14
REPRINT DA "PROC. 6TH INT, CONF, ON ANALYSIS AND OPT. OF SYSTEMS/ NI ZZA/ 1934, LECT. NOTES IN CONTROL AND INF, SCI., VOL. 52" INFINITI:; ZERO MODULE AND INFINITE POLE MODULE G. Conte Ist. Mat. Univ. Genova via L.B.Alberti 4 16132 Genova - ITALY A. Perdon Ist. Mat. Appi. Univ. Padova via Belzon.i 7 35100 Padova - ITALY SUMMARY In this paper we introduce the notion of infinite zero module Z (G) and infinite oo pole module P (G) associated with a transfer function G(z). We show that Z (G) and co co P (G) describe the zero/pole structure at infinity of G(z) and we investigate their dynamical and system theoretic properties. Finally, we apply these concepts to the study of the inverses of G(z). INTP.ODUCTION Let G(z) denote a rational transfer function matrix of dimensions pxm. In this paper we introduce two abstract algebraic òbjects, callcd respectively "infinite zero module" and "infinite pole module" and denoted by Z (G) and P (G), which describe the zero/pole structure at infinity of G(z). More brecisely, Z (G) and P (G) are finitely CU PO generated torsion K |f z |-modules (and hence finite dimensionai K-vector spaces) whose definition i n tlerms of G(z) is based on a dynamical characterization of zeros and po- les at infinity. Moreover, when S(z) •- diag {z z r ) is the non trìvial part of the Smith-MacMillan form at infinity of G(z), the following representations hold : 2 (G) ~ * K|lz -1 U/ z^iKlIz™1! and P (G) - • K|| a" 1/ z""Vi d z~ 1 11. co V ^<0 Vj L >0 Both Z (tì) and P (G) can be described using special representations of G(z) : Au -* AY co oo -1 of the form G(z) = T (z)V(z), where T(z) and V(z) are matricès with entries from -1 , -1„ K|z [|, coprirne in the appropriate sense. In faet, dcnoting by \i U and u Y the Kg z [|- -1 m p modules of series in z with coefficients from U - K and Y = K respectively, we have that Z (G) is isomorphic to the torsion submodule of f< Y/V(z)fì U and that P (G) co OO 00 00 is isomorphic to Q Y/T(z) Q Y. QO CO When G(z) is strictly proper, the connection between Z (G) and the geometrie defini-

Infinite zero module and infinite pole module

  • Upload
    univpm

  • View
    3

  • Download
    0

Embed Size (px)

Citation preview

R E P R I N T DA " P R O C . 6 T H I N T , C O N F , ON A N A L Y S I S A N D O P T . OF S Y S T E M S /

NI ZZA/ 1 9 3 4 , L E C T . N O T E S IN C O N T R O L A N D INF, S C I . , V O L . 5 2 "

INFINITI:; ZERO MODULE AND INFINITE POLE MODULE

G. Conte

I s t . Mat. U n i v . Genova v i a L . B . A l b e r t i 4

16132 Genova - ITALY

A. Perdon

I s t . Mat. A p p i . Univ. Padova v i a Belzon.i 7

35100 Padova - ITALY

SUMMARY

I n t h i s paper we i n t r o d u c e t h e n o t i o n o f i n f i n i t e z e r o module Z (G) and i n f i n i t e oo

p o l e module P (G) a s s o c i a t e d w i t h a t r a n s f e r f u n c t i o n G ( z ) . We show t h a t Z (G) and co co

P (G) d e s c r i b e t h e z e r o / p o l e s t r u c t u r e a t i n f i n i t y o f G(z) and we i n v e s t i g a t e t h e i r

d y n a m i c a l and system t h e o r e t i c p r o p e r t i e s . F i n a l l y , we a p p l y these concepts t o the

s t u d y o f the i n v e r s e s o f G ( z ) .

INTP.ODUCTION

Let G(z) denote a r a t i o n a l t r a n s f e r f u n c t i o n m a t r i x of dimensions p x m . I n t h i s

paper we i n t r o d u c e two a b s t r a c t a l g e b r a i c òbjects, c a l l c d r e s p e c t i v e l y " i n f i n i t e z e r o

module" and " i n f i n i t e p o l e module" and denoted by Z (G) and P (G), which d e s c r i b e t h e

z e r o / p o l e s t r u c t u r e a t i n f i n i t y o f G ( z ) . More b r e c i s e l y , Z (G) and P (G) ar e f i n i t e l y CU PO

g e n e r a t e d t o r s i o n K|f z |-modules (and hence f i n i t e d i m e n s i o n a i K-vector spaces) whose

d e f i n i t i o n i n tlerms o f G(z) i s based on a dynamical c h a r a c t e r i z a t i o n o f zeros and po-

le s a t i n f i n i t y . Moreover, when S(z) •- d i a g {z z r ) i s the non trìvial p a r t

o f t h e S m i t h - M a c M i l l a n form a t i n f i n i t y o f G ( z ) , the f o l l o w i n g r e p r e s e n t a t i o n s h o l d :

2 (G) ~ * K|lz - 1 U/ z^iKlIz™1! and P (G) - • K|| a" 1/ z""Vi d z~111. co V^<0 Vj L>0

