 Existence Conditions of Asymptotically Stable 2-D Feedback Control
Systems on the Basis of Block Matrix Diagonalization
Giido Izuta
Yonezawa Women’s Junior College, Yonezawa, Yamagata, Japan
Keywords:
2-D Systems, Asymptotic Stability, Feedback Control, Block Matrix Diagonalization.
Abstract:
This work is concerned with the existence of asymptotically stable 2-D (2-dimensional) systems by means of a
feedback control model represented by the system of partial difference equations and their Lagrange solutions.
Thus, the goal is to establish a controller thatprovides a feedback control system with state variables depending
solely on its Lagrange solution in the sense that the solution to the variable state is not a linear combination
of other Lagrange solutions. Roughly speaking, the results showed that, to achieve such a control system,
the controller has to diagonalize the block matrices of the matrices composing the system description model.
Finally, a numerical example is presented to show how the controller is designed in order to generate a stable
feedback control with given Lagrange solutions.
1 INTRODUCTION
Discrete 2-D control systems have been a subject
of research for over half a century now; and many
approaches have been proposed to study the stabil-
ity and design of feedback control systems. As far
as the mathematical model is concerned with, either
the Roesser system description (Givone and Roesser,
1972) or the FM model (Attasi, 1973; Fornasini and
Marchesini, 1978; Fornasini and Marchesini, 1980),
which are essentially equivalent to one another in the
sense that it is possible to transform the partial dif-
ference equations describing one model into the other
(Kaczorek, 1985), is taken as the system representa-
tion to handle this type of control systems. In addi-
range of mathematical ﬁelds. Among those, the z
transform showed to be a very effective strategy to in-
vestigate systems with single input and single output
corresponding to what is called the class of ‘delayless
systems’ in the ordinary 1-D system counterpart the-
ory; and a great deal of stability criteria and controller
design procedures were accomplished for 2-D con-
trol systems (Anderson and Jury, 1974; Bose, 1982;
Tzafestas, 1986; Lim, 1990).
During the time period around the turning of the
century, a remarkable shift in the paradigm from han-
dling those problems on the grounds of analytical
mathematics into the incorporation of computational
methods, which in control system theory is closely as-
sociated to the energy method framework, took place.
In fact, this approach relies basically on the Lyapunov
stability theory as well as the optimization algorithms,
and has undergone tremendous developments hand in
hand with the advances in computational techniques;
so that the formalization of the control system de-
sign based on the linear matrix inequalities gained
widespread popularity and established its status as a
standard procedure to handle not only the ordinary
1-D but also multi-input multi-output n-D systems
(Piekarski, 1977; Du and Xie, 2002; Pazke et al.,
2004; Izuta, 2007; Rogers et al., 2007). However,
this procedure has been pointed out to yield conserva-
tive controllers in the control theory sense. Also it is
worth noting that many other techniques have ﬂour-
ished and contributed to the broadening of the ﬁeld
(Jerri, 1996; Zerz, 2000; Cheng, 2003; Elaydi, 2005;
Izuta, 2012; Izuta, 2014a).
Unlike the theoretical accounts aforementioned,
this paper examines the existence conditions of
asymptotically stable 2-D feedback control systems
from the Lagrange method frame of reference, which
has shed light on the stability analysis problem (Izuta,
2010; Izuta, 2014b). In these reports, the Lagrange
method was used in conjunction with Jordan matrix
transformation in order to transform the original sys-
tem of partial difference equations into a system com-
posed by only diagonal matrices, which allow the
transformed system be solved analytically.
Motivated by these results, this paper aims to ﬁg-
Izuta, G.
Existence Conditions of Asymptotically Stable 2-D Feedback Control Systems on the Basis of Block Matrix Diagonalization.
DOI: 10.5220/0005975604630470
In Proceedings of the 13th International Conference on Informatics in Control, Automation and Robotics (ICINCO 2016) - Volume 1, pages 463-470
ISBN: 978-989-758-198-4
c
463 ure out the conditions that the controller has to ful-
ﬁll so as to form a feedback control system hav-
ing deﬁned asymptotically stable Lagrange solutions.
