Вы находитесь на странице: 1из 1

IEEE TRANSACTIONSON AUTOMATICCONTROL,VOL. 38, NO.

5, MAY 1993

830

Assume (4.1) is also true for n = k.


Consider the case that n = k 1.
Since
A[,+11 = [ U i j .Awl],

The algorithm given in the Proposition 2.15 of the paper to


compute the maximal set of indistinguishable state pairs (1,) is
as follows:

Let Z,

the ( p , q)th entry of the Kronecker power A[,+ 1 is found in the


( p o l , q o 1)st block, i.e., uPoloA[k]
and is located at the
( p 1 p 2 ...
. p n - + 1, qlq2 ... qn- + 1)st entry of the block.
By using the assumption for n = k, the ( p , q)th entry of the
Kronecker power is equal to

uPOqOuP1~l

* * * Pkqk

Equation (4.1) is also true for n = k + 1.


Equation (4.1) is a way to zoom into the general entry xtq.
The first factor of x p q , upoqo,is contributed from the submatm,
ACn-l],containing xpq. The second factor of xpq is supplied
uPo90
from the submatrix U , ~ , ! A [ -which
~ ~ , is a submatrix of the
previous submatrix, containing xpq. This process of constructing
xpq goes on until the last submatrix (the nth submatrix, which is
equal to the matrix A ) containing xpq has been reached.
Example: Find the (30,60)th entry of the 10th Kronecker
power of the matrix

Solution: Convert 29 (= 30
numbers.

zk+

=Y

x Y,and iterate

{(x, y ) E z k I fop ( ( x , y ) , y )

Terminate when

= 1,. Then

Z,,

n 1, # 4

for some

yl.

= zk.

However, the above algorithm is incorrect because for any alive


plant A with A I being D-stable, the above algorithm always
yields ,Z = Z,. This can be easily verified by applying the
algorithm to the example automaton in Fig. 6 of the paper
where the state pairs (1,2) and (2, 1) fall in Z, even though they
are distinguishable pairs.
The problem arises because 0, and the algorithm in Proposition 2.15 both consider events y which may not be defined from
both x and y in an element ( x , y ) of 0, or 1,. To remove the
difficulty, one may modify either the definition of 0, or the
algorithm itself so that only events y E d ( x )n d ( y ) are considered. However, the former modification may preclude the use
of 0, in testing observability. Therefore, we modify the algorithm as follows:
Modified Algorithm:

- 1) and 59 (= 60 - 1) into binary

29 = OOOOO11101 (base 2).


59 = oooO111011 (base 2).
The (30,60)th entry is equal to

To show that this algorithm works correctly let the algorithm


terminate at some k = N with Z = IN+= Z,. Since Z, 2 ZM
and any ( x , y ) E Zk - Zk+l cannot be an indistinguishable pair
according to Lemma 2.14, I, 2 ,Z Vk. Hence, Z 2 Z., Also, it is
obvious from the iterative step and the termination condition of
the algorithm that all pairs in Z are indistinguishable. Hence,

u&uO1u~lulOuO1ull~

which is simplified to
a&&a10a113.

z z,.

&FERENCES

[l] Chiu H. Choi, Time-varying Riccati differential equations with


known analytic solutions, IEEE Trans. Automat. Contr., vol. 37,
no. 5, pp. 642-645, May 1992.
[2] Chiu H. Choi and A. J. h u b , Efficient matrix-valued algorithms
for solving stiff Riccati differential equations, IEEE Trans.
Automat. Contr.,vol. 35, no. 7, pp. 770-776, July 1990.

Comments on Conditions for Stable Zeros of


Sampled Systems
Per Hagander
Abslraet-The

Comments on Observability of Discrete Event


Dynamic Systems
Supratik Bose, Amit Patra, and Siddhartha
Mukhopadhyay
Ab--This
note points out that the algorithm introduced in the
paper to compute the maximal set of indistinguishable state pairs is
incorrect under the given definition of the automaton 0, and suggests a
correction.

Manuscript received October 25, 1991; revised September 17, 1992.


The authors are with Department of Electrical Engineering, Indian
Institute of Technology, Kharagpur-721302India.
fEEE Log Number 9208718.
C. M. Ozveren and A. S. Willsky, Observability of discrete event
dynamic systems, IEEE Trans. Automat. Contr.,vol. 35, pp. 797-806,
July 1990.

result of the above paper is incorrect.

COUNTEREXAMPLE
When sampling a continuous-time system G ( s ) using zeroorder hold the zeros of the resulting discrete-time system H(z)
become complicated functions of the sampling interval T. The
basic facts were collected in [l], while subsequent papers have
tried to formulate sharper conditions for stable zeros, e.g., [2].

Manuscript received Februaly 10, 1993.


The author is with the Department of Automatic Control, Lurid
Institute of Technology, P.O. Box 118, S-221 00 Lund, Sweden.
fEEE Log Number 9209268.
M. Ishitobi, Conditions for stable zeros of sampled systems, IEEE
Trans.Automat. Contr., vol. 37, no. 10, pp. 1558-1561, 1992.

0018-9286/93$03.00 0 1993 IEEE

Вам также может понравиться