TELKOM
NIKA
, Vol.14, No
.4, Dece
mbe
r
2016, pp. 14
08~141
6
ISSN: 1693-6
930,
accredited
A
by DIKTI, De
cree No: 58/DIK
T
I/Kep/2013
DOI
:
10.12928/TELKOMNIKA.v14i4.4997
1408
Re
cei
v
ed Se
ptem
ber 21, 2016; Revi
se
d No
vem
ber
8, 2016; Acce
pted No
vem
b
er 23, 201
6
Planning and Coordination in Hierarchie of Intelligent
Dynamic Systems
Alexan
der Y
a
. Fridman
Institute for Informatics an
d Mathematic
al
Mode
lli
ng,
Kola Sci
enc
e Centre of the R
u
ssia
n
Acad
e
m
y
of Scie
nces
,
24A F
e
rsman s
t
r., 184209 Ap
atit
y
Murma
nsk
reg., Russia
T
e
l./fax +
7
815
55 74
05
0, e-mail: fridma
n@ii
mm.ru
A
b
st
r
a
ct
On the bas
is o
f
the know
n pr
incip
l
e
of inter
a
ct
ions
pred
icti
on (Mes
a
rovic)
, our earl
i
er
propos
e
d
increm
ental c
o
ordination pr
inciple is
extended ov
er hier
archical collec
tiv
e
s of intelligent dynam
i
c
systems
(IDSs) after Gennady Osipov. Such
systems adm
i
t
arbitrary types of va
r
i
ables in their
state vector and
thereby
all
o
w
investig
atin
g more ge
ner
al dy
na
mic syste
m
s
than “class
ical
” o
nes d
e
fin
ed
in nu
merica
l stat
e
spaces. Us
ing
the conc
ept of
effectiv
e N-atta
ina
b
il
ity (Osipo
v), a straightfor
w
ard proce
dur
e of pl
an
nin
g
fo
r
hier
archic
al co
l
l
ectives of IDS
is deve
l
o
ped.
As soon
as a
pla
n
for reac
hi
ng a
goa
l state
from the c
u
rre
nt
one is fo
un
d, e
ffective imple
m
entatio
n of
this
plan r
equ
ires f
o
r coord
i
n
a
tion
of IDSs taking
their parts in t
h
e
collectiv
e. W
e
consi
der b
o
th aspects of coo
r
din
a
tion
(co
o
r
d
in
abi
lity w
i
th respect to the c
oord
i
nator
’
s
ta
sk
and c
oor
din
abi
lity in r
e
lati
on t
o
the g
l
o
bal ta
sk) and
in
fer
n
e
cessary c
ond
i
t
ions of th
e co
ordi
nab
ility for
a
local
l
y org
ani
zed hi
erarchy of
IDSs.
Ke
y
w
ords
:
intelligent dynamic system
,
increm
ent
al coordi
nation, direct
planning, loca
lly or
gani
z
e
d
hier
archy
Copy
right
©
2016 Un
ive
r
sita
s Ah
mad
Dah
l
an
. All rig
h
t
s r
ese
rved
.
1. Introduc
tion
Hierarchi
c
al
control
be
cam
e
sta
nda
rd fo
r comp
l
e
x systems d
ue to
i
n
crea
sing
difficulty of
centralized
manag
eme
n
t for su
ch sy
stem
s. It wa
s ne
ce
ssary
to divide the deci
s
ion
-
ma
king
pro
c
e
ss i
n
to several levels to get rea
s
o
nable
co
m
p
le
xity of optimization ta
sks
on ea
ch of th
em.
Ho
wever,
ad
vent of multil
evel
hie
r
arch
ical
system
s rai
s
ed
a n
e
w
p
r
obl
em o
f
matchin
g
a
nd
coo
r
din
a
ting
the de
cisio
n
s made o
n
different cont
rol levels (see,
for insta
n
ce, [1-3]). The
key
probl
em
s in
the d
e
velopm
ent of
su
ch
system
s
ar
e:
sp
eci
a
lizatio
n of
su
bsy
s
tems withi
n
t
heir
inherent pro
b
l
ems
an
d co
ordin
a
tion of
co
ntrol
imp
a
c
ts
at diffe
re
nt levels of t
he hi
erarchy. In
other word
s,
the tasks fo
r
sub
-
sy
stem
s
and th
eir q
u
a
lity criteri
a
are ne
ce
ssa
r
y
to form
such
a
way that
the
sha
r
ed
pe
rformance
o
f
th
eir
tas
k
s
a
llows
s
u
bs
ys
tems to
pe
r
f
or
m a g
l
o
b
a
l
tas
k
fo
r
entire hie
r
a
r
chy (com
patibi
lity postulate [4]).
In gene
ral, th
e con
s
tru
c
tio
n
of the
co
ordinati
on pri
n
ciples (in pa
rticula
r
,
the
i
n
te
ractio
ns
predi
ction
p
r
i
n
cipl
e [4])
re
quire
s fo
r
se
ekin
g
sati
sfa
c
tory
solutio
n
s
at the l
e
vel of the l
o
wer
deci
s
ive elem
ents, whi
c
h is con
s
iste
nt with the
modern method
s of decentra
li
ze
d
control. Befo
re
this
, it is
necess
ary to solve three problems
: fi
rst, t
o
build
a m
e
tric in
the
sta
t
e sp
ace of t
h
e
system;
se
co
nd, to spe
c
ify a coo
r
dinati
on p
r
in
cipl
e, and
thi
r
d,
to plan com
b
ine
activities of
the
lowe
r-l
evel el
ements for reaching
the
gene
ral
goal
of the
whole
system.
Let
us con
s
ide
r
the
existing ways of solving these p
r
obl
em
s.
To
solve th
e
first
of the
mentione
d p
r
oblem
s, the
r
e a
r
e
kn
own
metho
d
s to
metri
z
e
spa
c
e
s
of
dat
a and
kno
w
le
dge in
order to build
t
heir
h
i
era
r
chical ta
xonomie
s (e.g., [5,
6]). Such
metrics are base
d
on plant
characte
risti
c
tables
of the taxonomy elements
that are not suita
b
l
e
for dynami
c
system
s. In e
c
on
omic
appl
ication
s
of hi
era
r
chical sy
stem
s (fo
r
ex
ample, [7]), t
hey
only u
s
e fina
ncial
indi
cato
rs. T
h
is limits the
gen
eral
ity of estimat
i
on a
nd
cont
rol for th
e
sta
t
e
element
s cha
r
acte
ri
zed by
different unit
s
of me
a
s
u
r
e
.