Moreover, the kind of systems concerned with here
corresponds to the class of ‘systems with delays’ in
the 1-D case.
Finally, the paper is organized as follows: in sec-
tion 2, the 2-d control system and the concepts neces-
sary throughout the paper are presented; the asymp-
totic stability conditions are establish in section 3; a
numerical example is provided in 4; and some ﬁnal
remarks are enunciated in section 5 .
2 PRELIMINARIES
In this section, the concepts and deﬁnitions used
throughout this paper are presented. Firstly, the 2-d
control system to be investigated is presented here.
Deﬁnition 2.1. Let the 2-d control system be given by
the following system of partial difference equations.
x
p
(i+ 1, j)
x
q
(i, j+ 1)
=
A
11
A
12
A
21
A
22
x
p
(i, j)
x
q
(i, j)
+
B
11
B
12
B
21
B
22
x
p
(i δ
p
, j σ
p
)
x
q
(i δ
q
, j σ
q
)
+
C
11
C
12
C
21
C
22
u
p
(i, j)
u
q
(i, j)
(1)
where x
p
(i, j) and x
q
(i, j) are vectors representing the
states of the system and given by
x
p
(i+ 1, j) =
def
x
1
(i+ 1, j)
.
.
.
x
n
(i+ 1, j)
x
q
(i, j+ 1) =
def
ˆx
1
(i, j+ 1)
.
.
.
ˆx
n
(i, j+ 1)
(2)
with the indices i and j (i, j Z) indicating that the
system is discrete and doubly indexed. In addition,
the inputs of the system u
p
(i, j) and u
q
(i, j) are and
are deﬁned similarly to (2).
Assumption 2.2. Hereafter the matrices C
11
, C
12
,
C
21
, C
22
are all real valued non-singular matrices.
In order to simplify the notations, the following
(i, j) =
p
(i, j)
q
(i, j)
=
def
1
(i, j)
.
.
.
n
(i, j)
ˆ
1
(i, j)
.
.
.
ˆ
n
(i, j)
,
= x
or
= u
x(i+ 1, j+ 1) =
def
x
1
(i+ 1, j)
.
.
.
x
n
(i+ 1, j)
ˆx
1
(i, j+ 1)
.
.
.
ˆx
n
(i, j+ 1)
x(i δ, j σ) =
def
x
1
(i δ
1
, j σ
1
)
.
.
.
x
n
(i δ
n
, j σ
n
)
ˆx
1
(i
ˆ
δ
1
, j
ˆ
σ
1
)
.
.
.
ˆx
n
(i
ˆ
δ
n
, j
ˆ
σ
n
)
(3)
Thus, a compact notation for (1) is represented by
x(i+ 1, j+ 1) = Ax(i, j) + Bx(i δ, j σ)
+ Cu(i, j)
(4)
Remark 2.3. Small bold faced letters stand for vec-
tors whereas their non-bold faced counterparts mean
the variables. Capital letters describe either matrices
or block matrix.
Deﬁnition 2.4. A 2-d feedback control system is a
system accomplished from (1) by setting the inputs
variables be the states of the system itself. That is
to say
x(i+ 1, j+ 1) =
¯
Ax(i, j) +
¯
Bx(i δ, j σ)
(5)
in which
u(i, j) = Fx(i, j) Gx(i δ, j σ)
¯
A = A CF
¯
B = B CG
(6)
Remark 2.5. Under the notations of (1), (6) yields
¯
A
11
= A
11
C
11
F
11
C
12
F
21
¯
A
12
= A
12
C
11
F
12
C
12
F
22
¯
A
21
= A
21
C
21
F
11
C
22
F
21
¯
A
22
= A
22
C
21
F
12
C
22
F
22
¯
B
11
= B
11
C
11
G
11
C
12
G
21
¯
B
12
= B
12
C
11
G
12
C
12
G
22
¯
B
21
= B
21
C
21
G
11
C
22
G
21
¯
B
22
= B
22
C
21
G
12
C
22
G
22
(7)
In this paper, the system is asymptotically stable
if all the state variables vanish as the indices increase.
