TELKOM
NIKA
, Vol.11, No
.5, May 2013, pp. 2301
~
230
8
ISSN: 2302-4
046
2301
Re
cei
v
ed
Jan
uary 10, 201
3
;
Revi
sed Fe
brua
ry 26, 20
13; Accepted
March 11, 20
13
Fuzzy Dynamic Scheduling of Multi-loop NCS
Fang He*
1,a
, Jiajia
Pang
1,b
, Qiang Wan
g
2,c
, Zhijie Z
h
ang
1,d
1
School of Elec
trical Eng
i
ne
eri
ng, Univ
ersit
y
of Jinan, Jin
an,
Chin
a
2
School of
Mec
han
ical En
gi
ne
erin
g, Univers
i
ty of Jin
an, Jin
a
n
, Chin
a
*Corres
p
o
ndi
n
g
author, e-ma
i
l
: hefang
75
88
@16
3
.com*
a
, 4596
36
121
@qq
.
com
b
, me_w
a
ngq@ujn.edu.cn
c
,
cse_zzj@ ujn.e
du.cn
d
A
b
st
r
a
ct
T
here ar
e
mor
e
pro
b
l
e
ms
of
netw
o
rk data
trans
mi
ssi
on
i
n
multi-
loo
p
N
C
S than
in s
i
n
g
le
loo
p
NCS. It mayb
e
lead to th
at o
ne or
mor
e
of close
d
-l
o
op sy
stems b
e
co
me
unstab
l
e i
n
multi-lo
op N
C
S. In
order to solve
this proble
m
,
a
novel al
gor
ithm of fu
zz
y
dyna
mic sche
duli
ng for mult
i-loo
p
NCS is pu
t
forw
ard in this pap
er. F
i
rstly, the req
u
ire
m
ent
of the
samp
li
n
g
peri
od of NC
S is discusse
d. T
he necessity
of
dyna
mic sche
d
u
lin
g for
mu
lti-l
oop
NCS is
an
aly
z
e
d
. A
nd th
en, the
alg
o
rith
m of fu
zz
y
dyn
a
mic sch
edu
lin
g
for multi-
lo
op
NCS is pro
pos
ed. T
he id
ea i
s
describ
ed fo
rm thre
e parts
.T
he princi
ple
of fu
zz
y
dy
na
mi
c
sched
uli
ng is
ana
ly
z
e
d. T
h
e
fu
zz
y c
ontro
ll
er is des
i
gne
d
.
T
he fu
zz
y
dy
na
mic sch
edu
l
e
r is construct
ed.
Finally, the si
mulati
on
mod
e
l
of a mu
lti-lo
op
NCS w
i
th
fu
zzy dyna
mic sch
edu
ler is set u
p
usin
g T
r
ueT
i
m
e
softw
are. T
he relev
ant an
alysi
s of simul
a
tion
is given.
T
he result of study can
prov
e that the perfor
m
a
n
c
e
of each c
l
os
ed
control
loo
p
b
a
sed
on th
e s
a
me n
e
tw
ork platfor
m
is
i
m
p
r
oved
by a
ddi
ng fu
zz
y
dy
na
mi
c
sched
uler i
n
th
e mu
lti-lo
op N
C
S.
Ke
y
w
ords
: mu
lti-loo
p
NCS, fu
zz
y
co
ntrol, dy
na
mic
sch
edu
li
ng, the sa
mpl
i
n
g
peri
od, pri
o
rity
Copy
right
©
2013 Un
ive
r
sita
s Ah
mad
Dah
l
an
. All rig
h
t
s r
ese
rved
.
1. Introduc
tion
Comp
ared with traditional
control sy
stem with p
o
i
n
t to point control mo
d
e
, NCS
(Net
wo
rked
Control Syste
m
) ha
s a
d
v
anta
ges of
less wi
ring,
resou
r
ce
sha
r
ing, and conve
n
ient
of
fault diagno
si
s and m
a
inte
nan
ce. On th
e other h
and,
there are so
me que
stion
s
maybe eme
r
ge
and n
eed to
be solved, su
ch a
s
delay
of netwo
rk
transmi
ssion, seq
uen
ce error
of
tran
smitting
data pa
cket, packet lo
ss,
etc. All this questio
n
s h
a
ve bad influe
n
c
e on the p
e
rforman
c
e of
NCS.
The pe
rform
ance of NCS depen
ds
on not only
t
he co
ntrol
strategie
s
but
also
sched
uling
strategi
es of
netwo
rk. So it is
necessa
ry to design a reasona
ble
ne
twork sche
dul
ing algo
rithm
to
optimize the
perfo
rman
ce
of overall NCS.
The current rese
arche
s
o
n
netwo
rk
sch
edulin
g
mo
stly focus o
n
th
e appli
c
ation
layer of
netwo
rk.