Methods of
expert jud
g
m
ent pro
c
e
s
sin
g
like [8] are ne
ither focu
se
d on the dynam
ic co
ntrol p
r
o
b
lems.
As for the se
con
d
pro
b
l
em, coo
r
din
a
tion is mo
stly investig
ated with re
gard to
informatio
n coordi
nation [9
,
10], behaviors
coo
r
din
a
ti
o
n
[11] or coo
r
dination of pe
er-ran
k obj
ect
s
[12,
13], whil
e mo
st
com
p
lex sy
stem
s are hi
era
r
chical
an
d n
e
ed dyn
a
mic
coo
r
din
a
tion
of
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
1693-6
930
Planning
a
nd Coo
r
din
a
tion in
Hierarchie of
Intelligent Dynam
ic System
s (Alexa
nd
er Ya)
1409
intera
ction
s
a
m
ong sub
s
ystems of different levels in d
e
ci
sion ma
kin
g
.
In the literature, the third probl
em (pl
a
nni
ng
) co
nce
r
ns me
ch
ani
cal aspe
cts of
robots
functioni
ng [14-1
6
] and
d
oes
not co
nsider info
rm
ati
onal inte
rlin
ks amo
ng
sub
s
ystem
s
, whi
c
h
inclu
de their
own d
e
ci
sion
make
rs.
In co
nne
ction
with th
e a
b
o
v
e-de
scrib
ed,
a
ch
a
nge
-ba
s
ed
(g
ra
dient
for
contin
uou
s
states
and in
creme
n
tal – for di
screte
state
s
)
gene
rali
zed
criterio
n wa
s
prop
osed in [
17] for the
state
estimation
in hiera
r
chi
c
al
d
y
namic syste
m
s
that
allows to a
nalyze
nume
r
ical sta
t
e eleme
n
ts.
In
terms of mult
i-obje
c
tive op
timization (e.
g
., [18]), th
is
crite
r
ion bel
o
ngs to the gl
obal criteria
with
weig
ht coeffi
cient
s inversely prop
ortio
nal to the
tol
e
ran
c
e
s
of
scala
r
criteri
a
. This id
ea lo
oks
rea
s
on
able
si
nce the mo
re
important is
a crite
r
i
on for the whole sy
stem, the less its devation
s
are ad
missibl
e
from the Co
ordin
a
tor’
s po
int of view.
On this
basi
s
, we have d
e
v
eloped a
co
ordin
a
tion
p
r
i
n
cipl
e for hi
erarchical sy
ste
m
s [17]
that impleme
n
ts the intera
ction
s
pre
d
ict
i
on prin
ci
pl
e [4] for eleme
n
ts of a hierarchical syst
em,
taking into
accou
n
t the me
thod of en
suring stability
o
f
local control
signal
s in th
e colle
ctives
of
automata [19
]. Our coo
r
di
nation techni
que u
s
e
s
the necessa
ry and sufficie
n
t condition
s of
coo
r
din
ability for a locally o
r
gani
ze
d
hierarchy of dynamic sy
stems.
In terms of system an
alysis, the p
r
op
o
s
ed
pr
in
cipl
e to
coo
r
di
nate
hierarchi
c
al system
s
corre
s
p
ond
s to the externa
l
(obje
c
tive) a
ppro
a
ch
to assessin
g the effectivene
ss of subsyste
ms
within a m
e
tasystem. Thi
s
principle
stat
es as
follows:
sub-obj
ect
s
tasks
will be
coordi
nated
wi
th
respec
t to the Coordinator’s
task
, if the s
i
gn
of the
gradie
n
t of the gen
erali
z
ed Co
ordi
nat
or’s
crite
r
ion
for i
t
s current
do
minant
scala
r
criteri
on
wil
l
coi
n
cid
e
with si
gn
s of g
r
adient
s of thi
s
gene
rali
zed
criterio
n for
all
cu
rre
nt valu
es of
scal
a
r
crite
r
ia fo
r su
b-obj
ect
s
. Efficien
cy of thi
s
techni
que
wa
s illustrated b
y
its simulatio
n
for a netwo
rk obje
c
t [17].
Thus, the ea
rlier develo
p
e
d
coo
r
din
a
tio
n
prin
ci
ple fits only to hierarchical syst
ems with
quantitative q
uality crite
r
ia
and n
u
me
rical state
sp
a
c
es. In this
p
aper,
we ext
end the i
dea
of
increme
n
tal
coo
r
din
a
tion
to the sy
ste
m
s a
d
mitti
ng
other type
s of variabl
es as
well. T
he
approa
ch
we
prop
ose is implemente
d
belo
w
fo
r intelligent d
y
namic
syst
ems (I
DSs)
[20]
desi
gne
d for mod
e
ling
co
mplex dyna
m
i
c
system
s i
n
the state sp
ace
s
with arbitrary
type
s of
state eleme
n
ts.
To illustrat
e
our ide
a
of coordi
nation,
we
intro
d
u
c
e
IDSs para
d
i
h
m [21] first, and then
descri
be me
a
n
s for pl
anni
n
g
and
coo
r
di
nation in this
formali
s
m si
n
c
e existin
g
a
plan to re
ach
a
goal state i
s
a necessa
ry con
d
ition for
coo
r
din
a
ting.
2. Sy
nopsis
of IDS [13, 2
0
, 21]
An IDS is describ
ed a
s
a di
screte dyna
m
i
c syste
m
.
D = < X
,
N,
,
>,
(1)
whe
r
e:
X
is
a
topologi
cal
state spa
c
e
x
X
with the
proximity rela
tion
;
N
is
the s
e
t of natural
numbe
rs, whi
c
h ma
rk di
screte points of time;
is the set of all sub
s
ets of
X
;
:
is a clo
s
ure functi
on with the fo
llowing
pro
p
e
r
ty:
if
х
, then
х
(
х
);
(2)
:
N
is a tran
sitio
n
functio
n
wit
h
the
p
r
op
erti
es co
rre
sp
on
ding
to req
u
irements
for tran
sition functio
n
s in th
e “cla
ssical”
control theo
ry:
(
x,
0) =
x
for any
x
,
(
(
x,
t
1
),
t
2
) =
(
x,
t
1
+ t
2
).
(3)
If an IDS is
based o
n
rul
e
s that
co
ntain se
ts
of formul
as
of a
ce
rtain la
ng
uage
L,
dynamics of this IDS in ma
rkovia
n ca
se
i
s
de
scribe
d b
y
the followin
g
equatio
n:
x
(
t
+ 1
)
=
(
(
(
x(
t
)
u
(
t
)
(
t
))),
(4)
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 16
93-6
930
TELKOM
NIKA
Vol. 14, No. 4, Dece
mb
er 201
6 : 1408 – 141
6
1410
Whe
r
e:
u
(
t
)
U
(
t
)
U
L
is
a
set of
facts that a
r
e ad
ded to
the
state
x
(
t
) (co
n
trol
si
gn
als);
(
t
)
(
t
)
L
is a se
t of facts tha
t
appea
r a
s
a re
sult of u
npre
d
icta
ble
cha
nge
s in I
D
S’s
environ
ment (disturban
ce
s).