ICINCO 2016 - 13th International Conference on Informatics in Control, Automation and Robotics
464 Deﬁnition 2.6. 2-D feedback control system (5) is
said to be asymptotically stable if all the solutions
x
1
(i, j), ..., ˆx
n
(i, j) fulﬁll the conditions described by
lim
(i+ j)
| x
1
(i, j) |→ 0
.
.
.
lim
(i+ j)
| ˆx
n
(i, j) |→ 0
(8)
Furthermore, an asymptotically stable Lagrange
solution is an analytic solution given by
Deﬁnition 2.7. A non-null asymptotically stable La-
grange solution to the state variable is the equation
x
(i, j) =
n
k=1
I
k
α
i
k
β
j
k
+
n
k=1
ˆ
I
k
ˆ
α
i
k
ˆ
β
j
k
(9)
where I
s and
ˆ
I
s are the initial values and non-null
real numbers α
s,
ˆ
α
s , β
s and
ˆ
β
s are such that
|α
|, |
ˆ
α
|, |β
|, |
ˆ
β
| < 1 .
Finally, this paper handles the following problem.
Problem 2.8. To establish conditions on the feedback
controller matrices for the feedbackcontrol system (5)
have asymptotically stable Lagrange solution given
by
x
1
(i, j)
.
.
.
x
t
(i, j)
=
I
1
α
i
1
β
j
1
.
.
.
I
n
α
i
n
β
j
n
ˆ
I
1
ˆ
α
i
1
ˆ
β
j
1
.
.
.
ˆ
I
m
ˆ
α
i
m
ˆ
β
j
m
(10)
such that the numbers |α
r
| < 1, |
ˆ
α
r
| < 1 , |β
s
| < 1 and
|
ˆ
β
s
| < 1 , r, s.
Remark 2.9. The initial condition for (10) yield
x(0, 0) =
"
x
p
(0, 0)
x
q
(0, 0)
#
=
K
p
0
0 K
q
1
.
.
.
1
(11)
where
K
p
=
"
I
1
0
0 I
n
#
, K
q
=
"
ˆ
I
1
0
0
ˆ
I
n
#
(12)
3 RESULTS
Henceforth the solution to the problem is presented
gathered in a theorem. However, before doing it, a
very basic result is given for the sake of the compu-
tational procedures that are to be followed in the nu-
merical example.
Proposition 3.1. Let the controller be as in (6) and
(7). Then, their deﬁning matrices are computed as
F
11
= C
1
11
(A
11
¯
A
11
)
C
1
11
C
12
(C
1
11
C
12
C
1
21
C
22
)
1
×
C
1
11
(A
11
¯
A
11
) C
1
21
(A
21
¯
A
21
)
F
21
= (C
1
11
C
12
C
1
21
C
22
)
1
×
C
1
11
(A
11
¯
A
11
) C
1
21
(A
21
¯
A
21
)
F
12
= (C
1
22
C
21
C
1
12
C
11
)
1
×
C
1
22
(A
22
¯
A
22
) C
1
12
(A
12
¯
A
12
)
F
22
= C
1
22
(A
22
¯
A
22
)
C
1
22
C
21
(C
1
22
C
21
C
1
12
C
11
)
1
×
C
1
22
(A
22
¯
A
22
) C
1
12
(A
12
¯
A
12
)
(13)
and
G
11
= C
1
11
(B
11
¯
B
11
)
C
1
11
C
12
(C
1
11
C
12
C
1
21
C
22
)
1
×
C
1
11
(B
11
¯
B
11
) C
1
21
(B
21
¯
B
21
)
G
21
= (C
1
11
C
12
C
1
21
C
22
)
1
×
C
1
11
(B
11
¯
B
11
) C
1
21
(B
21
¯
B
21
)
G
12
= (C
1
22
C
21
C
1
12
C
11
)
1
×
C
1
22
(B
22
¯
B
22
) C
1
12
(B
12
¯
B
12
)
G
22
= C
1
22
(B
22
¯
B
22
)
C
1
22
C
21
(C
1
22
C
21
C
1
12
C
11
)
1
×
C
1
22
(B
22
¯
B
22
) C
1
12
(B
12
¯
B
12
)
(14)
Proof. It follows straightforwardly from assumption
2.2.