Th
e
y
mainly d
r
a
w
from the
ta
sk sch
edul
i
n
g
algo
rithm
s
of
CP
U, such a
s
the
typical
FP
(Fix Priority),
the RM (Rate
Monotoni
c) [
2
], the
EDF (Earlie
st Dea
d
line First) [3], etc. FP and
RM
sched
uling al
gorithm
s bel
o
ng to the stati
c
sche
du
lin
g method, which the flexibility of sched
uli
n
g
pro
c
e
ss i
s
p
oor. Even EDF is dyna
mic sche
duli
ng, whi
c
h di
stribute
s
dyn
a
mic the p
r
i
o
rity
according to
time deadli
n
e
of each
cont
rol loo
p
, its p
e
rform
a
n
c
e o
f
adaptive is
poor. Be
side
s
sched
uling
al
gorithm
s a
b
o
v
e, some
sch
o
lars
p
r
op
ose dynami
c
scheduli
ng al
go
rithm ba
se
d
on
"dead
zo
ne"
[4], static
sched
uling
a
l
gorithm
ba
sed o
n
"time
win
d
o
w
" [5
], and dyn
a
m
ic
sched
uling al
gorithm of M
E
F-TO
D (Ma
x
imum Erro
r First – Try On
ce Di
scard) [
6
].
Based
on th
e origi
nal E
D
F sch
eduli
ng algo
rithm
,
we put forward fuzzy
dynamic
sched
uling al
gorithm fo
r multi-loo
p
NCS to r
edu
ce the bad inf
l
uen
ce of net
work
conflict
and
improve the p
e
rform
a
n
c
e o
f
NCS.
This pa
per is organi
zed
in
to 6
se
ction
s
incl
udin
g
thi
s
se
ction. Se
ction
2 a
naly
z
e
s
the
influen
ce of the sa
mpling
perio
d on pe
rforman
c
e
of
NCS. Sectio
n
3 discus
se
s
the necessity of
dynamic sch
edule
for m
u
lti-loop
NCS. The fu
zz
y d
y
namic
sche
duling
algo
rit
h
m of m
u
lti-l
oop
NCS i
s
propo
sed
in
se
ctio
n 4. It in
clud
e
s
several
p
a
rt
s: an
alyzin
g t
he p
r
in
ciple
o
f
fuzzy
dyna
mic
sched
uling; g
i
ving the design of fuzzy controlle
r; co
nstru
c
ting th
e fuzzy dyna
mic sche
dule
r
.
Simulation an
d analysi
s
are done in
se
ction 5. Finally, concl
u
si
on
s are p
r
e
s
ente
d
.
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 23
02-4
046
TELKOM
NIKA
Vol. 11, No
. 5, May 2013 : 2301 – 230
8
2302
2. The Influe
nce of Samp
ling Period on the Perfor
mance of
NCS
In
gen
eral
co
ntrol system
with contin
uo
us si
g
nal
s, th
e feed
ba
ck e
r
ror si
gnal
re
ceived by
the co
ntrolle
r is
continu
o
u
s
. So the
sa
mpling p
e
ri
od
doe
s not
affect the
effect
of co
ntrol. B
u
t in
the control
system
with di
screte
digital
sign
al, sampl
i
ng p
e
rio
d
d
e
termine
s
th
e
time interval
of
feedba
ck si
g
nal to control
l
er [7]. NCS is a
spe
c
ia
l
ki
nd of digital
control sy
ste
m
. Its control
l
er
receives p
e
ri
odically the feedba
ck sig
nal
sample
d by sen
s
o
r
s.
Although net
work sou
r
ce are sha
r
ed b
y
multip
le tasks, network permits itsel
f
to b
e
occupi
ed by
a task in
some
spe
c
ifie
d time. In
this case, not
only the sa
mpling p
e
ri
o
d
of
feedba
ck sig
nal but also the time interval of
the network tra
n
smi
ssi
on for fee
dba
ck
sign
al will
influen
ce on
the perfo
rma
n
ce of the
cl
ose
d
l
oop
co
ntrol sy
stem.
As a re
sult, sampli
ng pe
ri
od
whi
c
h is
set too long o
r
to
o sho
r
t may lead to t
he ou
tput of NCS
diverge
n
ce. For exampl
e, we
set up
a sim
u
lation mo
del
of a sin
g
le l
oop
NCS
o
n
l
y
includ
es
on
ly one contro
lled plant
s, o
ne
controlle
r no
de, one
sen
s
or no
de an
d
one a
c
tuato
r
nod
e usi
n
g
TrueTim
e software[8]. Fi
xed
sampli
ng pe
riod is u
s
uall
y
set for sin
g
le loop
NCS, in which
multiple tasks of node
s run
according to time division
multiplexing
model to tran
smit data thro
ugh net
work.
By adjusting the sam
p
ling
perio
d of NCS
，
we ca
n ge
t some cu
rve
s
as sho
w
n i
n
Figure
1.
h
pre
s
e
n
ts the sam
p
lin
g peri
od of
NCS. Fig
u
re
1 a)
sho
w
s the re
sp
on
se
output when
the
value of
the
sampli
ng
pe
ri
od of
NCS i
s
set
t
oo sm
al
l.
Figure 1 c) s
hows the resp
on
se outp
u
t
whe
n
the
val
ue of th
e
sa
mpling
pe
riod
of NCS i
s
se
t too big.
The
s
e
cu
rve
s
p
r
o
v
e that it i
s
v
e
ry
importa
nt for NCS to
ad
opt a p
r
op
er sam
p
ling
p
e
riod
within
a ra
nge to
ensure
a g
o
o
d
perfo
rman
ce
of control system.