The traje
c
tory (4) is stabl
e, if the
function
is mo
notoni
c and
is monoto
n
ic with
respec
t to the s
t
ate vec
t
or (4) [21]. In
what
follo
ws
we a
s
sum
e
that any referenced IDS
m
eets
these con
d
itions.
2.1. IDSs Architecture
IDSs allo
w for kn
owl
edg
e rep
r
e
s
entat
ion bot
h by rules a
nd se
mantic net
wo
rks [13].
Furthe
r on, we will con
s
i
der the rul
e
-based ID
Ss.
In such an IDS, its kno
w
ledge ba
se (KB)
comp
ri
se
s a set of rule
s of
the following
format:
D
,
A
,
C
,
(5)
Whe
r
e:
C
is
a pre
c
on
ditio
n
(co
ndition
)
of a rule;
A
is a set of facts a
dde
d after appli
c
atio
n of the rule
;
D
is a set of facts d
e
leted
after appli
c
ati
on of the rule
.
C, A
и
D
are sets of form
ul
as of the lang
uage
L
.
Any rule must
meet the rela
tion.
A
D =
.
(6)
Any rule belo
ngs to only o
ne of the two cla
s
ses: RD
or RS.
Every rule of
the RD cl
ass
contai
ns
an a
c
tion
a
pplied
to the extern
al enviro
n
me
nt by an
executive bo
dy or a pro
c
edure t
hat compute
s
and
assi
gn
s a variabl
e with
certai
n value
s
of
some
data
b
a
s
e
attribute
s
con
s
id
erin
g t
heir val
u
e
s
a
v
ailable in
th
e current
stat
e. The
s
e
acti
ons
result in
ch
a
nge
s of th
e I
D
S’s
datab
ase state.
T
h
is
gro
up of
rul
e
s de
scribe
s
ch
ang
es of the
system’
s
stat
e in time and is call
ed “rule
s
of (dia
ch
ron
i
c) tra
n
sitio
n
”.
Rule
s of the RS cla
ss a
r
e
bind with no
ac
tion
s, they do not chan
ge the enviro
n
ment;
rathe
r
, they
chang
e the
kn
owle
dge
of it. In othe
r
wo
rds, they
rep
r
ese
n
t the th
e
o
ry of the
su
bject
domain.
Then the IDS’
s kn
owl
edg
e base is:
R = <
RS
,
RD
>
.
(7)
2.2. IDSs Goal-Seeking
Behav
i
our
As noted a
b
o
v
e, an IDS ca
n be de
scrib
e
d
by the relati
ons
(1)
– (4
). To com
p
ly with the
prop
erty (2
), no RS-cla
ss rules a
r
e supp
ose
d
to be ca
pable of re
mo
ving facts [13
]
.
Then
(
,
1) is the
tra
n
si
tion fun
c
tion,
and
{
(
(
x,
i
))
i
N
} de
scribe
s an o
r
bit
o
r
a
trajecto
ry of the dynami
c
system.
Dynami
cs of
a rule
-ba
s
e
d
IDS (in the Ma
rkov
ca
se) i
s
descri
bed by
equatio
n (4
).
Let u
s
con
s
id
er the
relatio
n
shi
p
b
e
twee
n archite
c
tu
re
of the
kn
owl
edge
ba
se
R
(
7
)
a
nd
prop
ertie
s
of the model
(1)
– (4) [20].
Suppo
se,
L
(
R
) is a set of formula
s
from t
he lang
uage
L
, which occu
r in the rule
s
of
R
.
Defini
tion 1
.
If
X
is a
set
of IDS’s stat
es, the
n
the
pair
of poi
nts (
x
0
,
x
1
) in
the
sp
ac
e
X ×
X
is called
N
-attainabl
e one, if the
r
e exist
control sig
nal
s
U
(
j
) (
j
= 0, 1,…,
N
-
1), for whic
h
x
1
x
(
N
) with
the initial conditions
x
(0)
x
0
U
(0), w
h
e
r
e
x
(
t
),
t
N
are
solution
s
of the IDS’
s
state equ
atio
n (4).
Defini
tion 2
.
If a pair of p
o
ints (
x
0
,
x
1
) i
s
N
-attaina
b
l
e
and eve
r
y fact of
x
(
N
) d
oes
not
occur i
n
more
than one
rul
e
within the
corres
pon
ding
trajecto
ry, then the pai
r of
points (
x
0
,
x
1
) is
called effectively
N
-attainable.
Let
L
(
R
)
be a set of facts. If a sequ
ence of rule
s
П
1
,
П
2
,
…,
П
k
from
R
is gi
ven, the
set
of
f
a
ct
s S
(
П
1
,
П
2
,
…,
П
k
) derive
d
after ap
plicatio
n
of
rule
s fro
m
this se
que
nce i
s
define
d
by
indu
ction:
S
(
П
1
) =
\
D
1
A
1
;
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
1693-6
930
Planning
a
nd Coo
r
din
a
tion in
Hierarchie of
Intelligent Dynam
ic System
s (Alexa
nd
er Ya)
1411
S
(
П
1
,
П
2
,…,
П
i
) =
S
(
П
1
,
П
2
,…,
П
i
-1
) \
D
(
П
i
)
A
(
П
i
).
(8)
Defini
tion 3
. A rule
П
i
is ca
lled admi
ssi
bl
e one, if there
is a cont
rol
U
i
-1
, for whic
h:
C
i
(
S
(
П
1
,
П
2
,…,
П
i
-1
)
U
i
-1
.)
(9)
Defini
tion 4
. The seque
nce of rules
and
controls
П
=
<(
П
1
,
U
1
), (
П
2
,
U
2
), …, (
П
k
,
U
k
)>
is
calle
d a plan
to achieve th
e state
from the current
state
, if:
1) ea
ch rule from this
seq
u
ence is admi
s
sible;
2)
S
(
П
1
,
П
2
, …,
П
k
).
Theorem
[2
0
]
. Fo
r eve
r
y p
a
ir
of poi
nts
(
x
0
,
x
1
)
X ×
X
, the pla
n
П
= <(
П
1
,
U
1
),
(
П
2
,
U
2
),
…, (
П
k
,
U
k
)>
exists, if and only if the pair (
x
0
,
x
1
) i
s
N
-attainable.
