
ights of co
nn
ection hi
dde
n
layer and out
put layer.
The output of
the output layer is
1
()
(
)
()
m
ij
j
j
yk
W
k
k
(
1,
2
i
)
(12)
RBF neu
ral n
e
twork lea
r
ni
ng error i
s
de
fined
1
()
()
()
()
2
T
dd
E
yk
y
k
yk
y
k
(13)
The erro
r sig
nal for onlin
e learni
ng is d
e
f
ined
2
2
1
1
()
(
)
2
i
i
J
kE
k
(14)
Network pa
ra
meters upd
ating equ
ation
s
are de
sig
ned
as follo
ws:
(1
)
(
)
(
)
w
wk
w
k
J
k
w
(15)
(1
)
(
)
(
)
c
ck
c
k
J
k
c
(16)
(1
)
(
)
(
)
kk
J
k
(17)
Whe
r
e ,
w
,
c
and
is pa
rame
ters
of lea
r
ni
ng la
w. Network converge
nce
ca
n b
e
ensure
d
by the above pa
ra
meters upd
ate algorith
m
.
4. Simulation and Analy
s
is
The effective
ness of the
control algo
ri
thm
is illustrated by this
pape
r, para
m
eters of
dynamic m
o
d
e
l are
15
D
,
8
C
.
The de
sired traje
c
tory of X and Y axes for X-Y NC table system a
r
e:
2c
o
s
1
.
5
dX
t
x
;
2s
i
n
2
dY
t
x
PD cont
rolle
r gain
s
are:
dia
g
{
2
0
,
2
0
}
p
K
;
dia
g
{
3
0
,
3
0
}
d
K
Paramete
rs o
f
controlle
r are:
0.
6
w
,
0.5
c
,
0.
5
.
X-Y NC table
system mov
e
ment initial values
a
r
e ze
ro. The sim
u
l
a
tion re
sults
as follow.
The Figu
re 3
- Figure 6 are traj
ecto
ry
tracki
ng g
r
a
ph and tracki
ng error
cu
rve grap
h in th
is
scheme, Fig
u
r
e 7 - Fig
u
re
8 ar
e control moment grap
h.
As are sh
own from Fig
u
re 3, even in
the initial po
sition errors a
r
e larg
er, RB
F neu
ral
network feedforward PI
D
controller can still
trac
k fastly expected
posit
ion trajectory in a
relatively s
h
ort time (t=
5
s); As
c
an be
s
e
en from
th
e figure 4 in
the initial velocity error la
rger
ca
se
s. Neura
l
network PID co
ntrolle
r
stil
l ca
n
g
uarant
ee the
preci
s
e tra
c
king
sp
eed
at ab
out
t =
4s. Co
ntrol m
o
ment is not
big , as ca
n b
e
sho
w
n fro
m
the Figure 5.
RBF n
eural n
e
twork i
s
still
in a
lea
r
nin
g
peri
od
i
n
initi
a
l sta
ge in
th
e control
p
r
o
c
e
ss,
at
this
time, ne
ural network controlle
r with
convent
io
n
a
l PID fe
edb
ack
cont
rolle
r wo
rk tog
e
th
er to
meet trackin
g
error preci
s
e
s
of the joint angle,
with learni
ng, n
eural n
e
two
r
k co
ntrolle
r can
achi
eve bette
r cont
rol effect.
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
1693-6
930
Self-Learning PID Control f
o
r X-Y NC Posi
tion Tabl
e with Uncertaint
y .... (Hu Xiaoping)
347
(a) X pos
ition trajec
tory track
i
ng
(b) Y pos
it
ion trajec
tory
track
i
ng
Figure 3. Position trajecto
ry tracki
ng cu
rves of X-Y table
(a) X speed traj
ecto
ry tracki
ng
(b) Y spe
ed traje
c
t
o
ry trackin
g
Figure 4. Velocity trajecto
ry tracki
ng cu
rves of X-Y table
(a) Co
ntrol input curves of X
(b)
Control in
put curve
s
of
Y
Figure 5. Con
t
rol input cu
rv
es of X-Y tabl
e
5. Conclusio
n
The traj
ecto
ry tracking
control p
r
o
b
le
ms of u
n
certain X-Y NC table sy
ste
m
with
uncertainty a
r
e con
s
ide
r
e
d
. Self-learni
ng control
strategy ba
sed
on ra
dial ba
si
s fun
c
tion ne
ural
netwo
rk i
s
propo
sed by thi
s
pap
er.