0
0.
2
0.
4
0.
6
0.
8
1
-1
-0
.
5
0
0.
5
1
1.
5
x 1
0
7
ti
m
e
(
s
)
r,
y
(rad
/
s
)
a)
h
=3ms
0
0.
2
0.
4
0.
6
0.
8
1
0
50
10
0
15
0
20
0
25
0
30
0
35
0
40
0
45
0
ti
m
e
(
s
)
r,
y
(ra
d /
s
)
b)
h
=12m
s
0
0.
2
0.
4
0.
6
0.
8
1
-10
0
0
10
0
20
0
30
0
40
0
50
0
60
0
70
0
ti
m
e
(
s
)
r,
y
(rad /
s
)
c)
h
=28m
s
Figure 1. Re
spon
se outp
u
t curve
s
u
s
ing
different sa
m
p
ling pe
riod
s
The multi-l
o
o
p
NCS are
u
s
ed
widely
which in
clu
d
e
s
several cl
osed loop
co
ntrol system
based on the
same n
e
two
r
k platform,
su
ch a
s
Figu
re
2. Each cl
ose
d
loop control
system o
w
n
s
itself cont
rolle
d plant, controller, se
nsor
and a
c
tuator.
The ch
oice of the sampling
period is mo
re important fo
r multi-loo
p
NCS than singl
e loop
NCS. Multi-lo
op NCS have
some featu
r
e
s
as follo
w:
1) Different cl
ose
d
loo
p
NCS
on same
n
e
twor
k
platform ha
s itself d
i
fferent p
r
ope
r range
of sam
p
ling
period. The response output of
NCS
will diverge if
sam
p
ling period
beyond
the
rang
e.
2) Be
cau
s
e
of different
close
d
loo
p
NCS on
sam
e
netwo
rk
plat
form, phe
no
menon
of
confli
ct of data transmi
ssi
on maybe h
appe
ns. It l
eads to that multi-loo
p
NCS can n
o
t to be
sched
uled if sampling p
e
rio
d
of each
clo
s
ed lo
op NCS is set un
sui
t
able.
3) Even
sam
p
ling period
of each
closed loop
NCS
is
set suitab
le, schedule is
still
necessa
ry for different
clo
s
ed loo
p
NCS
on same
net
work
platform
to ensure an
d improve th
e
perfo
rman
ce
of each
clo
s
e
d
loop NCS.
3. The Nec
e
s
s
it
y
of NCS D
y
namic Sc
heduling
Based
on
sy
stem
stru
cture sho
w
n of t
h
ree
cl
osed l
oop
s
NCS i
n
Figure 2, a
si
mulation
model of a m
u
lti-loop
NCS
is built usin
g Ture
Time software. It is sh
own a
s
Figu
re 3.
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
2302-4
046
Fuzzy Dynam
ic Sche
dulin
g
of Multi-loop
NCS (Fang
He)
2303
The m
u
lti-loo
p
NCS in
clu
des thre
e
co
ntrol lo
op
s.
Each
co
ntrol
loop i
s
a
cl
ose
d
loo
p
control
syste
m
which in
clude
s
sen
s
o
r
, cont
rolle
r,
actuato
r
and
plant. Ea
ch
co
ntrol
loo
p
indep
ende
nt carry out thei
r tasks. But they sha
r
e
sa
me network to tran
smit the
i
r data. In Fig
u
re
3, ea
ch m
o
d
e
l of net
wo
rk nod
es
ado
pts T
r
ueT
i
m
e t
oolbox of
TrueTime
Kern
el mod
u
le. E
D
F
sched
uling al
gorithm i
s
used
for TrueTi
m
e Network
module.
Figure 2. System stru
cture
of three clo
s
e
d
loop
s NCS
r,
y
3
r,
y
2
r,
y
1
t
T
o
W
o
r
k
s
pac
e5
r
T
o
W
o
r
k
s
pac
e4
y3
T
o
W
o
r
k
s
pac
e3
y2
T
o
W
o
r
k
s
pac
e2
y1
T
o
W
o
r
k
s
pac
e1
S
t
ep1
A/
D
S
nd
S
ens
or
3
A/
D
S
n
d
S
ens
or
2
A/
D
S
n
d
S
ens
or
1
sn
d
1
sn
d
2
sn
d
3
sn
d
4
sn
d
5
sn
d
6
sn
d
7
sn
d
8
sn
d
9
sn
d
1
0
rc
v
1
rc
v
2
rc
v
3
rc
v
4
rc
v
5
rc
v
6
rc
v
7
rc
v
8
rc
v
9
rc
v
1
0
Ne
t
w
o
r
k
Rc
v
S
n
d
I
n
t
e
r
f
er
enc
e
[i
n
s
]
[a
c
r
1
]
[r
]
[
c
or
1]
[
c
or
3]
[
c
or
2]
[a
c
r
2
]
[a
c
r
3
]
305.
25
s
+
2
1.
32s
+
1
9
.
38
2
DC
S
e
r
v
o
3
305.
25
s
+
2
1.
32s
+
1
9
.
38
2
DC
S
e
r
v
o
2
3
05.
25
s
+
21.
32s
+
19.