In [13], an algorithm is p
r
opo
sed to se
arch for a se
quen
ce of ad
missi
ble rul
e
s and their
relevant controls that make
up the plan to achi
eve the state
from the curre
n
t state
.
A
ccor
d
i
ng
to the prin
cipl
es of dyn
a
mi
c p
r
og
rammi
n
g
, this al
go
r
i
th
m w
o
rk
s
"bac
kw
ar
d
in
time
" (
s
tar
t
in
g
fro
m
the target
sta
t
e). Ho
wever,
this app
roa
c
h is difficult t
o
con
s
id
er th
e rule
s of the
RD
cla
ss
sin
c
e
they not
alwa
ys have
a
n
in
verse
o
perato
r
. In
co
n
n
e
c
tion with
th
e written
a
bove, we no
w
p
r
o
p
o
se
a planni
ng proce
dure that works in "live time
" starting
from the initia
l IDS’s state.
2.3. Direct Pl
anning for I
D
Ss
The a
bove-cited theo
rem
determi
ne
s the ne
ce
ssary an
d sufficient
con
d
itions fo
r
existen
c
e a p
l
an to tran
sfe
r
a
sy
stem from an initial
state
x
0
to th
e end
state
x
1
. If we toughen
con
d
ition
s
of this theore
m
and re
quire for effective
N
-attain
abil
i
ty of a pair
of points (
x
0
,
x
1
)
rathe
r
than their
N
-attain
ability (see
Definition 2
)
, we ca
n obtain a dire
ct planni
ng alg
o
rith
m
simila
r to th
e
idea
s
of de
ri
vative-based
control
im
ple
m
ented i
n
th
e "cla
ssical"
automatic co
ntrol
theory.
Let
x
(
t
+
1
)
be an IDS’s st
ate vector ob
tained by
sol
v
ing the equ
ation (2) at the step
t + 1.
Then
x
(
t
+
1) \
x
(
t
) a
r
e the ne
w fa
cts that have
appe
are
d
on
this sta
ge of i
n
feren
c
e. If the
IDS is on a trajecto
ry, which has
the pro
perty of the effective
N
-attainability, then the emergi
ng
facts shoul
d not repe
at all along thi
s
traj
ectory
. Conseque
ntly, the
followin
g
relat
i
on is true:
1
1
︶
︵
\
1
︶
︵
N
t
t
x
t
x
.
(10)
Then for
synthesi
s
a pla
n
by dire
ct infe
rence, it is possi
ble
to cal
c
ulate the inte
rse
c
tion
of already ap
peared ne
w facts at ea
ch
step
k
:
1
0
︶
︵
\
1
︶
︵
k
t
t
x
t
x
k
(11)
And cho
o
se the cu
rrent co
ntrol so a
s
:
k
k
x
k
x
))
(
1)
(
(
\
.
(12)
Hence, the procedure of
the direct planning will look as follows:
1) Let
x
1
is a t
a
rget st
ate an
d
x
0
is the initial state of an
IDS.
2)
k
:=
0,
k
:
.
3) Let
x
(
k
) b
e
the
cu
rrent
state. If
x
1
x
(
k
), then s
t
op.
Als
o
, let
П
k
:= {
П
1
k
,
П
2
k
, …,
П
l
k
} i
s
the set of ad
missi
ble rule
s at the step
k
.
4) Ap
ply a rul
e
from
П
k
an
d
ch
eck the
co
ndition
(12
)
. If it hold
s
, then
k
:=
k
+
1 a
n
d
go to
Step 3, otherwise sel
e
ct a
nother
(not a
pplied yet) ru
le from
П
k
an
d return to the begin
n
ing
o
f
Step 4. If none of the admissi
ble
rul
e
s
result in ful
f
illing the
condition (12), then the step
k
is
con
s
id
ere
d
a
failure; the
rul
e
led to it is
marked
as a
dead
-en
d
; ba
ckt
ra
ck to the
previou
s
ste
p
of
inference;
k
:=
k
- 1 a
nd g
o
to Step 3.
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 16
93-6
930
TELKOM
NIKA
Vol. 14, No. 4, Dece
mb
er 201
6 : 1408 – 141
6
1412
It is possible t
o
sho
w
that, if there exist a
n
effective (in
the sen
s
e of
Definition 2
)
plan to
achi
eve the state
x
1
from the state
x
0
, then the d
e
scribed
pro
c
e
d
u
re
will co
mp
lete con
s
truct
i
on
of this plan. The sp
eed of
implementin
g the plan wi
l
l
incre
a
se, if the algo
rithm will look th
ro
ugh
the entire
se
t of admissi
b
l
e rule
s at Step 4 and,
if there a
r
e several rule
s that satisfy the
con
d
ition (1
2), choo
se on
e of them,
П
*k
, for whi
c
h the f
o
llowin
g
rel
a
tion is true:
x
1
\
x*
(
k
)
x
1
\
x
i
(
k
),
(13)
Whe
r
e
x*
(
k
) i
s
the state af
ter appli
c
atio
n the rule
П
*k
and
x
i
(
k
) are
the states af
ter appli
c
atio
n of
any other rule
s meeting the
conditio
n
(12
)
.
After
obtaini
ng
a plan, we have
to proceed with coordination duri
n
g
fulfilling
thi
s
plan.
In
[13], the p
o
ssibility to
con
t
rol inte
ra
ctio
ns
within
a
te
am of "
pee
r"
IDSs
(the
on
es
with
identi
c
al
kno
w
le
dge b
a
se
s) i
s
de
scribed. An exa
m
ple of su
ch
a gro
up can
serve vehi
cle
s
involved in ro
ad
traffic. We b
e
lieve, it is of interest to invest
igate hi
era
r
chical system
s, which
include IDS
s
as
element
s of different levels. This will be
done in the n
e
xt section.
3. Coordina
tion in a Collectiv
e of IDSs
As in [4]
and
without l
o
ss
of gen
erality,
we
co
nsid
er
a two
-
level I
D
Ss
syste
m
(Figure
1)
whe
r
e
the to
p-level
de
sisi
on m
a
ker (Coordi
nato
r
) DM
0
se
nd
s
co
ordin
a
ting
si
gnal
s
(adju
s
ti
ng
para
m
eters o
f
the quality criter
ia
of the l
o
we
r-l
evel IDSs)
i
to the
subordinate
d
IDSs
DM
1
– DM
n
and re
ceive
s
their feedb
ack sig
nal
s
w
i
. For si
mplicity, we assume that
all IDSs i
n
the lower level
are of the
sa
me type, i.e.
have
the sa
me KBs (7
). Such a tea
m
may take pa
rt, for example, in
the solutio
n
of any con
s
truction ta
sk b
y
a collect
ive
of robots, o
ne of whi
c
h
serve
s
a
s
a j
o
b
coo
r
din
a
tor. Subordinate
d
IDSs
inte
ra
ct only via
the co
ntrolle
d process
P
and h
a
ve
no
informatio
n concerni
ng th
e states
of other IDS
s
in
the sa
me leve
l, that is, the entire
system
is
locally organi
zed.