1)
Neu
r
al net
wo
rk PID hyb
r
id
controller
ba
sed
o
n
the a
ugmente
d
va
riable m
e
thod
is de
signe
d.
the co
ntrol p
r
eci
s
ion
of the system i
s
en
sur
ed, this
m
e
thod
can
sp
eed u
p
erro
rs conve
r
ge
nce
in early pha
se.
2)
Improved sto
c
ha
stic gra
d
i
ent
algo
rithm
is
de
sign
ed t
o
en
sure onli
ne re
al-time
adju
s
tment of
0
2
4
6
8
10
0
0.
5
1
1.
5
2
2.
5
3
3.
5
4
t/s
Pos
i
t
i
on t
r
ac
k
i
ng
of
X/
m
des
ired
pos
it
ion of
X
real pos
i
t
ion of
X
0
2
4
6
8
10
0
0.
5
1
1.
5
2
2.
5
3
3.
5
4
t/s
Po
s
i
t
i
on
t
r
ac
k
i
ng of
Y
/
m
de
s
i
red
pos
it
ion
of
Y
rea
l
p
o
s
i
t
i
on
of
Y
0
2
4
6
8
10
-2
-1
0
1
2
3
t/s
V
e
l
o
ci
ty tr
a
cki
n
g
o
f
X
/
m
.
s-
1
d
e
s
i
r
ed v
e
lo
c
i
t
y
of
X
r
e
a
l
v
e
l
o
ci
ty o
f
X
0
2
4
6
8
10
-3
-2
-1
0
1
2
3
4
5
t/
s
V
e
l
o
ci
ty tr
a
cki
n
g
o
f
Y
/
m.s-
1
d
e
s
i
r
ed v
e
loc
i
t
y
of
Y
r
eal v
e
lo
c
i
t
y
of
Y
0
5
10
15
20
-10
-5
0
5
10
15
20
t/s
C
ont
r
o
l
in
pu
t
X/
N
.
m
0
5
10
15
20
-10
-5
0
5
10
15
t/
s
C
o
nt
r
o
l i
n
pu
t
Y/
N
.
m
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 16
93-6
930
TELKOM
NIKA
Vol. 12, No. 2, June 20
14: 343 – 34
8
348
the netwo
rk p
a
ram
e
ters, the improve
d
al
gor
ithm
can
speed u
p
the learni
ng spee
d.
3)
The control
mech
ani
sm i
s
an
alysi
s
ed
and
simulate
d by the pap
er, sim
u
lation
results
sho
w
that the control method is
effective.
This cont
rol scheme ca
n
achieve go
od
co
nt
rol ef
fect, and it has hi
ghe
r a
pplication
value.
Ackn
o
w
l
e
dg
ments
Proje
c
t sup
p
o
r
ted by Zhejia
ng Provin
cial
Natural Scie
n
c
e Fou
ndatio
n (No. LY1
4
F
0300
05),
Zhejian
g
Pro
v
incial Edu
c
a
t
ion Dep
a
rtm
ent Scien
c
e
Re
sea
r
ch Pro
j
ect (No.Y201
3300
00).
Zhejian
g
Pro
v
incial Scie
nce and Te
chn
o
l
ogy Proje
c
t (No. 201
3C31
10)
Referen
ces
[1]
Amor J, Willia
m A. Exp
e
rime
ntal stud
y
on
a
c
tive vi
br
ation
control
of a sin
g
le-
L
ink fl
e
x
ib
l
e
man
i
pu
lato
r
usin
g
tools of
fuzz
y
l
o
g
i
c and ne
ural n
e
t
w
o
r
ks.