Theorem 3.2. Let the 2-D feedback control system be
given by (5). If there exist diagonal matrices
¯
A
11
,
¯
A
12
,
¯
A
21
,
¯
A
22
,
¯
B
11
,
¯
B
12
,
¯
B
21
and
¯
B
22
in the sense of the
feedback controller (14) and such that the following
conditions all are satisﬁed then the control system is
asymptotically stable and its Lagrange solutions can
be established.
1. the diagonal entries of the diagonal matrices
¯
A
11
+
¯
B
11
¯
B
1
21
¯
A
21
, and
¯
A
22
+
¯
B
22
¯
B
1
12
¯
A
12
have all
non-null absolute values less then unit.
2. the diagonal entries of diagonal matrices
Existence Conditions of Asymptotically Stable 2-D Feedback Control Systems on the Basis of Block Matrix Diagonalization
465 ¯
A
1
21
¯
B
21
α
δ
1
1
0
.
.
.
0 α
δ
n
n
(15)
and
ˆ
β
ˆ
σ
1
1
0
.
.
.
0
ˆ
β
ˆ
σ
n
n
¯
A
1
12
¯
B
12
(16)
have all positive values less then unit.
Proof. Firstly, note that (5) can be rewritten as
x
p
(i+ 1, j) =
¯
A
11
x
p
(i, j) +
¯
A
12
x
q
(i, j)
+
¯
B
11
x
p
(i δ, j σ)+
¯
B
12
x
q
(i δ, j σ)
x
q
(i, j+ 1) =
¯
A
21
x
p
(i, j) +
¯
A
22
x
q
(i, j)
+
¯
B
21
x
p
(i δ, j σ)+
¯
B
22
x
q
(i δ, j σ)
(17)
Now substituting (10) into the ﬁrst equation in
(17) leads to the subsystem described by
I
1
α
i+1
1
β
j
1
.
.
.
I
n
α
i+1
n
β
j
n
¯
A
11
I
1
α
i
1
β
j
1
.
.
.
I
n
α
i
n
β
j
n
¯
B
11
I
1
α
iδ
1
1
β
jσ
1
1
.
.
.
I
n
α
iδ
n
n
β
jσ
n
n
=
¯
A
12
ˆ
I
1
ˆ
α
i
1
ˆ
β
j
1
.
.
.
ˆ
I
n
ˆ
α
i
n
ˆ
β
j
n
¯
B
12
ˆ
I
1
ˆ
α
i
ˆ
δ
1
1
ˆ
β
j
ˆ
σ
1
1
.
.
.
ˆ
I
n
ˆ
α
i
ˆ
δ
n
n
ˆ
β
j
ˆ
σ
n
n
(18)
whereas the second equation in (17) provides the set
of equations gathered as
ˆ
I
1
¯
α
i
1
¯
β
j+1
1
.
.
.
ˆ
I
n
¯
α
i
n
¯
β
j+1
n
¯
A
22
¯
I
1
¯
α
i
1
¯
β
j
1
.
.
.
¯
I
n
¯
α
i
n
¯
β
j
m
¯
B
22
ˆ
I
1
¯
α
i
¯
δ
1
1
¯
β
j
¯
σ
1
1
.
.
.
ˆ
I
n
¯
α
i
¯
δ
n
n
¯
β
j
¯
σ
n
n
=
¯
A
21
I
1
α
i
1
β
j
1
.
.
.
I
n
α
i
n
β
j
n
¯
B
21
I
1
α
iδ
1
1
β
jσ
1
1
.
.
.
I
n
α
iδ
n
n
β
jσ
n
n
(19)
Splitting the matrices further in order to single out
the terms related to the index, (18) yields
K
p
α
i+1
1
β
j
1
.
.
.
α
i+1
n
β
j
n
¯
A
11
K
p
α
i
1
β
j
1
.