Acco
rdi
ng to the prin
ciple
s
of decentrali
zed
control, the purpo
se o
f
coordi
nating
signal
s
i
, co
ming
from the
Coordinator
to
sy
stem
s of the lower level, is
a
specifi
c
ation of
such
con
d
ition
s
for the tasks they have to manag
e t
hat they would i
s
sue pro
p
e
r
co
ntrol sig
nal
s
m
i
,
whi
c
h ensure fulfillment of both their own
goals and the
task
of the whol
e sy
stem.
Corre
s
p
ondin
g
ly, there e
x
ist two con
c
ept
s of co
ordin
ability for the lo
we
r-level sy
ste
m
s:
coordinability with
respect
to t
he
task of
the Coordi
nator and
co
ordinability in relation to t
he
global
task. T
here
a
r
e
sev
e
ral
mod
e
s o
f
coo
r
di
nation
[4], the m
e
th
od of
interact
ions p
r
edi
ctio
n
looks the
mo
st suitable fo
r a
colle
ctive
of IDSs
. In i
m
pleme
n
ting
this meth
od,
the Co
ordinat
or
inform
s the subordinate
sy
stem
s about i
t
s desi
r
ed val
ues of interactions amo
ng them and e
a
ch
of the lower-l
e
vel system
s is trying to
reac
h the
corresp
ondi
ng p
r
edetermine
d
value, assum
i
ng
that the other subsystem
s
w
ill operate properly as
well.
Figure 1. A two-l
e
vel syst
em of deci
s
io
n-ma
kin
g
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
1693-6
930
Planning
a
nd Coo
r
din
a
tion in
Hierarchie of
Intelligent Dynam
ic System
s (Alexa
nd
er Ya)
1413
3.1. Coordinabilit
y
w
i
th
Resp
ect to the Coordinator’s Task
Whe
n
inte
rpreting the
met
hod of i
n
tera
ction
s
p
r
edi
ct
ion for the t
w
o-l
e
vel
coll
ective of
IDSs, we
assume th
at at the begin
n
i
ng of every
stage
of the
operation
t
j
, the Coo
r
di
n
a
tor
inform
s DM
i
about their d
e
sired state
s
x
i
to be achi
eved at this stage. Thu
s
,
we shoul
d take
i
=
x
i
in Fi
gu
re
1. As fee
d
back
signal
s
from the
lo
wer-l
evel IDSs, the Coo
r
din
a
tor u
s
e
s
the
i
r
states
x
i
a
s
t
he m
o
st
co
mpre
hen
sive
set
s
of lo
ca
l inform
ation.
The
n
n
a
tura
l formul
ation
for
coo
r
din
ability conditio
n
s
with respect to
the Coo
r
din
a
tor’s ta
sk re
quire
s that e
a
ch
sub
o
rdi
n
ated
IDS can
achi
eve the
de
sired
st
ate from
its
cu
rrent state
x
i
.
Obviou
sly, this re
qui
res for exi
s
te
nce
a plan to a
c
hieve
x
i
(
t
j
) fr
om
x
i
(
t
j
) [13], i.e. existence a se
que
nce of cont
rol
impact
s
on t
he
environ
ment, whi
c
h would
allow to
com
e
clo
s
e
r
to the desi
r
e
d
stat
e (in g
ene
ral,
in the pre
s
e
n
ce
of distu
r
b
a
n
c
es).
It is also
cle
a
r that, g
enerally spea
king,
every
DM
i
may n
eed
differe
nt tim
e
to
achi
eve a given state; t
herefore it is po
ssible to propo
se tw
o ap
pro
a
ch
es to the orga
nization
of
interac
t
ions
among s
u
bsys
tems
in
time.
Either
th
e Coo
r
dinato
r
shall
ge
nerate the
spe
c
ified
states, taki
ng
into acco
unt the potential of all lo
wer-le
vel IDSs to achieve them i
n
a single cy
cle
of the system
, then you ca
n syn
c
hr
oni
ze the internal
time of the
IDSs, or the I
D
Ss sh
all have
the
event-d
riven
planni
ng. Preferen
ce fo
r one of
the
above a
pproache
s is
d
e
termin
ed by
the
spe
c
ificity of the subj
ect a
r
ea. Fo
r sim
p
licity, it
is further a
s
sume
d
that the Coo
r
dinato
r
’
s
time
and lo
cal tim
e
s of all
DM
i
are
synchro
n
ize
d
, and it
s in
cre
m
ent i
s
a
c
cepted e
qual to 1. Th
us,
t
T
=
{0, 1,
2,…}.
Und
e
r coo
r
di
nation
by
the
method of
intera
ctio
n
s
p
r
e
d
iction, di
stu
r
ban
ce
s will
o
c
cur i
n
a
sy
st
em DM
i
,
if the states
of the lower-l
e
vel system
s are diffe
rent
from the o
n
e
s
set by th
e
Coo
r
din
a
tor.
More d
e
tail d
e
scriptio
n of the distu
r
ban
ce
s is only p
o
ssible after
a more
spe
c
i
f
ic
descri
p
tion
of the p
r
obl
em
to be
solve
d
by a
co
ll
ective of IDSs
and th
e envi
r
onment; that
is
beyond the p
u
rpo
s
e of the
given paper.
Howeve
r, in
view of (4), we can interp
ret a necessary
coo
r
din
ability con
d
ition
wi
th re
spe
c
t to
the Co
ordi
n
a
tor’s t
a
sk a
s
a
requi
rem
ent to move
"as
clo
s
e a
s
po
ssible" to end st
ates for all do
minat
ed IDS
s
by the end of the current control ste
p
:
i
(
t
)
(
t
),
x
i
(
t
+ 1
)
X
,
i
I
;
u
i
*
(
t
)
U
(
t
):
x
i
*(
t
+
1) \
x
i
(
t
+ 1)
x
i
(
t
+
1) \
x
i
(
t
+
1),
(14)
Whe
r
e:
x
i
*(
t
+ 1)
=
(
(
(
x
i
(t
)
u
i
*(
t
)
i
(
t
))) is th
e be
st po
ssi
ble
state for the
DM
i
by the end
of
a control step
;
x
i
(
t
+ 1
)
=
(
(
(
x
i
(t
)
u
i
(
t
)
i
(
t
))) i
s
a
n
y other
attainable
state
for the
DM
i
by
this
instant.