Both Z (tì) and P (G) can be d e s c r i b e d u s i n g s p e c i a l r e p r e s e n t a t i o n s o f G(z) : Au -* AY co oo -1 o f t h e form G(z) = T ( z ) V ( z ) , where T ( z ) and V(z) are matricès w i t h e n t r i e s from -1 , -1„ K|z [|, coprirne i n t h e a p p r o p r i a t e sense. I n f a e t , d c n o t i n g by \i U and u Y th e Kg z [|-

-1 m p modules o f s e r i e s i n z w i t h c o e f f i c i e n t s from U - K and Y = K r e s p e c t i v e l y , we

have t h a t Z (G) i s i s o m o r p h i c t o the t o r s i o n submodule o f f< Y/V(z)fì U and t h a t P (G) co OO 00 00

i s i s o m o r p h i c t o Q Y/T(z)Q Y. QO CO

When G(z) i s s t r i c t l y p r o p e r , the c o n n e c t i o n between Z (G) and the g e o m e t r i e d e f i n i -

t i o n o f zero s t r u c t u r e a t i n f i n i t y i s g i v e n by t h e e x i s t e n c e o f a K|| z I-isomorphism

f : S*/R* > Z (G). Here, assuming t h a t G(z) has the c a n o n i c a l r e a l i z a t i o n (X,A,B,C),

S* denotes, as u s u a i , t h e minimum c o n d i t i o n a l l y i n v a r i a n t subspace o f X c o n t a i n i n g

Im B and R* denotes t h e maximum r e a c h a b i l i t y subspace o f X c o n t a i ned i n Ker C. The

q u o t i e n t S*/R* i s endowed w i t h a n a t u r a i module s t r u c t u r e i nduced from those o f AU

and AY.

Since t h e zeros a t i n f i n i t y p l a y a fundamental r o l e i n many c o n t r o l problems

(such as : i n v e r s i o n o f l i n e a r systems, c a u s a i f a c t o r i z a t i o n , feedback e q u i v a l e n c e ) ,

the main m o t i v a t i o n t o i n t r o d u c e Z (G) and P (G) i s the i n t e r e s t i n h a v i n g a n a t u r a i CO co

a l g e b r a i c t o o l i n o r d e r t o handle such z e r o s .

Here we a p p l y t h e n o t i o n s o f i n f i n i t e z e r o module and i n f i n i t e p o l e module t o

the s t u d y o f t h e i n v e r s e s o f G ( z ) . We o b t a i n , f i r s t o f a l i , t h a t g i v e n a ( r i g h t o r

l e f t ) i n v e r s e H{•) o f G(z) t h e i n f i n i t e z e r o module Z (G) i s contaìned, i n an appro-CO o r i a t e sense, i n t h e i n f i n i t e p o l e module P (H) o f H ( z ) . The analogous r e s u l t , COn-

CO

c e r n i n g t h e ( f i n i t e ) z e r o module Z(G) and t h e ( f i n i t e ) p o l e module, i . e . t h e c a n o n i -

c a i s t a t e space, X(H) o f H(z) was proved by B.Wyman and M.Sain i n [ 8 ] ,

Then, t h e f a c t t h a t M a c M i l l a n degree H(z) = dim X(H) + dim P (H), enables us t o s t a t e

K K « the f o l l o w i n g r e s u l t : f o r any i n v e r s e H(z) o f G ( z ) , M a c M i l l a n degree H(Z) > dim Z(G)

K + dim Z (Gii Moreover, r e c a l l i n q t h a t g e n e r a l i z e d o r d . H(z) = dim X(H) + dim X (H) K oo K K « (see [ 2 | , [ 6 J ) , t h e e x i s t e n c e o f a c a n o n i c a ! p r o j e c t i o n <J> : X (H) -> P (H) and i t a

CO «XJ s p e c i a l p r o p e r t i e s a l l o w us t o s t a t e the f o l l o w i n g r e s u l t : f o r any i n v e r s e H(z) o f G( z ) , a e n e r a l i z e d o r d . H(z) > dim Z (G) + dim '/. (G) + (number o f c y c l i c submodules i n

K K °° d i r e c t sum d e c o m p o s i t i o n o f Z ( G ) ) .

co

I t i s p o s s i b l e t o - f i n d exampìes i n which t h e lower bounds d e s c r i b e d above cannot be

reached. Howevep:, f u r t h e r i n v e s t i g a t i o n s on t h i s l i n e seem a b l e t o g i v e r e s u l t s con-

c e r n i n g a c o n s t r u c t i v e c h a r a c t e r i z a t i o n o f the m i n i m a l i n v e r s e o f a g i v e n G ( z ) .

l.PRELIMINARIES AND NOTATIONS

ven a f i e l d K and a f i n i t e d i m e n s i o n a i K-vector spao- S AS denotes the s e t -1

o f ( f o r m a i ) L a u r e n t s e r i e s i n z w i t h c o e f f i c i e n t s i n S, i . e . s e r i e s o f the forni co — t

s = E s z , s e S and t è Z- The o r d e r o f s i s d e f i n e d as t h e ind e x o f t h e f i r s t non-t= k t t zero c o e f f i c i e n t s . A s e r i e s s w i l l be s a i d o r o o e r i f f i t s o r d e r i s 0 and s t r i c t l y t p r o p e r i f f i t s o r d e r i s g r e a t e r than 0. ^

- t

fiS denoted t h e p o l y n o m i a l subset o f AS, i . e . t h e s e t o f elements o f the form E s z ,

and fi S denotes the power s e r i e s subset o f AS, i . e . the s e t o f elements o f the form 00

^J S Z .. 0 t

AK i s a f i e l d and AS i s a AK-vector space; (e . e ..., e } denotes b o t h t h e ca-1 2 r

n o n i c a l b a s i s o f S over K and t h a t o f AS over AK. The f i e l d AK c o n t a i n s t h e r i n g

o f p o l y n o m i a l s K[ Z] , t h e r i n g o f power s e r i e s K 11 z Ij and the f i e l d o f f r a c t i o n s

K(z) . As a consequence A S i s , i n p a r t i c u l a r , a KÌ zi -module and, a l s o , a K (l z \ -module.