.
.
α
i
n
β
j
n
¯
B
11
K
p
α
iδ
1
1
β
jσ
1
1
.
.
.
α
iδ
n
n
β
jσ
n
n
=
¯
A
12
K
q
ˆ
α
i
1
ˆ
β
j
1
.
.
.
ˆ
α
i
n
ˆ
β
j
n
¯
B
12
K
q
ˆ
α
i
ˆ
δ
1
1
ˆ
β
j
ˆ
σ
1
1
.
.
.
ˆ
α
i
ˆ
δ
n
n
ˆ
β
j
ˆ
σ
n
n
(20)
on the other hand (19) produces
K
q
ˆ
α
i
1
ˆ
β
j+1
1
.
.
.
ˆ
α
i
n
ˆ
β
j+1
n
¯
A
22
K
q
ˆ
α
i
1
ˆ
β
j
1
.
.
.
ˆ
α
i
n
ˆ
β
j
n
¯
B
22
K
q
ˆ
α
i
ˆ
δ
1
1
ˆ
β
j
ˆ
σ
1
1
.
.
.
ˆ
α
i
ˆ
δ
n
n
ˆ
β
j
ˆ
σ
n
n
=
¯
A
21
K
p
α
i
1
β
j
1
.
.
.
α
i
t
β
j
t
¯
B
21
K
p
α
iδ
1
1
β
jσ
1
1
.
.
.
α
iδ
n
n
β
jσ
n
n
(21)
Yet, equations (20) and (21) can be recasted as
K
p
α
1
0
.
.
.
α
n
¯
A
11
K
p
¯
B
11
K
p
α
δ
1
1
β
σ
1
1
0
.
.
.
0 α
δ
n
n
β
σ
n
n
×
α
i
1
β
j
1
.
.
.
α
i
n
β
j
n
=
¯
A
12
K
q
¯
B
12
K
q
×
ˆ
α
ˆ
δ
1
1
ˆ
β
ˆ
σ
1
1
0
.
.
.
0
ˆ
α
ˆ
δ
n
n
ˆ
β
ˆ
σ
n
n
×
ˆ
α
i
1
ˆ
β
j
1
.
.
.
ˆ
α
i
n
ˆ
β
j
n
(22)
and
ICINCO 2016 - 13th International Conference on Informatics in Control, Automation and Robotics
466 K
q
ˆ
β
1
0
.
.
.
ˆ
β
n
¯
A
22
K
q
¯
B
22
K
q
ˆ
α
ˆ
δ
1
1
ˆ
β
ˆ
σ
1
1
0
.
.
.
0
ˆ
α
ˆ
δ
n
n
ˆ
β
ˆ
σ
n
n
×
ˆ
α
i
1
ˆ
β
j
1
.
.
.
ˆ
α
i
n
ˆ
β
j
n
=
¯
A
21
K
p
¯
B
21
K
p
α
δ
1
1
β
σ
1
1
0
.
.
.
0 α
δ
n
n
β
σ
n
n
×
α
i
1
β
j
1
.
.
.
α
i
n
β
j
n
(23)
respectively. Note that these equations have terms de-
pending only on αs and βs on one side of the equal-
ity, and
ˆ
αs and
ˆ
βs on the other side. Furthermore
this equality has to hold for all values that the indices
assume. Since, in general, the pairs of αs and βs,
and the corresponding pairs
ˆ
αs and
ˆ
βs are not equal,
these equalities are valid only if the following expres-
sions are valid.
K
p
α
1
0
.
.
.
α
n
¯
A
11
K
p
¯
B
11
K
p
α
δ
1
1
β
σ
1
1
0
.
.
.
0 α
δ
n
n
β
σ
n
n
= 0
(24)
¯
A
12
K
q
¯
B
12
K
q
×
ˆ
α
ˆ
δ
1
1
ˆ
β
ˆ
σ
1
1
0
.
.
.
0
ˆ
α
ˆ
δ
n
n
ˆ
β
ˆ
σ
n
n
= 0
(25)
K
q
ˆ
β
1
0
.