In [13], it is shown that for stabili
zation o
f
a
trajectory of an IDS, i.e. for comp
ensation of
disturban
ce
s
influen
ce, it is enoug
h to ap
ply the followi
ng co
ntrol:
u
(
x
(
t
+ 1
)
,
t
+ 1))
=
(
(
x
(
t
+ 1)
),
t
+ 1)) \
(
(
(
(
(
x
(
t
)
δ
(
t
)),
t
)),
t
+
1)).
(15)
Considering (4), the relati
on (15) will look like:
u
(
x
(
t
+ 1
)
,
t
+ 1))
=
x
(
t
+ 2 /
t
+ 1)) \
x
(
t
+ 2
/
t
).
(16)
Whe
r
e:
x
(
t
+ 2 /
t
+
1) is the predi
ction f
o
r the value o
f
the state
x
at the instant t
+
2 made at t
h
e
time t
+
1 in
absen
ce
of di
sturb
a
n
c
e
s
;
x
(
t
+ 2 /
t
) is th
e p
r
edi
ction
for th
e valu
e
of the
state
x
at
the insta
n
t t
+
2 mad
e
at
the time
t
co
nsid
erin
g di
sturba
nces,
wh
ich exi
s
ted at
that point, a
nd
assumin
g
ab
sen
c
e of di
sturba
nces at the time t + 1.
Formul
as (15
)
, (1
6) allo
w to take into
acco
unt u
n
a
voidable
del
ay of control
sig
nal
s
becau
se of u
npre
d
icta
ble
distur
ban
ce
s
from the environment [4].
Given (4
) an
d (16
)
, the neces
sa
ry condition for
coo
r
din
ability with re
spe
c
t to the
probl
em of the Coo
r
din
a
tor (14) ta
ke
s the form:
i
(
t
)
(
t
),
x
i
(
t
+ 1
)
X
,
i
I
u
i
*
(
t
)
U
(
t
):
x
i
*(
t
+ 2 /
t
+ 1)) \
x
i
(
t
+ 2 /
t
)
x
i
(
t
+ 2 /
t
+ 1)
) \
x
i
(
t
+ 2
/
t
),
(17)
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 16
93-6
930
TELKOM
NIKA
Vol. 14, No. 4, Dece
mb
er 201
6 : 1408 – 141
6
1414
3.2. Coordinabilit
y
in Rel
a
tion to the
Global Task
Let us a
s
su
me first that
the two-lev
e
l sy
stem of
IDSs is si
n
g
le-p
urpo
se, and this
purp
o
se i
s
to
achi
eve a
giv
en exte
rnal
st
ate of the
Co
ordin
a
tor
x
0
X
0
. In the ab
ove-me
ntione
d
example with
the con
s
tru
c
tion, the purp
o
se
x
0
can b
e
formali
z
ed
as a cle
a
r d
e
s
cription of the
expected results (e.g., a drawing of
the
building). Then the
system
will be coordi
nated in rel
a
tion
to the task of achi
eving
x
0
, if the Coordinator will be able to find a set of predicted values
x
i
X
,
i
I
at the ti
me
t
so th
at (after thei
r i
s
suan
ce to
the
sub
o
rdi
nated
IDSs
and
sub
s
eq
uent im
pa
ct
of these
syst
ems on the environm
ent) t
he
Coordinat
or’s
state
will move closer to
x
0
. T
o
pr
ese
n
t
this statem
en
t more formall
y
, let us detail the task of the Coo
r
din
a
tor.
For th
e Co
ordin
a
tor, di
sturb
a
n
c
e
s
are
deviatio
n
s of th
e
curre
n
t state
s
of its
sub
o
rdi
nated
IDSs fro
m
th
eir given
valu
es, an
d
c
ontrol sig
nal
s a
r
e
expecte
d
sta
t
es of the l
o
wer-
level IDSs. A
c
cordi
ngly, af
ter receiving t
he fee
dba
ck
sign
als f
r
om t
he
subo
rdi
nat
ed IDS
s
(a o
ne-
step del
ay is assu
med t
o
exist for each IDS
a
nd for re
acti
on of the e
n
vironm
ent),
the
Coo
r
din
a
tor’
s state equatio
n can b
e
rep
r
ese
n
ted simil
a
r to (4
):
x
0
(
t
+ 4)
=
0
(
0
(
0
(
x
0
(t
)
u
0
(
t
)
0
(
t
))),
(18
)
Whe
r
e:
0
(
t
):
X
1
X
2
…
X
n
0
is the functio
nal
mappin
g
of
t
he imp
a
ct of
deviation
s of
the
curre
n
t state
s
of subo
rdin
ates IDS
s
fro
m
their
given
values u
pon
the gene
ral
state of the j
ob;
u
0
(
t
) is a ge
n
e
ral de
scri
ption of the pre
d
icted jo
b’s
state for the ne
xt step.
Relatio
n
(18
)
sho
w
s that
th
e Coo
r
dinato
r
DM
0
sh
ould
solve
t
w
o disparate
ta
sks: first,
to
evaluate the
curre
n
t pro
g
ress an
d to d
e
velop a
st
ra
tegy for furth
e
r
solving the
probl
em o
n
th
e
basi
s
of this
asse
ssm
ent;
se
con
d
, to all
o
cate
tasks f
o
r the
next
step am
ong
th
e subo
rdin
ated
IDSs. The
r
ef
ore, the stru
cture of the
Coo
r
din
a
tor
sho
u
ld be p
r
ese
n
ted in th
e form sh
own in
Figure 2.
Here: BOSS
s
t
ands
for the B
lock to O
bjectify the
current S
tate
of solvin
g the
proble
m
and to develo
p
S
trategie
s
for furthe
r a
c
tion; its state e
quation i
s
de
scribe
d by the relation
(18
)
;
BCA
i
denote
s
the B
lock
s to C
orrect A
ctions, which are respo
n
sibl
e for m
appin
g
a
gene
rali
zed d
e
scriptio
n of the job state predi
cted
for
the next step to the
anticipated state
s
of
their su
bo
rdin
ated DM
i
.
Then, by an
alogy with (16), the formul
ated a
b
o
v
e coo
r
din
a
b
ility statement for a
colle
ctive of IDSs in relatio
n
to the globa
l task can be
written in the
form:
t
T
,
x
i
(
t
)
X
i
,
i
I
,
x
0
X
0
x
i
*
(
t
+4
)
X
i
:
x
0
*(
t
+4
) \
x
0
x
0
*(
t
) \
x
0
.