IEEE Transactions
on In
strum
e
ntation and
Measur
e
m
ent
. 200
5; 54(3): 12
00-1
208.
[2]
Farzane
h A, T
a
le
bi
HA, R
a
jn
ikant V. A
Sta
b
le
Ne
ural
N
e
tw
o
r
k-Bas
ed O
b
server
W
i
th A
pplic
atio
n t
o
F
l
exibl
e
-Jo
i
nt Mani
pul
ators.
IEEE Trans on Neur
al Networ
ks
. 2005; 17(
1)
: 118-12
9
[3]
Dub
o
w
s
k
y
S,
Papa
do
pou
los
E G.
T
he kin
e
m
atics D
y
n
a
mi
cs and
contr
o
l
of free-fl
yin
g
s
pace r
o
b
o
tic
s
y
stems.
IEEE Trans on Ro
bo
tics and Auto
mation
. 19
93; 9(
5): 531-5
43.
[4]
Feng B M, Ma G C, Xie W
N, W
ang C H.
Desig
n
of Inte
llig
ent Co
ntrol
l
e
r for Space
R
obot i
n
task
space.
Jour
na
l of Astronautics
in Chi
nese
. 2
0
07; 28(4): 1
49-
154.
[5]
Che
ah C
C
S,
Ka
w
a
m
u
ra SA
, Lee K
H
. T
u
nin
g
for T
a
sk-Space F
e
ed
ba
ck Contro
l of
Rob
o
t W
i
th
Uncertain Jacobian Matrix
.
IEEE Trans on Automatic Control
. 200
1; 46(8):
1313-
13
18.
[6] Chen
L.
Adaptive and robust composite c
o
ntrol of
coor
dinated motion of
s
pace robot
system
with
pris
matic j
o
int
.
Procee
din
g
s
of the 4th
w
o
rld Co
ngr
ess
on Intel
lig
ent
Contro
l an
d A
u
tomatio
n
,
Shan
gh
ai, Chi
na. 200
2: 125
5
–12
59.
[7]
Z
hang
W
,
F
a
n
g
Y, Ye
X. Ad
a
p
tive
Neur
al
N
e
t
w
o
r
k
Rob
u
st
Contro
l for S
p
a
c
e R
obot
w
i
th
Uncerta
i
nt
y
.
TELKOMNIKA.
2013; 1
1
(3): 5
13-5
20.
[8]
W
ang SD,
Lin
CK. Adaptiv
e c
ontrol
of ro
b
o
t
mani
pul
ator us
ing fuzz
y
comp
ensator.
Fu
z
z
y
Sets Syst
.
200
0; 110(
3): 351-3
63.
[9]
W
ang CH, T
s
ai
CH, Li
n W
S
. Robust fuzz
y mo
del
-fo
llo
w
i
n
g
c
ontrol
of rob
o
t mani
pul
ators.
IEEE Trans.
Fu
z
z
y
Syst
. 20
00; 8(4): 46
2-4
69.
[10]
Lin
C K. No
n-
sing
ular term
in
al sl
idi
ng m
o
d
e
l co
ntrol
of robot ma
ni
pul
ators usi
ng fuzz
y
w
a
ve
le
t
net
w
o
rks.
IEEE Trans. Fu
z
z
y
Syst
. 2009; 16
0(12): 17
65-
17
86.
[11]
Yoo BK, Ha
m W
C
. Adap
tive contro
l o
f
robot mani
p
u
lator
usin
g fuzz
y
c
o
mp
en
sator.
IEEE
T
r
ans.F
u
z
z
y
S
yst
. 2000; 8(2): 186-1
99.
[12]
Z
hang
X D, Ji
a Q X, S
un H
X,
C
hu M. T
he R
e
se
arch
of Space
Ro
b
o
t F
l
e
x
ib
le J
o
i
n
t T
r
ajector
y
Contro
l.