38
2
DC S
e
r
v
o
1
Rc
v
r
Snd
C
ont
r
o
l
l
e
r
3
Rc
v
r
Snd
C
ont
r
o
l
l
e
r
2
Rc
v
r
Snd
C
ont
r
o
l
l
e
r
1
Cl
o
c
k
Rc
v
D
/
A
A
c
t
uat
or
3
Rc
v
D
/
A
A
c
t
uat
or
2
Rc
v
D
/
A
A
c
t
uat
or
1
[s
c
s
1
]
[
c
os
3]
[s
c
s
3
]
[
c
os
2]
[s
c
s
2
]
[
c
os
1]
[i
n
s
]
[
c
os
2]
[r
]
[
c
os
3]
[a
c
r
3
]
[
c
or
2]
[r
]
[
c
or
3]
[s
c
s
2
]
[s
c
s
1
]
[r
]
[
c
os
1]
[a
c
r
2
]
[
c
or
1]
[a
c
r
1
]
[s
c
s
3
]
Figure 3. The
simulation m
odel of
a
multi-loop NCS wi
thout
sched
ul
er
Figure 4 sho
w
s respon
se
output
cu
rves of three con
t
rol loop
s system, in which
y
1
,
y
2
and
y
3
re
spe
c
tively is
re
spon
se
output
s of th
e 1
st
control l
oop, t
he 2
nd
cont
ro
l loop
an
d th
e 3
rd
control loop.
h
re
pre
s
e
n
ts the
sam
p
ling pe
riod
of
each cl
ose
d
loop
NCS. The three
PID
controlle
rs
a
dopt the
sa
me pa
ram
e
ters: th
e p
r
o
portion
al
coe
fficient
K
P
is 4.2;
K
I
inte
gral
coeffici
ent is
8 and differe
n
t
ial coefficie
n
t
K
D
is 0.02. In
this simulatio
n
, the other p
a
ram
e
ters are:
netwo
rk type
is ‘Roun
d Ro
bin’; the rate
of netwo
rk tra
n
smi
ssi
on i
s
80kbit/s; the
sampli
ng p
e
ri
od
of three control loop re
sp
e
c
tively is
T
1
=4ms
, T
2
=5
ms
, T
3
=6ms.
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 23
02-4
046
TELKOM
NIKA
Vol. 11, No
. 5, May 2013 : 2301 – 230
8
2304
0
0.
2
0.
4
0.
6
0.
8
1
0
50
100
150
200
250
300
350
400
450
500
ti
m
e
(
s
)
r,
y
1
(ra
d
/
s
)
Loop1
a)
h
=4ms
0
0.
2
0.
4
0.
6
0.
8
1
0
50
10
0
15
0
20
0
25
0
30
0
35
0
40
0
45
0
50
0
ti
m
e
(
s
)
r,
y
2
(ra
d /
s
)
l
oop2
b)
h
=5ms
0
0.
2
0.
4
0.
6
0.
8
1
-1
2
-1
0
-8
-6
-4
-2
0
2
x 1
0
4
ti
m
e
(
s
)
r,
y
3
(ra
d
/
s
)
Loop3
c)
h
=6m
s
Figure 4. Re
spon
se outp
u
t curve
s
of thre
e loop
s witho
u
t sch
edul
er
The outp
u
t of the 1
st
cont
ro
l loop an
d the
2
nd
control l
o
op ca
n qui
ckl
y
reach their
desi
r
ed
value. It means they tran
smit data thro
ugh net
work
i
n
time accord
ing to the respective sampl
i
ng
cy
cle.
B
u
t
t
h
e
3
rd
control lo
op ha
s to take a long time
waiting for d
a
ta tran
smission. It leads that
the feed
ba
ck sig
nal
ca
n n
o
t go
ba
ck controlle
r i
n
ti
me. The
cont
rolle
r
can
not
finish
calcul
ating
and adj
usting
the controll
e
d
plant. The resp
on
se outp
u
t of the 3
rd
control loop fails to reach the
expecte
d val
ue finally. Th
is
simulatio
n
prove
s
t
hat
it nece
s
sa
ry
for multi-l
oop
NCS
to ad
d
a
sched
uler.
4. Fuzzy
D
y
n
a
mic Sched
u
ling Algorithm of Multi-l
oop NCS
At prese
n
t, the resea
r
ch o
n
informatio
n sched
uling of
NCS is mo
stl
y
sche
duling
of open
loop. Based
on theo
ry of feedba
ck control,
con
s
i
derin
g the
sampling
pe
riod’s i
n
fluen
ce on
system p
e
rfo
r
mance, this p
aper
puts fo
rward a multi-l
oop fuzzy dy
namic
sche
d
u
ling alg
o
rith
ms,
whi
c
h in
cludi
ng calculatin
g prio
rity, determinin
g
the
prio
rity and
sched
uling ta
sks of n
e
two
r
k
node
s.
4.1. Principle of Fuzz
y
Dy
namic Scheduling
Princi
ple of fuzzy dynami
c
sche
dulin
g
is dynam
i
c
distrib
u
te the
priority of e
a
ch lo
op
according to
respon
se
error of ea
ch lo
op in mu
lti-l
o
op NCS. Co
nsid
erin
g rat
e
of error, two-
dimen
s
ion
a
l f
u
zzy controller i
s
used to
dete
r
mine
the p
a
ra
mete
r of
prio
rity. Figure 5
sho
w
s
schemati
c
of
fuzzy dyna
mical
sche
du
ling
alg
o
rithm
.