W i t h r e s p e c t t o these s t r u c t u r e s fiS t u r n s o u t t o be a k| z]-submodule and fi S t u r n s , OD

o u t t o be a K 1 z il -submodule. I n the f o l l o w i n g we w i l l c o n s i d e r t h e q u o t i e n t modules:

PS := A S/fiS Kl zi-module

I'*S :=AS/fi S and F S:= AS/z *Ù S K || z | -modules. CO CO do CO

The elements o f TS cari be u n i q u e l y r e p r e s e n t e d as s t r i c t l y p r o p e r s e r i e s , t h e p r o d u c t

by z b e i n g the u s u a i p r o d u c t by z i n As f o l l o w e d by t r u n c a t i o n o f the p o l y n o m i a l p a r t .

The elements o f V S ( f * S ) can be uni q u e l v r e o r e s e n t o d as p o l y n o m i a l s ( p o l y n o m i a l s co co

w i t h o u t Constant t e r m ) . The p r o d u c t by z i s the visual p r o d u c t i n AS f o l l o w e d by

t r u n c a t i o n o f t h e s t r i c t l y p r o p e r p a r t ( r e s p . : the p r o p e r p a r t ) .

L e t u and V .be K-vector spaces o f dimension m and p r e s p e c t i v e l y . I n the f o l l o w i n g

no d i s t i n c t i o n w i l l be made b e t w e e n A K - l i n e a r maps between AU and AY and pxm m a t r i c e s

a s s o c i a t e d t o them w i t h r e s p e c t t o the c a n o n i c a l b a s i s . -1

Any p xm m a t r i x w i t h e n t r i e s i n AK can be seen as a L a u r e n t s e r i e s i n z w i t h m a t r i x oo - t

c o e f f i c i e n t s , i . e . G(z) = l G z , G p x m m a t r i x w i t h e n t r i e s i n K, G(z) w i l l be s a i d k t t

p o l y n o m i a l , s t r i c t l y p r o p e r , p r o p e r i f f a l i i t s e n t r i e s are such and i t w i l l be s a i d

r a t i o n a l i f f a l i i t s e n t r i e s b e l o n g t o K ( z ) ,

By a t r a n s f e r f u n c t i o n we mean a A K-linear r a t i o n a l . map G(z) between AU and AY.

Any t r a n s f e r f u n c t i o n G(z) g i v e s r i s e t o the f o l l o w i n g commutative diagrams.- 1.1,

where a l i modules are K[ z)-modules and a l i maps are Kj zj-homomorphisms, and 1.2, where -1 -1 a l i modules are K|| z [1-modules and a l i maps are K|| Z ||-homomorphisms.

AU • AY j- | TI c a n o n i c a l p r o j e c t i o n

G , 0 0 - t 1.1 fiU TY p i ( pfcz ) = y }

i , i c a n o n i c a l i n c l u s i o n U x Y

B • c

X =Im Ga~ fiU/Ker G** i s a f.g. t o r s i o n K( Z] -module. De f i n i n g A: X -*• X as Ax = zx,

(X, A, B, C) i s a minimal r e a l i z a t i o n of the s t r i c t l y proper p a r t of G(z) (see(71)

G AU • AY f i ' , j ' c a n o n i c a l i n c l u s i o n i ' + ir' 1 G 1. 2 iì U p ' • F Y ir ' c a n o n i c a l p r o j e c t i o n

9 - t . j ' ! ^ B ' ^ ^ C- i p Q p Q ( l y t z ") = - y Q

Y B ' » C'

X - Im G - fi U/Ker G i s a f.g. t o r s i o n K|Z 1-module; d e f i n i n g A': X -*• X as c x > p o o * p - " o o c o

A'x = z x, (X ,A', B',C) i s a minimal g e n e r a l i z e d s t a t e space r e a l i z a t i o n of the oo

polynomial p a r t of G(z) ( s e e [ 2 ] ) .

L e t H(co) denote the Smith MacMillan form of G(l/to) and l e t p (co) = to J q (co) .with i i

q, (0) ^ 0 f o r i = 1 , . . , r=rank G, be the non-zero(diagonal) elements i n H(co) . Then,

the Smith-MacMillan form a t i n f i n i t y of G(z) i s the p x m matrix

'S(z) i

0 where S(z) = diag {z .., z V ) r ) v < ...< v

1— — r

The r-uple { v, ,.., v } i s c a l l e d s t r u c t u r e at i n f i n i t y of G(z) (see t s l 6.5). I r — """""" ~* 1 —

The Smith-MacMillan form a t i n f i n i t y can be obtained a l s o by the f o l l o w i n g procedure.