(19)
If you do
not
sep
a
rate
BCAs withi
n
the
Coo
r
din
a
tor, i
t
sho
u
ld
re
cei
v
e distu
r
ba
nces
as a
vector
with the followin
g
co
mpone
nts:
i
0
=
x
i
(
t
+ 1) \
x
i
(
t
+ 1),
i
I,
(20)
And DM
0
will
dire
ctly gene
rate
x
i
as its o
u
tput sign
als.
Figure 2.
The
Coordinato
r
’
s
structu
r
e
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
1693-6
930
Planning
a
nd Coo
r
din
a
tion in
Hierarchie of
Intelligent Dynam
ic System
s (Alexa
nd
er Ya)
1415
4. Results a
nd Discu
ssi
on
Cha
nge
-ba
s
e
d
ma
nage
me
nt procedu
re
s
see
m
to
b
e
effective
in
differe
nt ap
p
lication
s
for
control
systems. In
the p
r
evoiu
s
and th
e
give
n pu
blication
s
, we h
a
ve
prop
osed
su
ch
method
s to coordi
nate inte
ractio
ns
amo
ng compo
nen
ts of a hie
r
a
r
chi
c
al
system
and to pl
an i
t
s
gene
ral be
ha
vior for both
nume
r
ic a
n
d
non-nume
r
ic
metrics upo
n the state spaces of the
s
e
comp
one
nts.
The
possibilit
y to co
nst
r
u
c
t hierarchi
e
s for
solv
ing
mu
lti-purpo
se ta
sks i
s
evident
as well.
Then, the
go
al state
of th
e Co
ordinato
r
(
x
0
in Figu
re
2) will
de
p
end
o
n
time and sho
u
ld b
e
sele
cted
withi
n
the
system
with th
e u
s
e
of a
prefe
r
e
n
ce
rel
a
tion
on the
set of
goal
s [13]. S
u
ch
probl
em
s
ca
n a
r
ise, for
example, in
probl
em
s of
stru
cture
con
t
rol for the
virtual
enterpri
s
e
s
[22, 23]. However, they req
u
ire for a
sep
a
rate con
s
ide
r
ation.
5. Conclusio
n
For lo
cally o
r
gani
zed
hierarchical colle
ctives
of int
e
lligent dynami
c
sy
stem
s, we hav
e
found ne
ce
ssary co
ndition
s of coo
r
di
na
bility both with re
spe
c
t to the Co
ordi
nat
or’s ta
sk and
in
relation
to th
e glo
bal ta
sk of the
whole
hie
r
archy. B
e
sid
e
s,
we
h
a
ve p
r
opo
se
d a
procedu
re for
dire
ct synthe
sis
a
plan
to control su
ch
a
hie
r
a
r
ch
y. Furthe
r
re
sea
r
ch
in
this fiel
d ou
ght to
co
ver
at least the followin
g
directions: se
arch
for
necessa
ry
condition
s of
coordina
bility and plannin
g
;
developm
ent
of spe
c
ific
co
ordin
a
tion alg
o
rithm
s
fo
r
certain
cla
s
se
s of job
s
; loo
k
ing fo
r way
s
to
prevent
confli
cts a
m
ong th
e de
cisi
on m
a
ke
rs
re
sp
on
sible fo
r diffe
rent sub
s
yste
ms withi
n
th
e
hiera
r
chy; extending i
dea
s of increme
n
tal co
ordi
natio
n to othe
r problem
s
like l
ogisti
cs, virtu
a
l
enterp
r
i
s
e
s
, etc.
Ackn
o
w
l
e
dg
ements
The autho
r would like to than
k the Ru
ssi
an
Fou
nda
tion for Basic Research
es (gra
nt
s
14-0
7
-0025
7,
15-07-047
60
, 15-0
7
-0
275
7, 16-29-044
24, and
16
-2
9-12
901
) fo
r
partial fu
ndin
g
of
t
h
is re
sea
r
ch.
Referen
ces
[1]
T
a
kahashi S, Kijima K, Sato R
.
Applie
d Gene
ra
l S
y
stems R
e
search o
n
Organiz
a
tions. Spr
i
ng
er. 200
4.
[2]
Delg
ad
o J, P
u
jol
J, San
g
u
e
sa R. Em
ergenc
e of C
o
o
r
din
a
tion
in S
c
ale-F
r
e
e
Net
w
o
r
ks.
Web
Intelli
genc
e an
d Agent Syste
m
s
. 20
03; 1: 131-1
38.
[3]
Lee G, C
h
o
ng
NY. Self-org
an
izing
Mob
ile
S
ensor
Net
w
ork:
Distrib
uted T
opol
og
y C
ontro
l
F
r
ame
w
ork.
In: Hand
book
of Researc
h
on Ambi
ent
In
tellig
enc
e
and Sm
art Enviro
nments:
T
r
ends and
Perspectiv
e
s. IGI Global. 2011: 558-584.
[4]
Mesarovic MD, Macko D,
T
a
kah
a
ra Y. T
h
eor
y of Hier
a
r
c
hical
M
u
ltil
ev
el S
y
stems. Ne
w
-
York a
n
d
Lon
do
n: Acad. Press. 1970.
[5]
Baroni M, Dinu G, Krusze
w
s
ki G.
Don
’
t co
unt, pred
ict! a systematic
c
o
mp
ariso
n
of co
ntext-counti
n
g
vs. context-pre
d
ictin
g
se
ma
ntic vectors
. 52
n
d
Ann
ual M
eet
ing
of the
Ass
o
ciati
on for C
o
mputatio
nal
Lin
guistics. 20
14: 1: 238-
247.
[6]
Be
ydo
un G, Garcia-S
anch
e
z
F
,
Vincent-T
o
rres CM
, Lop
e
z
-Lorca AA, M
a
rtinez-B
ejar
R. Providi
n
g
metrics and
automatic e
n
hanc
ement
fo
r hierarch
ical
taxo
nomi
e
s.
Informatio
n Processi
ng &
Mana
ge
me
nt.
201
3; 49(1): 67
-82.
[7]
Manteg
na
RN.
Hier
a
rchic
a
l S
t
ructure in
F
i
n
anci
a
l Mark
ets
.
T
he Euro
pe
a
n
Physic
a
l J
o
u
r
nal B.
19
99;
11: 193-
19
7.
[8]
Jonso
n
RA, Wicher
n DW
. Applie
d Multiv
ari
a
te St
atistical
Anal
ys
is. 6th edn.