Acco
rdi
ng t
o
differe
nces
e
1
…e
n
of in
puts
r
1
…r
n
with ou
tputs
y
1
…y
n
a
nd
rates of
dif
f
eren
ce
ec
1
… e
c
n
, the
prio
rities of e
a
ch l
oop
p
1
…p
n
are
obtaine
d by f
u
zzy inferen
c
e [9]. The
sa
mpling
node
s
of NCS are
usu
a
lly drive
n
by cl
ock. O
n
ce
sched
uler i
s
use
d
, the nodes of NCS tran
smit t
heir
data or re
cei
v
e their data throug
h network
according
to arrang
ement of
sch
edul
er.
Figure 5. Sch
e
matic dia
g
ra
m of fuzzy dynamical sche
duling alg
o
rit
h
m
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
2302-4
046
Fuzzy Dynam
ic Sche
dulin
g
of Multi-loop
NCS (Fang
He)
2305
The
core of t
h
is
scheduling
algorithm i
s
to apply fuzzy control met
hod to
determine the
prio
rity of con
t
rol loop i
n
m
u
lti-loop
NCS
.
After
priority
is dete
r
min
e
d, the sche
du
ler de
cid
e
s th
e
orde
r
of adj
u
s
ting
different
loop
s
acco
rding to
t
he
p
r
iority of
co
ntrol lo
op. An
d
the
cont
rol l
oop
with the high
est prio
rity firstly posse
sse
s
prio
rity
of network data transmi
ssion to
finish its task.
4.2. Design
of Fuzz
y
Controller
The al
go
rithm
of fuzzy dyn
a
mic
sche
duli
ng fo
cu
se
s o
n
the
distri
but
ion of
prio
rity
of ea
ch
NCS
co
ntrol
l
oop [1
0]. Fig
u
re
6 i
s
sche
matic
diag
ra
m of fu
zzy i
n
feren
c
e
of p
r
i
o
rity. In loo
p
o
f
multi-loop NCS
r
i
is refe
rence input;
y
i
is a feedba
ck sig
nal of outputs;
e
i
is the differen
c
e
o
f
referen
c
e inp
u
t with outpu
t feedback of
loop;
ec
i
i
s
rate of error,
ec
i
(
k
)
= e
i
(
k
)
- e
i
(
k-
1)
;
p
i
is t
he
prio
rity of loop. The input
s
of fuzzy co
ntroller are
e
i
an
d
ec
i
. The out
put of fuzzy controlle
r is
p
i
.
It
is assum
ed that actual in
p
u
t range of
e
i
and
ec
i
i
s
[-1
,
1], and actual output ran
ge of
p
i
is
[1, 5].
The ratio fa
ct
or
K
e
and
K
ec
both equ
al 4. The qua
ntitative factor
K
q
equal 1. After fuzzifi
cation,
e
i
and
ec
i
is re
spectively tran
slated into
E
i
and
EC
i
.
The fuzzy discou
rse dom
ai
ns are define
d
as follo
w:
E
i
:
{-4, -3, -2, -1, 0, 1, 2, 3,
4}
EC
i
:
{-4, -3, -2, -1, 0, 1, 2,
3, 4}
P
i
:
{1, 2, 3, 4, 5}
and fuzzy set
s
are a
s
sign
e
d
as follo
w:
E
i
:
{ NB, NS,
ZE, PS,
PB}
EC
i
:
{ NB, NS, ZE, PS
, PB}
P
i
:
{PS,
S, M,
B, PB}
NB re
pre
s
e
n
ts ne
gative bi
g; NS rep
r
e
s
ent
s n
egative small; ZE
rep
r
e
s
ent
s zero; PS
rep
r
e
s
ent
s p
o
sitive small;
PB represent
s po
sitive
big
;
S represent
s small; M re
pre
s
ent
s mid
d
le;
B represents big. The m
e
mbe
r
ship fu
nction of tria
ngula
r
type is used for
E
and
EC
. The
membe
r
ship functio
n
of Ga
uss type is used for memb
ership
P
.
Fuzzy rule
s a
r
e
determi
ne
d ba
se
d o
n
t
he a
nalysi
s
t
o
the
relatio
n
s
hip
of p
e
rfo
r
mance of
control
syste
m
with it
s p
r
iority of task.
The
pr
in
cipl
e of setting f
u
zzy co
ntrol
rule
s i
s
that t
he
output of fu
zzy controller
can
truly
refl
ect the
stat
e
of loo
p
ta
sk in o
r
d
e
r to
rational
allo
cate
prio
rity of ea
ch control
loo
p
in
m
u
lti-loo
p
NCS [11].
Note th
at the
bigge
r the
value of
prio
rit
y
is,
the lower the
prio
rity of co
ntrol loo
p
is.
In other
wo
rd
, the smalle
r
the value of
prio
rity is, the
highe
r the
priority of cont
rol lo
op i
s
.