.
.
ˆ
β
n
¯
A
22
K
q
¯
B
22
K
q
ˆ
α
ˆ
δ
1
1
ˆ
β
ˆ
σ
1
1
0
.
.
.
0
ˆ
α
ˆ
δ
n
n
ˆ
β
ˆ
σ
n
n
= 0
(26)
¯
A
21
K
p
¯
B
21
K
p
α
δ
1
1
β
σ
1
1
0
.
.
.
0 α
δ
n
n
β
σ
n
n
= 0
(27)
Hence, (24) and (27) yield
K
p
α
1
0
.
.
.
α
n
¯
A
11
K
p
¯
B
11
K
p
(
¯
B
21
K
p
)
1
¯
A
21
K
p
= 0
(28)
taking into account the hypothesis of the theorem (28)
translates into
α
1
0
.
.
.
0 α
n
=
¯
A
11
+
¯
B
11
¯
B
1
21
¯
A
21
(29)
As for β
( = 1, ··· , n), they are computed from
(27) , which with the hypothesis of the theorem lead
to
β
σ
1
1
0
.
.
.
0 β
σ
n
n
=
¯
A
1
21
¯
B
21
α
δ
1
1
0
.
.
.
0 α
δ
n
n
(30)
Similarly, from (26) and (25)
K
q
ˆ
β
1
0
.
.
.
0
ˆ
β
n
¯
A
22
K
q
¯
B
22
K
q
(
¯
B
12
K
q
)
1
¯
A
12
K
q
= 0
(31)
and it turns out that the hypothesis of the theorem
brings up
ˆ
β
1
0
.
.
.
ˆ
β
n
=
¯
A
22
+
¯
B
22
¯
B
1
12
¯
A
12
(32)
In the same way,
ˆ
α
, = 1, ··· , n are computed
from (25) according to
ˆ
α
ˆ
δ
1
1
0
.
.
.
0
ˆ
α
ˆ
σ
n
n
=
ˆ
β
ˆ
σ
1
1
0
.
.
.
0
ˆ
β
ˆ
σ
n
n
¯
A
1
12
¯
B
12
(33)
Finally, (29), (30), (32), and (33) establish the
claim of the theorem.
Existence Conditions of Asymptotically Stable 2-D Feedback Control Systems on the Basis of Block Matrix Diagonalization
467 In practical terms, the key to solving the problem
comes down to the determination of the diagonal ma-
trices. The following remark provides a directive to
roughly evaluate the diagonal matrices.
Remark 3.3. It is clear from (15) and (16) that the
diagonal entries of
¯
A
1
21
¯
B
21
as well as
¯
A
1
12
¯
B
12
must
have absolute values smaller than unit in order to be
able to establish Lagrange solutions. Thus, in prac-
tice, begin focusing on these diagonal matrices trying
to set the product at relatively small values; then set
each of the matrices individually. Once these matri-
ces are deﬁned, handle (29) and (32) to deﬁne the re-
maining diagonal matrices. The controller matrices
are computed only after an estimate of these diagonal
matrices are obtained.
In the next section, a numerical example is pre-
sented to show how to make the calculations.