Write G(z) - z G(z) where s i s the minimum int e g e r such t h a t G(z) i s proper and l e t

M(z) be the Smith form of G(z) with r e s p e c t to the r i n g KIZ 1 1 As KIZ 1 I i s a l o c a i -1

r i n g whose maximal i d e a l i s generated by z , the non-zero (diagonal) elements of -1 s

M(z) can be assumed to be powers of z . Then z M(z) i s the Smith-MacMillan form a t s

i n f i n i t y of G(z) . I n p a r t i c u l a r G(z) * h^(z)z M(z)B (z) where B ^ z ) , BgU) are b i c a u s a l

matrices ( i . e . i n v e r t i t i l e i n the r i n g of proper m a t r i c e s ) .

G(z) i s s a i d to have a pole a t as of order -y for any negative v i h i t s s t r u c t u r e i 1

at <*>, and, analogously, i t i s s a i d to have a zero a t w of order w f o r any p o s i t i v e v i n i t s s t r u c t u r e a t °°. The t o t a l number of poles (zeros) a t » i s then E „ (-v.) j • v, <0 i ( Z n v.) . v, >0 ì

Supnose now t h a t v. < 0 f o r i = l , . . , k and that v > 0 for i = k + l , . . , r . Denotine i — ì

by S = diag{ z 1 , z k } and by S = diag {z z r } , a coprirne f a e t o n z i -

t i o n of the Smith-MacMillan forra a t °° of G(z) by proper m a t r i c e s i s given by

S(7) 1 0

0 ] 0

-1 -1 e (z) \p (z) - 4> (z) e (z) R L where

0 1 \ 1 0

0 S~ 1

0 1 0 i

, + I

s 0 * («) « I 0 I , 1 R I r-k 1

1 I 1 m-r

'S 0

di (a). «1 0 : I r-k

p-r/

Coprirne f a c t o r i z a t i o n s o f G(z) by proper matrices are then given by

G(z) = [B, ( z ) e ( z ) ] [ B* (Z)>|> ( Z ) ] _ 1

1 ^ Z R

G(z) = U ( z ) B ~ ( z ) ] ~ [ e(z)B •(*)] 2 -1 -1

I n the fo l l o w i n g , f o r any coprirne f a c t o r i z a t i o n G(z) = V ( Z ) T r (z) or G(z) = ( z ) V ( z )

by proper m a t r i c e s , V(z) w i l l be c a l l e d proper numerator of G(z) and T ( z ) , T (z) w i l l R L

be c a l l e d proper denominatorsof G ( z ) . I t can be proved as i n [5] t h a t E ( Z ) i s the

Smith form,with r e s p e c t to Kfz | , of every proper numerator of G(z) and t h a t ^ (z)

(resp. ( 2 ) ) has the sanie n o n t r i v i a l i n v a r i a n t f a c t o r s of any proper denominator of Li

G(z) .

2. INFINITE ZERO MODULE

The aim of t h i s s e c t i o n i s to define the modula of i n f i n i t e zeros of a t r a n s f e r

f u n c t i o n G ( z ) . I t s r e l a t i o n s with the classìcal notion of zeros a t i n f i n i t y and i t s

system t h e o r e t i c i n t e r p r e t a t i o n are i n v e s t i g a t e d .

DEFINITION 2.1 Given a t r a n s f e r function G(z) i t s i n f i n i t e zero module Z (G) i s d e f i -

ned by : •1 (fi Y) + n U Z (G) =

Ker G + fi U co

To motivate the d e f i n i t i o n given above, l e t us consider the case ra = p = 1. L e t

u(z) be an element i n AK and l e t k be i t s order, then, i f k < 0, u(z) i s s a i d to have

-k modes a t i n f i n i t y . C e r t a i n of these modes raay f a i l to appear i n the response of

the system y ( z ) = G ( z ) u ( z ) , i . e . ord y(2) =^k > k , and t h i s f a c t i s i n t e r p r e t e d as

thè presence of zeros a t i n f i n i t y i n G ( z ) . So i n d e f i n i n g the a b s t r a c t module we con­

s i d e r e x c i t a t i o n s which produce response having no modes a t i n f i n i t y , and we ignore

both proper inputs (which have no modes a t i n f i n i t y whose absence can be detected i n

the output) and Ker G ( s i n c e i d e n t i c a l l y zero outputs are of l i t t l e i n t e r e s t ) .

-1 PROPOSITION 2.2 Z (G) i s a f i n i t e l y generated t o r s i o n KÌz Smodule

• - •-— — — CO

..V

1 Proof. G(z) i s Kflz ! - l i n e a r and 0 Y i s a f i n i t e l y generated K| z 1-module, then

co G * (Q Y ) / k e r G and Z (G) , which can be viewed as a quotient of the previous one,

co co are f i n i t e l y generated KJ.Z J-modules. Everv element of Z (G) i s the equivalence c l a s s [ u] modulo Ker G + il U o f some

co co

u e AU such t h a t G ( z ) u i s proper. Let k be the degree of the polynomial p a r t of u; -k -k -k then G ( z ) ( z u) i s proper and z u i s proper. Therefore z [ u] = 0 and Z (G) i s

co

t o r s i o n .

PROPOSITION 2.3 Z (G) i s isomorphic to the t o r s i o n submodule of fi Y/ e ( z ) f i U, where • co co co

e(z) i s the Smith form of any proper numerator of G ( z ) .