Fuzzy set PB co
rre
sp
on
ds to
a lo
w
prio
rity, and
PS
corre
s
p
ond
s t
o
a high
prio
ri
ty. Table 1 sh
ows the ru
le
s of fuzzy cont
rol to pri
o
rity
of each co
ntro
l
loop in multi-l
oop NCS.
Figure 6. Sch
e
matic dia
g
ra
m of fuzzy
inferenc
e of priority
Table 1. Rul
e
s of fuzzy con
t
rol
P
i
EC
i
NB NS
ZE
PS
PB
E
i
NB PS
PS
B
S
PS
NS
PS S
M
B PS
ZE
PS M PB M PS
PS
PS B
M
S PS
PB PS
S
S
PS PS
By the fuzzy inferen
c
e a
n
d
defuzzificatio
n
usin
g ce
nte
r
of gravity method, the tru
e
value
of the output finally
p
i
can b
e
got. Equation (1
) is the formul
a of cen
t
er of gravity method.
55
11
/
kk
k
kk
pp
p
p
(1)
Accordi
ng to
the pri
o
rity calc
ulated out
of control loop,
NC
S will
execute task of
control
loop in
seq
u
e
n
ce
acco
rdin
g to prio
rity level of
each l
o
op, and
reali
z
e the dynami
c
sch
edulin
g
of
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 23
02-4
046
TELKOM
NIKA
Vol. 11, No
. 5, May 2013 : 2301 – 230
8
2306
multi-loo
p
NCS. Becau
s
e
fuzzy dyna
mic
sched
ul
i
ng con
s
id
ers
error
an
d
e
rro
r ratio of each
control loop, the control loop whi
c
h pe
rforman
c
e
de
viates from set point worst is adjusted
in
time. So this kind of dyna
mic sche
dulin
g is hel
pful to
improve the
whol
e perfo
rmance of NCS.
4.3. Design
of Fuzz
y
D
y
namic Sche
duler
Based
on th
e multi-loo
p
NCS mo
del
as sho
w
n in
Figure 3, th
e fuzzy sch
e
duler i
s
adde
d into this simul
a
tion NCS mod
e
l. Once the
sch
edule
r
is add
ed, the sen
s
o
r
sen
d
s data
to
the net
work
a
c
cordi
ng to
sche
duling
of f
u
zzy sch
edul
er in
stea
d of
dire
ctly tran
smitting data t
o
the co
ntrolle
r throu
gh the
netwo
rk. F
u
zzy sche
duli
n
g
of multi-loo
p
s
is
used to
avoid conflict
of
data tran
smi
s
sion of differe
nt control lo
o
p
. Figur
e 7
shows the sim
u
lation struct
ure dia
g
ram of
multi-loo
p
fuzzy sch
eduli
ng. The out
put sign
al of three control loop
y
1
,
y
2
, and
y
3
ar
e
respe
c
tively conne
cted
to t
he in
put p
o
rts of fuzz
y
sche
duler. A
nd th
e outp
u
ts
of fuzzy controll
e
r
p
1
,
p
2
and
p
3
are al
so
re
spectively co
n
necte
d to the
input port
s
of the fuzzy
sched
uler. T
h
e
other pa
rt of dynamic
sche
duling
simulat
i
on model
of
multi-loo
p
NCS is same a
s
Figure 3.
-
K-
k9
-
K-
k8
-
K-
k7
-
K-
k6
-
K
-
k5
-
K-
k4
-
K-
k3
-
K
-
k2
-
K
-
k1
In
1
In
2
In
3
In
4
In
5
In
6
Ou
t
1
S
c
hed
ul
e1
Me
mo
r
y
2
Me
mo
r
y
1
Me
m
o
r
y
F
u
zz
y L
o
g
i
c
C
o
nt
r
o
l
l
er
6
F
u
zz
y L
o
g
i
c
C
o
nt
r
o
l
l
er
2
F
u
z
zy L
o
g
i
c
C
o
n
t
ro
l
l
e
r1
[y
3
]
[y
2
]
[y
1
]
[r
]
[y
2
]
[y
1
]
[y
3
]
[s
c
h
]
Figure 7. The
simulation
structu
r
e di
agram of multi-lo
op fuzzy sche
duling
In Figure 7, the modul
e of sch
edule
r
is
bu
ilt using T
r
ueTime Ke
rn
el module. By double
clickin
g
this
module of schedul
er withi
n
Matlab/si
m
u
link, the int
e
rnal
st
ru
cture of sch
edul
er is
sho
w
n a
s
Figure 8. Progra
m
of schedul
er mu
st
be written in form of M type file
into
sched
uler_ini
t file. Data of each
contro
l loop
ca
n be
transmitted t
h
rou
gh net
work
acco
rdin
g
to
the comm
and
from sched
ul
er_init file.