4 NUMERICAL EXAMPLE
Let the 2-D control system be given by
x
1
(i+ 1, j)
x
2
(i+ 1, j)
ˆx
1
(i, j + 1)
ˆx
2
(i, j + 1)
=
0.66 0.19 0.29 0.13
0.15 0.27 0.53 0.35
0.83 0.43 0.31 0.55
0.44 0.17 0.29 0.57
x
1
(i, j)
x
2
(i, j)
ˆx
1
(i, j)
ˆx
2
(i, j)
+
0.10 0.23 0.47 0.15
0.20 0.37 0.17 0.25
0.51 0.67 0.71 0.15
0.35 0.11 0.31 0.13
×
x
1
(i δ
1
, j σ
1
)
x
2
(i δ
2
, j σ
2
)
ˆx
1
(i
ˆ
δ
1
, j
ˆ
σ
1
)
ˆx
2
(i
ˆ
δ
2
, j
ˆ
σ
2
)
+
0.70 0.13 0.17 0.11
0.05 0.80 0.07 0.15
0.11 0.23 0.75 0.17
0.15 0.13 0.37 0.90
u
1
(i, j)
u
2
(i, j)
ˆu
1
(i, j)
ˆu
2
(i, j)
(34)
from which the composing matrices read
A
11
=
0.66 0.19
0.15 0.27
A
12
=
0.29 0.13
0.53 0.35
A
21
=
0.83 0.43
0.44 0.17
A
22
=
0.31 0.55
0.29 0.57
B
11
=
0.10 0.23
0.20 0.37
B
12
=
0.47 0.15
0.17 0.25
B
21
=
0.51 0.67
0.35 0.11
B
22
=
0.71 0.15
0.31 0.13
C
11
=
0.70 0.13
0.05 0.80
C
12
=
0.17 0.11
0.07 0.15
C
21
=
0.11 0.23
0.15 0.13
C
22
=
0.75 0.17
0.37 0.90
(35)
Here
¯
A
1
21
¯
B
21
as well as
¯
A
1
12
¯
B
12
are required to
satisfy
¯
A
1
21
¯
B
21
=
"
0.0667 0.000
0.0000 0.1667
#
(36)
as well as
¯
A
1
12
¯
B
12
=
"
0.0723 0.000
0.0000 0.2704
#
(37)
From these and keeping in mind the equations(29)
and (32), the diagonal matrices are set as
¯
A
11
=
0.11 0.00
0.00 0.67
¯
A
12
=
6.50 0.00
0.00 2.70
¯
A
21
=
1.50 0.00
0.00 3.00
¯
A
22
=
0.51 0.00
0.00 0.47
¯
B
11
=
0.05 0.00
0.00 0.01
¯
B
12
=
0.47 0.00
0.00 0.73
¯
B
21
=
0.10 0.00
0.00 0.50
¯
B
22
=
0.03 0.00
0.00 0.07
(38)
ICINCO 2016 - 13th International Conference on Informatics in Control, Automation and Robotics
468 Thus (7) gives the controller matrices
F
11
=
0.9422 0.5315
0.0816 0.0607
F
21
=
1.2451 1.3361
0.8320 3.7911
F
22
=
0.4778 1.6626
1.5606 0.1926
F
12
=
9.4028 0.3825
0.9157 3.0708
(39)
along with
G
11
=
0.1045 0.1509
0.1804 0.5314
G
21
=
0.4637 0.9148
0.1896 0.9113
G
22
=
0.8978 0.3677
0.0055 0.0334
G
12
=
0.2451 0.2494
0.1503 0.6415
(40)
Now taking into account expressions in (29) and
(30)
"
α
1
0
0 α
2
#
=
"
0.76 0.00
0.00 0.73
#
(41)
and
"
β
σ
1
1
0
0 β
σ
2
2
#
=
"
0.0667 0.0000
0.0000 0.1667
#
×
(0.76)
δ
1
0.00
0.00 (0.73)
δ
2
(42)
are computed. Note that as far as the diagonal values
of the resulting matrix on the right hand side of (42)
have absolute values less than unit, βs compose La-
grange solutions to the control system. On the other
hand (32) and (33) reduce to
"
ˆ
β
1
0
ˆ
β
2
#
=
"
0.9249 0.0000
0.0000 0.7289
#
(43)
and
"
ˆ
α
ˆ
δ
1
1
0
0
ˆ
α
ˆ
σ
2
2
#
=
"
0.0723 0.0000
0.0000 0.2704
#
×
(0.9249)
ˆ
σ
1
0.0000
0.0000 (0.7289)
ˆ
σ
2
(44)
Analogous comments to (42) hold here for the exis-
tence of Lagrange solutions.
If (41), (42), (43) and (44) are not Lagrange, (36)
and (37) are set to different values; in general to
smaller values, and the computations are carried out
all over again.