Proof. We prove the P r o p o s i t i o n showing t h a t Z (G) i s isomorphic to the t o r s i o n

submodule of fi Y/V(z)fi U where G(z) = T ( z ) V ( z ) i s a coprirne proper f a c t o r i z a t i o n . co co

At t h i s aim, we r e p r e s e n t the elements of Z (G) as f u] , for some u e AU such t h a t co

G ( z ) u i s proper. I n p a r t i c u l a r , V ( z ) u i s proper, i . e . V ( z ) u <= fi Y. Moreover, i f CO

fu] • [ u'1 , i . e . (u - u 1) e (Ker G + fi U), we have, s i n c e Ker G(z) • Ker V ( z ) , 1 co

V(z) (u - u 1 ) B V(z)fi ììi hs a consequenee, we can define a K [ [ Z 1-homotnorphism co

f i Z (G) -* fi Y/V(z)fi U by f ( [ u ] ) * [ V ( z ) u ] ( c l a s s of V ( z ) u i n fi Y/V(z)fi U) . 00 OO CO 00 oo

As Z (G) i s t o r r i o n , f ( Z (G)) i s contained i n the t o r s i o n submodule of Q Y/V(z)fl U, eo co co 00

Suppose t h a t f ( [ u ] ) « [ V ( z ) u ] = Os then V ( z ) u e V(«5)flJU and u <3 Ker V + QJJ =

* Ker G + fi B. Therefore ( u] * 0 and f i s i n j e c t i v e . Co

L e t f y ] be a t o r s i o n element i n fi Y / v ( z ) f i U. By coprimnessof T(z) and V ( z ) , there 1 * J OO CO

e x i s t proper m a t r i c e s A(z) and B(z) such t h a t y - T ( z ) A ( z ) y + V ( z ) B ( z ) y . Then [y] m **k

* [ T(as) A ( z ) y ] and there e x i s t s a p o s i t i v e i n t e g e r k such t h a t z T ( z ) A ( z ) y « V(z) u k k f o r a s u i t a b l e u i n fi U. L e t v = z u t B ( z ) y e AO? then V ( z ) v = V ( z ) z u + V ( z ) B ( z ) y

do k -1 k = T ( z ) A ( z ) y + V ( z ) B ( z ) y " y. Now, z u belongs to G (fi Y) because G ( z ) z u = 00

= T _ 1 ( z ) V ( z ) z k u = T " 1 ( z ) V ( z ) ( v - B ( z ) y ) = T _ 1 ( z ) (y - V ( z ) B ( z ) y ) - T~ ( Z ) ( T ( z ) A ( z ) y )

= A ( z ) y e fi Y. Then [ y ] = [ V ( z ) v ] = f ( [ u ] ) , and f i s onto the t o r s i o n submodule of co

fi Y/V(z)fi U. OO 00

The b a s i c property of Z (G), i n connection with the notion of zero a t i n f i n i t y we r e -

c a l l e d i n the previous s e c t i o n , i s pointed out by the following C o r o l l a r y .

COROLLARE 2.4 The i n v a r i a n t fàctors of Z (G) over Kfl2 1 | c o i n c i d e with the non t r i -— — — — — — ^ ^ — co

v i a l eleraents of S .

REMARK 2.5 The above C o r o l l a r y says t h a t Z (G) contains a l i the information about the 00

zero s t r u c t u r e a t i n f i n i t y of G ( z ) . More p r e c i s e l y , i f {v , ,v^} i s the s t r u c t u r e

a t i n f i n i t y of G ( z ) , the i n f i n i t e zero module has the following c a n o n i c a l decomposi-

t i c n i n t o a d i r e c t sum of c y c l i c submodules : Z (G) = © K|Z * J / z V ; IK|z J . co v <0

i REMARK 2.6 The K|z *]]-module G ^ (fi Y) which appears i n the d e f i n i t i o n of Z (G) i s the

" co oo

l a t e n c y k e r n e l of G(z) introduced i n [ 4 ] . We w i l l i n v e s t i g a t e more deeply i t s r e l a -

t i o n s with the s t r u c t u r e at i n f i n i t y i n the following.

3. GEOMETRIC CHARACTERIZATION

I n t h i s s e c t i o n we assume t h a t G(z) i s a p xm s t r i c t l y proper t r a n s f e r function

provided with the minimal r e a l i z a t i o n (X,A,B,C). Moreover, we assume, without l o s t of

g e n e r a l i t y , t h a t X = Im G** and, as a consequenee, t h a t C* i s the i n c l u s i o n (see 1.1). ,

I t i s known ([ 1 ] ) that the zero s t r u c t u r e a t i n f i n i t y of G(z) can be obtained

from the quotient S*/R*, where S* i s the minimum ( A , C ) - i n v a r i a n t subspace of X con-

t a i n i n g Im B, and R* i s the maximum c o n t r o l l a b i l i t y subspace of X contained i n Ker C.