1
sc
h
A/
D
In
te
r
r
u
p
ts
Rc
v
D/
A
Sn
d
Sc
h
e
dul
e
Mo
n
i
t
o
r
s
P
T
r
ue
T
i
m
e
K
e
r
n
el
T
e
r
m
i
n
at
or
2
T
e
r
m
i
n
at
or
1
Te
r
m
i
n
a
t
o
r
Sc
o
p
e
G
r
ou
nd
1
G
r
o
und
6
p3
5
p2
4
p1
3
y3
2
y2
1
y1
Figure 8. The
internal st
ru
cture of sched
uler
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
2302-4
046
Fuzzy Dynam
ic Sche
dulin
g
of Multi-loop
NCS (Fang
He)
2307
5. Fuzzy
D
y
n
a
mic Sched
u
ling Algorithm of Multi-l
oop NCS
By building si
mulation mod
e
l of multi-loop
NCS with
fuzzy dynam
ic sched
ule
r
, system
respon
se
curves of th
re
e
control lo
op
can b
e
g
o
t a
s
sho
w
n
Figu
re
9.
In thi
s
si
mulation,
sa
me
PID co
ntrolle
r an
d same
model
of co
n
t
rolled
obje
c
t
with same t
r
a
n
sfer fun
c
tion
are
ado
pted
in
different
co
ntrol lo
op
s. An
d type of
net
work is
sele
cted a
s
Ro
und
Ro
bin. T
he
other pa
rame
ters
are: net
work transmi
ssion
rate
is 50
0kbit/s, the sa
mpling pe
rio
d
s of thre
e control loo
p
s
are
T
1
=4m
s
,
T
2
=5ms and
T
3
=7m
s
. The t
i
me interval
adju
s
ting t
he p
r
iority o
f
fuzzy dyna
mic
sched
uler i
s
T
p
=50m
s.
h
repre
s
e
n
ts the
samplin
g pe
riod of each cl
ose
d
loop
NCS.
Comp
ari
ng
with Figu
re
4,
we
ca
n l
earn
that
all
the
outputs of th
ree
cont
rol l
o
ops can
conve
r
ge to
their de
sire
d value in Figure 9. Th
e 3
rd
loop n
o
longe
r ha
ppen
s diverg
ent
phen
omen
on.
Thi
s
i
s
be
cause fu
zzy
sche
dule
r
rea
s
on
able
arra
ngeme
n
t
seq
uen
ce
of d
a
ta
transmissio
n of different co
ntrolle
r. The over
all pe
rformance of NCS was imp
r
ov
ed obviou
s
ly.
0
0.
2
0.
4
0.
6
0.
8
1
0
50
100
150
200
250
300
350
400
450
500
ti
m
e
(
s
)
r,
y
1
(r
ad /
s
)
Loop1
a)
h
=4ms
0
0.
2
0.
4
0.
6
0.
8
1
0
50
10
0
15
0
20
0
25
0
30
0
35
0
40
0
45
0
50
0
time
(
s
)
r,
y
2
(ra
d
/
s
)
L
oop
2
b)
h
=5ms
0
0.
2
0.
4
0.
6
0.
8
1
0
50
10
0
15
0
20
0
25
0
30
0
35
0
40
0
45
0
50
0
ti
m
e
(
s
)
r,
y
3
(r
a
d
/
s
)
L
oop
3
c)
h
=7m
s
Figure 9. Re
spon
se outp
u
t curve
s
of
thre
e loop
s with fuzzy sched
ul
er
From the ‘schedul
e’ outpu
t of
TrueTim
e Ker
nel in F
i
gure 8, we can ob
serve t
he time
seq
uen
ce di
agra
m
of NCS network node runni
ng. It includ
es three sta
t
use
s
: no d
a
ta
transmissio
n, waiting data
transmissio
n and sendi
ng
data.
W
h
en
sc
he
du
le
r
is
n
’
t us
ed
in
F
i
g
u
r
e
4
,
th
e
1
st
co
ntrol loop
and t
he 2
nd
control
loop a
r
e
usu
a
lly stay t
w
o
statuses:
no data
tr
an
smissi
on or se
nding data.
B
u
t
the
3
rd
control loop al
ways
stay waitin
g d
a
ta tran
smi
s
sion be
ca
use
of netwo
rk
co
nflict. It eventually lead
s th
at the re
sp
on
se
output of the 3
rd
control lo
o
p
is diverg
ent
.
After sch
edul
er is u
s
ed in
Figure 9, data trans
missio
n of network node
s of thre
e control
loop
s is re
a
s
on
ably arra
nged by the
fuzzy dyna
mic sche
dul
er. They all
usually stay
tw
o
statuses: no
data tran
sm
issi
on or
sen
d
ing data
in
time. This pro
c
e
ss g
r
e
a
tly improves
the
perfo
rman
ce
of each
control loop of NCS.
6. Conclusio
n
Summari
zin
g
the work ab
o
v
e, some con
c
lu
sion
s are given as follo
w:
1) An impo
rt
ant premi
s
e
of implementi
ng dy
nami
c
netwo
rk
sche
duling in a m
u
lti-loop
NCS i
s
that the sa
mpling
perio
d of ea
ch clo
s
ed
cont
rol loop i
n
multi-loop
NCS
shoul
d be in
a
certai
n ra
nge
whi
c
h is n
e
ith
e
r too big no
r too small.