Journ
a
l of Astrona
uti
cs in Chi
nese
.
200
8; 29(6): 18
65-1
869.
[13]
Lin
CKH. Re
inf
o
rceme
n
t lear
n
i
ng c
ontrol
of r
obot ma
ni
pul
ators usi
ng fuzz
y
w
a
ve
let n
e
t
w
o
r
ks.
Fu
z
z
y
Sets Syst
. 2009; 160(1
2
): 176
5 -178
6.
[14]
Hon
g
Z
D, Yun
C, Che
n
L. Mo
deli
ng
an
d T
r
ajec
tor
y
T
r
ackin
g
Co
ntrol
of a F
r
ee-F
l
oati
ng Space Rob
o
t
w
i
t
h
F
l
e
x
i
b
le M
ani
pul
ators.
Jo
urna
l of Rob
o
t in Chi
nese
.
200
7,29(1):9
2-9
6
.
[15]
Le
w
i
s
F
L
, Ki
m YH. Intell
ig
ent o
p
timal
c
ontrol
of ro
bo
tic mani
pu
lato
rs usin
g
neur
al n
e
t
w
orks
.
Autom
a
tica
. 2
0
00; 36(9): 1
355
- 1364.
[16]
Psillak
i
s HE, A
l
e
x
a
ndri
d
is AT
.
Adaptiv
e ne
ur
al
motio
n
co
ntrol of n-
l
i
nk ro
b
o
t ma
ni
pul
ator
s subject to
unkn
o
w
n
distu
r
banc
es an
d s
t
ochastic p
e
rtu
r
batio
ns
. Pro. Inst. Electr. Eng, Co
ntrol T
heor
y Ap
pl.
200
6; 153(
2): 127-1
38.
[17]
Hon
g
Z
B
, C
h
e
n
L. S
e
lf-l
earn
i
ng c
ontrol
of s
pace
fle
x
i
b
l
e
mani
pul
ator
ba
se o
n
gauss
fu
nction
fuzz
y
neur
al net
w
o
rk
.
Engine
eri
ng Mecha
n
ics
in Chin
ese
. 2
008;
29(6): 172-
17
7.
[18]
W
e
i L
X
, T
i
an XJ, W
ang
HR,
and et a
l
. Ba
sed o
n
ad
apti
v
e fuzz
y
c
ontr
o
l
X-Y tabl
e sl
idin
g mo
de
control.
Electri
c
Drive.
200
9;1
1
(7): 59–
63.
[19]
Ma J, Z
hang W
,
Z
hu H. Adapt
ive C
ontrol
for Robotic Ma
nip
u
lators b
a
s
e
on RBF
Neu
r
al Net
w
ork.
TELKOMNIKA.
2013; 1
1
(3): 5
21-5
28.
[20]
W
en SH, Qian
g GB. Based o
n
improv
ed E
l
man n
e
t
w
ork
X-Y positi
oni
ng t
abl
e force/p
o
si
tion co
ntrol.
Machi
ne an
d H
y
drau
lic in C
h
i
nese
. 20
09; 33
(7): 117-1
21.
[21]
W
ang Z
H
, W
a
ng
XS,
Xu W
L
. X-Y
p
o
sitio
n
in
g tab
l
e
of t
he r
obust
ad
a
p
tive fricti
on c
o
mpe
n
satio
n
.
Journ
a
l of Sout
heast Un
iversit
y
. 2002; 32(
1): 69-7
2
.
[22]
Abdess
a
lam A
,
Anas M, Ali A. Design S
i
m
u
la
ti
on a
nd H
a
rd
w
a
r
e
imp
l
e
m
entatio
n of L
o
w
De
nsit
y
Parit
y
Check
Decoders us
ing Min-Sum
A
l
gor
i
thm.
Internati
o
nal
Jour
na
l of
Co
mp
uter Sci
e
nce Issu
es
.
201
2; 9(3): 83-
92.
Evaluation Warning : The document was created with Spire.PDF for Python.