I n the f o l l o w i n g wè w i l l c h a r a c t e r i z e S* and R* i n terms of the t r a n s f e r function G(z)

and then we w i l l / prove t h a t Z (G) i s Kj z ^ - i s o m o r p h i c to S*/R*. co

PROPOSITION 3.1 Define S = { s e AY, s i s s t r i c t l y proper and s = G ( z ) u f o r some

u e fiU } = z ^fi Y fi G(fiU) and R = { s e AY, there e x i s t u <= fiU and 'S s t r i c t l y proper co

such t h a t G(z).u = G ( z ) u * s } = G(z fi U) O G(fiU) . Then S = S* and R = R* «

Proof. We remark, f i r s t of a l i , that both S and R are contained i n Im G f l = X,

Since G(z) i s s t r i c t l y proper, we have Im B C S. To prove the (A,C)-invariance of

S we show t h a t A(S fi Ker C) C S. Any element of S, i n f a c t , i s of the form s =

= s ^ " 1 + = G(z)u, with u e fiU. As C(s) = p ^ i s ) = s ^ s e Ker C i f f s i = 0.

For such an element s, A(s) = zG(z)u = G ( z ) z u = ( s 2 z 1 + ) belongs c l e a r l y

to S.

The m'-.imality of S among the ( A , C ) - i n v a r i a n t subspaces containing Im B w i l l be

p r o v e d by c o n t r a d i c t i o n . Suppose t h a t V i s an ( A , C ) - i n v a r i a n t subspace of X con-

t a i n i n g Im B but not co n t a i n i n g S, i . e . G ( z ) u e V for every C o n s t a n t u and there

e x i s t polynomials u(z) such t h a t G ( z ) u ( z ) i s s t r i c t l y proper b u t G ( z ) u ( z ) does not

belong to V. L e t p( z ) be such a polynomial of minimum degree : deg p(z) > 1, as

G(z) u e V f o r every C o n s t a n t u. Therefore, we have p(z) = z q ( z ) + r , with r e U -1

and G ( z ) p ( z ) = zG(z)q(Z) + G ( z ) r . Now, z G ( z ) q ( z ) = G ( z ) p ( z ) - G ( z ) r = y Z + -2 -2 + y z + = z ( y ^ z + ) and, s i n c e deg q ( z ) < deg p(z) , G ( z ) q ( z ) = , -2

= y z + i s an element of V H Ker C. By the (A,C)-invariance of V,

AG(z)q(z) = z G ( z ) q ( z ) = v e V. Thus G ( z ) p ( z ) = v - G ( z ) r belongs to V a g a i n s t the

hypothesis.

R = R* i s proved i n [ 3 ] §'4.

REMARK 3.2 S*/R* has a n a t u r a i K[z ^-module s t r u c t u r e d e f i n i e d as fo l l o w s . L e t [ s] -1. ,

denote an element i n S*/R*, s = G ( z ) u ( z ) where u ( z ) = zu(z) + U q e fiU. Then z [ sj =

[ G ( z ) u ( z ) ] . D e f i n i t i o n i s c o n s i s t e n t , i n f a c t G ( z ) u ( z ) i s s t r i c t l y proper and hence

[ G ( z ) u ( z ) ] e S*/R», moreover i f [ s] • [ s 1 ] and s ! • G ( z ) v ( z ) , v(z) = z v ( z ) + V q , we

have G ( z ) u ( z ) - G ( z ) v ( z ) e R*, i . e . G ( z ) u ( z ) - G ( z ) v ( z ) • G(z)w(z) with w(z) s t r i c t l y -1 -1

proper. As a consequenee G ( z ) u ( z ) - G ( z ) v ( z ) = G ( z ) ( z w(z)) + G ( z ) ( z ( V q - u Q ) ) e

e R* and [ G ( z ) u ( z ) ] = [ G ( z ) v ( z ) ] .

PROPOSITION 3.3 Z (G) and S#/R* are isomorphic as K([Z ^-modules. oo

Proof. As G(z) i s s t r i c t l y p r o p e r , Z (G) = G (fi Y) / (Ker G + fi U). L e t [ s ] , s = oo co oo

• G ( z ) u ( z ) , be an element of S*/R*. Then s i s , i n p a r t i c u l a r , s t r i c t l y proper and

zu(z) e G _ 1 ( f i Y) . We define f : S»/R* * Z (G) as follows : f ({ s] ) m [ z u ( z ) ] ( c l a s s co co

of zu(z) i n Z ( G ) ) . D e f i n i t i o n i s c o n s i s t e n t , i n f a c t , i f [ s] = [ a ' \ s' = G ( z ) v ( z ) co

then G ( z ) ( u ( z ) - v ( z ) ) £ R*, i . e . G ( z ) ( u ( z ) - v ( z ) ) « G(z)w(z) with w(z) s t r i c t l y

proper. As a consequenee zu(z) - z v ( z ) = zw(z) + p ( z ) , with p(z) e Ker G, and

f z u ( z ) ] = [ z v ( z ) ] i n Z (G). f i s c l e a r l y K - l i n e a r and, moreover, f ( z [ s] ) -CO

- z _ 1 f ([ s] ) = 0. i n f a c t , i f s = G(z) (zu (z} + U q ) , f ( z [ s] ) = f ([ G (z) u(z) ] ) = -1 -1 ?

= [ z u ( z ) ] i n Z (G) . On the other hand, z f ([ s] ) = z [ z " u ( z ) + zu ] = co U

= [ z u ( z ) + u ] = ( z u ( z ) ] i n Z (G) s i n c e u G f i U. Hence f i s Kffz 1-lir.ear as i c v 0 CO 0 CO

S*/R* .and Z (G) a r e t o r s i o n .