5 FINAL REMARKS
This paper was concerned with the design of a feed-
back controller such that the overall system is asymp-
totically stable. The key point was the imposition of
a very speciﬁc Lagrange solution to then ﬁnd out the
condition for the existence of such a system.
The results showed that a solution can be found if
the controller can diagonalize the matrix blocks of the
matrices composing the model, which is represented
by the set of partial difference equations. A numerical
example was presented to show how the mechanics of
the calculations work..
REFERENCES
Anderson, B. D. O. and Jury, E. I. (1974). Stability of multi-
dimensional recursive ﬁlters. In IEEE Trans. Circuits
and Syst.Proc. IEEE, number 21, pages 300–304.
Attasi, S. (1973). Systemes lineaires homogenes a deux
indices. In Rapport Laboria, number 31.
Bose, N. K. (1982). Applied multdimensional systems the-
ory. England: Van Nostradand Reinhold Co.
Cheng, S. S. (2003). Partial difference equations. London:
Taylor & Francis.
Du, C. and Xie, L. (2002). H-inﬁnity control and ﬁltering of
two-dimensional systems. Berlin, Germany: Springer
Verlag.
Elaydi, S. (2005). An introduction to difference equations.
USA: Springer Science.
Fornasini, E. and Marchesini, G. (1978). Doubly indexed
dynamical systems: State space models and structural
properties. In Math. Syst. Th., number 12, pages 59–
72.
Fornasini, E. and Marchesini, G. (1980). Stability analysis
of 2-d systems. In IEEE Trans. Circ. and Syst., volume
CAS, pages 1210–1217.
Givone, D. D. and Roesser, R. P. (1972). Multidimensional
linear iterative circuits - general properties. In IEEE
Trans. Comp., volume C, pages 1067–1073.
Izuta, G. (2007). Stability and disturbance attenuation of 2-
d discrete delayed systems via memory state feedback
controller. Int. J. Gen. Systems, 36(3):263–280.
Izuta, G. (2010). Stability analysis of 2-d discrete systems
on the basis of lagrange solutions and doubly similar-
ity transformed systems. In Proc. 35th annual conf.
IEEE IES, Porto, Portugal.
Izuta, G. (2012). Networked 2-d linear discrete feedback
control systems design on the basis of observer con-
troller. In Proc. 16th Int. Conf. on System Theory,
Control and Computing IEEE CS, Sinaia, Romania.
Izuta, G. (2014a). Controller design for 2-d discrete linear
networked systems. In Proc. The Int. Electrical Eng.
Congress, Pataya, Thailand.
Izuta, G. (2014b). On the asymptotic stability analysis of
a certain type of discrete-time 3-d linear systems. In
Existence Conditions of Asymptotically Stable 2-D Feedback Control Systems on the Basis of Block Matrix Diagonalization
469 Proc. 11th Int. Conf. on Informatics in Control, Au-
tomation and Robotics (ICINCO 2014), Vienna, Aus-
tria.
Jerri, A. J. (1996). Linear difference equations with discrete
transform methods. Netherlands: Kluwer Acad. Pub.
Kaczorek, T. (1985). Two-dimensional linear systems.
Berlin: Springer Verlag.
Lim, J. S. (1990). Two-dimensional linear signal and image
processing. New Jersey, USA: Prentice Hall.
Pazke, W., Lam, J., Galkowski, K., and Xu, S. (2004).
Robust stability and stabilisation of 2d discrete state-
delayed systems. In Systems and Control, volume 51,
pages 277–291.
Piekarski, M. S. (1977). Algebraic characterization of ma-
trices whose multivariable characteristic polynomial
is hurwitzian. In Proc. Int. Symp. Operator Theory,
pages 121–126.
Rogers, E., Krzysztof, G., and Owens, D. H. (2007).
Control Systems Theory and Applications for Linear
Repetitive Processes. Springer, London.
Tzafestas, S. G. (1986). Multidimensional systems - Tech-
niques and Applications. New York: Marcel Dekker.
Zerz, E. (2000). Topics in Multidimensional Linear Systems
Theory. Springer.
ICINCO 2016 - 13th International Conference on Informatics in Control, Automation and Robotics
470 