2) T
he i
dea
of fuzzy dyna
mic
sched
uli
ng i
s
to
cal
c
ulate p
r
io
rity of ea
ch
clo
s
e
d
control
loop a
c
co
rdin
g to their error a
nd
rate o
f
erro
r n
e
two
r
k
paramete
r
s, and
to exe
c
ute
sched
uli
ng
according to
the value of priority. This i
dea ma
ke
those control loop whi
c
h output deviates
heavily from the expe
cted
value po
ssesse
s prio
rity of network tra
n
s
missio
n, and
be regul
ated
as
soo
n
as p
o
ssible. It ensures that ea
ch
clo
s
ed
c
ontro
l loop ca
n sat
i
sfy with the requireme
nts
o
f
steady an
d d
y
namic pe
rformance.
3) Th
e result of simul
a
tio
n
proves th
a
t
the perfo
rm
ance of e
a
ch
clo
s
ed
loop
cont
rol
system b
a
se
d on same
netwo
rk
platform is
im
pro
v
ed effectively by adding
fuzzy dyna
mic
scheduler in
multi-loop NCS.
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 23
02-4
046
TELKOM
NIKA
Vol. 11, No
. 5, May 2013 : 2301 – 230
8
2308
Ackn
o
w
l
e
dg
ement
This
wo
rk
was fina
ncially
sup
porte
d
by the Shan
dong P
r
ovin
cial Natural
Scien
c
e
Found
ation
(ZR20
12EEZ0
01), Ph
D Fo
undatio
n of
Un
iversity of Jina
n (No. X
BS1044), a
n
d
the
Grad
uate Ed
ucatio
n Innov
ation Proje
c
t of
Shandon
g Province (No. SDYC12
006
).
Referen
ces
[1]
Walsh GC, Octavian B, Bushnell
LG. As
y
m
ptotic be
hav
ior
of non
lin
ear
n
e
t
w
o
r
ke
d contr
o
l s
y
stems.
IEEE Transaction on Aut
o
m
a
tic Control
. 20
01
; 46(7): 109
3 -109
7.
[2]
Liu
CL,
La
yl
a
n
d
JW
. Sche
du
l
i
ng
al
gorit
hms
for mu
ltipro
gr
amming
i
n
a h
a
rd r
eal-tim
e
e
n
viro
nment.
Journ
a
l of the
ACM
. 1973; 2
0
(
1): 46-61.
[3]
Kevin J, Stanat
DF, Martel CU.
On Non-Pre
e
m
ptiv
e Sch
e
d
u
ling
of Peri
od
ic
and S
por
adic t
a
sks
. IEEE
Real-T
ime S
y
s
t
em S
y
mp
osiu
m. Heide
berg.
199
1: 10- 21.
[4]
Hon
g
SH, Kim
W
H
. Band
w
i
d
t
h alloc
a
tio
n
scheme i
n
CAN
protocol.
In P
r
oceedings of IEEE Control
T
heory an
d Ap
plicati
ons
. 2
0
0
0
; 147(1): 3
7
-4
4.
[5]
Otanez P, Mo
yn
e J, T
ilbur
y
D.
Usin
g D
ead
ban
ds to
Re
d
u
ce C
o
mmu
n
ic
ation
in
Netw
o
r
ked C
ontro
l
System
s
. In Procee
din
g
of the America
n
Co
ntro
l Co
nferen
ce. Anchora
ge.
2002: 6
15- 61
9.
[6]
W
a
lsh GC, Ye
H. Schedu
lin
g of Net
w
o
r
ke
d Contro
l S
y
ste
m
s.
IEEE Control System
s M
aga
z
i
ne
. 20
01
;
21(1): 57-
65.
[7]
W
i
dod
o RJ.
C
ontrol
S
y
stem
s in
Our D
a
il
y
Life.
T
E
LK
OMNIKA Indo
nesi
an J
our
na
l of E
l
ectrica
l
Engi
neer
in
g
. 2007; 5(1): 9-1
6
.
[8]
Emeka E, Jia
B, Derek R. M
ode
lin
g a
nd si
mulati
on too
l
for net
w
o
rked
control s
y
stem
s.
Sim
u
lati
on
Mode
lin
g Practice an
d T
heory
. 2012; 27(
9): 90–1
11.
[9]
Nasuti
on H, Ja
malu
ddi
n H, Syer
iff JM. Energ
y
an
al
ysis for
air con
d
itio
ni
n
g
s
y
stem
usin
g fuzz
y
log
i
c
control
l
er.
T
E
LKOMNIKA Indones
ian J
ourn
a
l of Electrica
l
Engi
neer
in
g
. 2011; 9(1): 1
39-
150.
[10]
Z
hang L, F
a
n
g
HJ. F
u
zz
y
c
ontrol
l
er des
ig
n for net
w
o
rk
e
d
control s
y
ste
m
w
i
th time-v
a
r
iant de
la
ys.
Journ
a
l of Systems E
ngi
ne
eri
ng an
d Electro
n
ics
. 200
6; 17(
1): 172-1
76.
[11]
Nasuti
on H, Ja
malu
ddi
n H, Syer
iff JM. Energ
y
an
al
ysis for
air con
d
itio
ni
n
g
s
y
stem
usin
g fuzz
y
log
i
c
control
l
er.
T
E
LKOMNIKA Indones
ian J
ourn
a
l of Electrica
l
Engi
neer
in
g
. 2011; 9(1): 1
39-
150.
Evaluation Warning : The document was created with Spire.PDF for Python.