To show t h a t f i s i n j e c t i v e , assume chat, f o r s = G(z ) u ( z ) , f ( [ s ] ) = 0. Then

zu(z) Ker G + fi U, i . e . zu(z) = v(z) + w ( z ) , v ( z ) e Ker G •? id w(z) proper. M u l t i -ÙO

-1 ' -1

p l y i n g by z and applying G(z) we have G(z)u(z) = G(z)(z w ( z ) ) , hence G(z)u(z) e

e R* and [ s] = 0 i n S* /R* . To show t h a t f i s s u r j e c t i v e , l e t us r e c a l i t h a t any element i n Z (G) i s the equi-

-1 valence c l a s s , modulo Ker G & Sì, U, of an element i n G (fi Y) . Then any element can

oo oo

be represented as [ z u ( z ) l , where u(z) i s a polynomial such t h a t G(z)u(z) i s strìctly

proper, and i t f o l l o w s t h a t [ z u ( z ) ] = f ( [ s ] ) w i t h s = G(z)u(z) e s* .

COROLLARY 3.4 The i n v a r i a n t f a c t o r s of S*/R* over K|z J describe the zero s t r u c t u r e

at i n f i n i t y o f G(z).

Proof. T r i v i a l by 3.3 and 2.4.

4. INFINITE POLE MODULE

To apply the no t i o n of i n f i n i t e zero rnodule t o the study of inverse t r a n s f e r

f u n c t i o n s , we need the dual notion of i n f i n i t e pole module. I t has been remarked i n

[ 8 ] t h a t the f i n i t e pole module of a r a t i o n a l G(z) in e s s e n t i a l l y the state space of-

a minimal r e a l i z a t i o n of the s t r i c t l y proper p a r t of 0 ( 2 ) . C l e a r l y , the generalized

state space of a minimal r e a l i z a t i o n of the polynomial p a r t of G(z) cannot be chosen

to represent the ' i n f i n i t e pole module we need since i t may contain a nondynamical

component (see [ 2 ] ) .

In the sarné way as i n section 2, where we considered the d e f i n i t i o n of Z (G), the co

case in = p = 1 suggests t o us the f o l l o w i n g abstract d e f i n i t i o n :

DEFINITION 4.1 Given a t r a n s f e r f u n c t i o n G(z) i t s i n f i n i t e pole module P (G) i s d e f i -. — co

ned by : , J G{fi U) + fi Y

P (G) = fi Y

,1 PR0P0SITI0N 4.2 P (G) i s a f i n i t e l v generated t o r s i o n KtL z 11-module whose n o n t r i v i a l —-—,———. & „ - 1 , + i n v a r i a n t f a c t o r s over Kil z U coincide w i t h the n o n t r i v i a l elements of S .

Proof. G(fi U) and fi Y are f i n i t e l y generated K[1 z 1-modules, then P (G) i s f i n i t e l y

- Ao-

generated. Any element i n P (G) i s the equivalence cl a s s , modulo Q Y, of some y = co co

= G(z)u, w i t h u e 9. U. I f y i s proper, [ y) = 0 . I f y has a polynomial p a r t of de­ce -k , -k. , aree k, z y i s proper and z [ yj = 0; hence P (G) i s t o r s i o n i

oo

To prove the second p a r t of the p r o p o s i t i o n , we show t h a t P (G) i s isomorphic t o

the t o r s i o n Kjz S-modul'e fi Y/T(z)fi Y where G(z) = T *(z)V(z) i s a coprirne proper 00 00

f a c t o r i z a t i o n . Remark, f i r s t of a l i , t h a t f o r any u e fi U, T(z)G(z)u = V(z)u belongs oo

t o fi Y. Therefore, T(G(fi U) + fi Y) C fi Y and there e x i s t s h : P (G) •> fi Y/T(z)fi Y c o co CO OO OO 00 00

such t h a t the f o l l o w i n g diagram, where the upper v e r t i c a l maps are canonical i n c l u -sions and the lower ones are canonical p r o j e c t i o n s , cornrautes :

T o, Y • T(z) f i Y

CO co

T 4.3 G(fi U) + fi Y fi Y — ' CO oo oo

p + 4- q P (G) • • fi Y/T(z)fi Y

CO CO co

Assume t h a t h(y) = 0 , w i t h y = p(G(z)u) and u e fi U. Then qTG(z)u = hpG(z)u = 0 and OO

TG(z)u G T( z ) f i Y. Let TG(z)u = Tv, v e fi Y, then G(z)u = v e fi Y and y = pv = 0 i n co oo oo

P (G); hence h i s i n j e c t i v e . 00

Let v be an element i n fi Y/T(z)fi Y, i . e . v = qy, y 6 fi Y. By coprìmness of T(z) and CO OO CO

V(z) there e x i s t proper matrices A(z) and B(z) such t h a t y = T(z)A(z)y + V(z)B(z)y

and qy = q ( V ( z ) B ( z ) y ) . Take u * B(z)y i n fi U, then hpG(z)u * qTG(z)B(z)y = oc

= qV(z)B(z)y = qy = v and h i s onto.