Pears
on
Prentice H
a
ll.
200
7.
[9]
EB F
o
routa
n
,
H Kh
otanl
ou. I
m
provin
g s
e
m
antic
c
l
usteri
ng
usi
ng
w
i
th O
n
tolo
g
y
a
n
d
rul
e
s.
Internatio
na
l
Journ
a
l of Elec
trical an
d Co
mputer Eng
i
n
eeri
ng (IJECE)
. 20
14; 4(1): 7-15.
[10]
Jian
g W
,
Z
hang J, Li J. A Mu
lti-ag
ent Sup
p
l
y
C
hai
n Inform
ation C
oor
din
a
t
ion Mod
e
Bas
ed o
n
Clo
u
d
Comp
uting.
T
E
LKOMNIKA Indon
esia
n Jour
nal
of Electric
al
Engin
eeri
ng.
2
013; 11(
11): 64
27-6
433.
[11]
W
i
cakson
o
H,
Khos
w
a
nto
H
,
Kus
w
ad
i S.
Behav
iors Co
ordi
natio
n
a
n
d
Lear
ni
ng on Autonom
ous
N
a
vi
ga
ti
on
o
f
Phy
s
i
c
al
R
o
bo
t.
T
E
LKOMNIKA T
e
leco
mmu
n
i
catio
n
Co
mp
ut
ing El
ectron
ics
and C
ontrol.
201
1; 9(3): 473
-482.
[12] Qian L
i
, Limi
n
Jia, Ch
ang
xu J
i
, Jian
lin
g H, Z
hon
g
y
u
an J. Si
mulati
ng o
n
Pa
sseng
ers Co
or
din
a
tion
w
i
th
Distributi
o
n
Se
rvice i
n
R
a
i
l
w
a
y Stati
on.
T
E
L
K
OMNIKA Ind
ones
ian
Jo
urn
a
l
of Electric
al
Eng
i
ne
eri
ng.
201
3; 11(1): 39
2-39
8.
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 16
93-6
930
TELKOM
NIKA
Vol. 14, No. 4, Dece
mb
er 201
6 : 1408 – 141
6
1416
[13]
Osipov G. Intellig
ent dy
namic
s
y
stems.
Scie
n
t
ific and T
e
c
h
n
i
cal Infor
m
ati
o
n Processi
ng
. 201
0;
37(5):
259
–2
64.
[14]
Is
w
a
nto, W
a
h
y
u
n
g
goro
O, Cah
y
a
d
i
AI. Quadr
ot
or Pat
h
Plan
ni
ng B
a
s
ed
on M
odifi
e
d
F
u
zz
y C
e
ll
Decom
positi
o
n
Algor
ithm.
T
E
LKOMNIKA T
e
leco
mmunic
a
tion
Co
mputi
ng El
ectron
ics
and
Co
ntrol
.
201
6; 14 (2): 6
55-6
64.
[15]
Jing
w
e
i L, Yifei T
,
Shaofen
g W,
Qingmeng T,
Dongbo L. Weld
ing
Robot Kinematics A
naly
s
is
and
T
r
ajector
y
Pl
a
nni
ng.
TELKOMNIKA Teleco
mmu
n
icati
on
Co
mp
uting E
l
e
c
tronics a
nd
Contro
l
. 201
6;
14(2A): 92-
10
0
.
[16]
Xu
eso
ng Y, Qing
hua W
,
Hammin L. An
Improve
d
Rob
o
t
Path Planni
n
g
Algorithm.
TELKOMNIKA
T
e
leco
mmunic
a
tion C
o
mputi
n
g Electron
ics a
nd Co
ntrol.
De
cember 2
0
1
2
; 10(4): 62
9-6
3
6
.
[17]
F
r
idman OV, F
r
idman A Ya.
Decreas
ing
D
i
ssemin
a
tio
n
of
Disturba
n
ces
w
i
t
h
in
Net
w
ork
S
y
stems b
y
Neur
al Net
w
or
ks.
T
E
LKOMN
I
KA Indonesi
a
n Journ
a
l of Electrical En
gi
neer
ing.
20
13;
11(9): 494
2-
494
8.
[18]
Marler
RT
, Arora JS. Surv
e
y
of multi-
obj
ecti
ve o
p
timizati
on
metho
d
s for
e
ngi
neer
in
g.
Struct Multidisc
Optim
. 20
04; 2
6
: 369-3
95.
[19] Stefanuk
V
L
.
Stability of Local Control
in
a System
with Contralinear
Interaction of
Subsystem
s
.
Europ
e
a
n
Cont
rol Co
nferenc
e
(ECC'
93), Gronin
gen. 1
993;
1: 117-1
19.
[20] Osipov
G.
Attaina
b
le S
e
ts an
d Know
le
dg
e Base Arch
itecture in
Discrete
Dyna
mic K
n
o
w
ledge-
bas
e
d
System
s
. W
o
r
kshop
"Ap
p
li
ed
Semi
otics: Co
ntrol Pr
obl
ems
(ASC
200
0)".
ECAI20
00.
14
th Euro
pe
an
Confer
ence
of Artificial Intel
lig
ence. Berl
in. 2
000: 39-
43.
[21]
Vino
grad
ov A
N
, Z
h
il
yak
o
va
L Yu, Osip
ov GS. D
y
na
mic Intell
ige
n
t
S
y
stems: I. Kno
w
l
e
dg
e
Repr
esentati
o
n
and
Basic
Al
gorithms.
C
o
mputer a
nd
Systems
Scie
nces
Internati
ona
l.
200
2;
41(
6):
953-
960.
[22]
Sokol
o
v B, F
r
idman A.
Integ
r
ated Situati
o
n
a
l Mod
e
ll
ing o
f
Industry-Busi
ness Process
e
s for Every
Stage of T
heir
Life Cycle
. 4th International IEEE C
onference “Intelligent S
y
stems” (IS
2008). Varna,
Bulg
aria. 2
008;
1: 8-40.
[23]
Sokol
o
v B, Iva
nov
D, F
r
idma
n A. Situ
atio
na
l Mo
d
e
l
lin
g for
Structural
D
y
namics
Co
ntrol
of Ind
u
str
y
-
Busin
e
ss Proc
esses
and
Su
ppl
y
Ch
ains. I
n
: Sgur
ev V,
Hadj
iski M,
Ka
cprz
y
k
Ja.
Ed
i
t
o
r
s
. Intelligent
S
y
stems: F
r
om
T
heor
y
to Pra
c
tice. Berlin
H
e
idel
ber
g: Sprin
ger-Verl
ag. 20
10: 279-
30
8.
Evaluation Warning : The document was created with Spire.PDF for Python.