REMARK 4.4 A consequence of 4.2 i s t h a t P (G) contains a l i the infor m a t i o n about the co

pole s t r u c t u r e a t i n f i n i t y of G(z). More p r e c i s e l y , i l {v , ,v } i s the s t r u c t u r e I r

at i n f i n i t y of G(z), thè i n f i n i t e pole module has the f o l l o w i n g canonical decomposition ' l ' I

i n t o a d i r e c t sum of c y c l i c submodules : P (G) = $ K|[ Z } / z -""KIZ ] . co V^>0

I n conclusion, the decomposition i n t o d i r e c t sums of c y c l i c submodules of Z (G) and 00

P (G) determines the non zero indices o f the s t r u c t u r e at i n f i n i t y of G(z). Moreover, 00

the s t r u c t u r e at i n f i n i t y contains a fiumber of zeros equal t o the d i f f e r e n c e (rank G - (number of c y c l i c submodules i n d i r e c t sum decompositions of Z (Q) and P (G)).

REMARK 4.5 I t i s easy t o see t h a t P (G) i s isomorphic t o the q u o t i e n t module — • — — co

<

fi u co

. This a l t e r n a t i v e representation p o i n t s out the r e l a t i o n between G (fi Y) H u

co co -1 . , P (G) and the latency kernel G (fi Y) (see l 4 J ) . This, together w i t h 2.6, gives an

i n s i g h t i n t o the connection between the concept of latency and the s t r u c t u r e at i n f i ­

n i t y . In p a r t i c u l a r , i t appears t h a t the latency kernel contains i n f o r m a t i o n on both

the i n f i n i t e zeros and the i n f i n i t e poles of G(z). However, as G (fi Y) i s not f i n i -co

t e l y generated unless G(z5 i s i n j e c t i v e , ([ 4 ] 6.16), Z (G) and P (G) are more handable CO 00

algebraic o b j e c t s .

In case G(z) i s i n j e c t i v e and s t r i c t l y proper, the latency indice:; {X , ,X } are 1 m

defined i n [ 4 ] i n the f o l l o w i n g way : l e t { d , ,d } be an ordered proper basis 1 m -1 of G (fi Y); then ord d < -1 and A = -ord d, - 1. Remarking t h a t the polynomial p a r t °° i i i

of any d. generates a c y c l i c submodule of order equa! t o - ord d, i n Z (G), we haye

t h a t the latency indices coincide w i t h the order of the i n f i n i t e zeros decreased by

1. As a consequence, G(z) i s non l a t e n t i f f a l i i t s i n f i n i t e zeros have order 1.

When G(z) i s proper, obviously G(fi U) C fi Y and P (G) = 0. Let now G(z) be a px m 00 co 00

t r a n s f e r f u n c t i o n of order k < 0. To c l a r i f y the r e l a t i o n between P (G) and X (G), the 00 co

generalized s t a t e space of the minimal r e a l i z a t i o n of the polynomial p a r t of G(z), l e t us consider the f o l l o w i n q diagram (see also 1.2) :

fi U

4.6 i d C'

Q U

P (G) 00

n Where TT* : V Y -> T* Y i s the projectìon TT* ( U Z f

r Y = AY/z * f i Y

r*.Y = AY/fi Y

+• u z + u ) = u z + 1 0 n + u^z and è i s the r e s t r i c t i o n of TT* (remark t h a t $ i s w e l l defined since TT*G = G * ) .

p p

PROPOfelTION 4.7 The morphism $ : X (G) •> P (G) i s s u r j e c t i v e . The c y c l i c submodules • a ? 00

of order k+1 of X ( G ) are mapped o r t o c y c l i c submodules of order k of P ( G ) .

Proof. The s u r j e c t i v i t y of <j> follows by the commuta t i v i ' t y of 4.6.

Let {x} be a c y c l i c submodule of order k+1 of X (G) , i . e . z 1 x f 0 i n T Y f o r i < 00 00 -k-1 -k-1 -k < k ahd z x = 0 i n T Y. I n other words, z x i s s t r i c t l y proper, z x i s pro-

co — i * — i — i - i per and z x has negative order f o r i < k. Then % $(x) ~ z ?r*x = v* (z x) 7* 0 f o r —k **k

i < k-1, Z 0 (x) = 7;* (z x) =0. Hence {<}> (x) } i s a c y c l i c submodule of P (G) of

order k.

- 4 2 -

REMARK 4.8 Let X (G) = © K|z 1/ z ^ K f z ] be the canonical decomposition of X (G) , 0 0 i °° i n t o a d\ j-ct sum of c y c l i c submodules. Then P (G) = ® K[z 1/ z ^ Kfz ] and the

i i n dices p -1 coincide w i t h the ind i c e s v of the pole s t r u c t u r e at. i n f i n i t y o f G(z). i i Moreover, denotino by G (z) the polynomial p a r t of G(z), we have by 4.7 and [ 2 ] :

p o i qeneralized ord. G (z) = dim X (G) = £ u > 2 (v +1) and (number o f indipendent p o i K 0 0 i i v i>0 i impulsive motions of G(z)) = dim P (G) = E v .

K 0 0 v i>0 i Hence, the d i f f e r e n c e between dim X (G) and dim P (G) i s equal t o the number of c y c l i c

K 0 0 K 0 0

submodules i n the d i r e c t sum decomposition of X^(G) or, e q u i v a l e n t l y , t o the number of

c y c l i c submodules i n the d i r e c t sum decomposition of P (G) plus the number of (non co ,

dynamical) c y c l i c submodules of order 1 of X (G). co

5. INVERSE TRANSFER FUNCTIONS

In t h i s section we i n v e s t i g a t e the connection between the i n f i n i t e zero module of

G(z) and the i n f i n i t e pole module of a ( r i g h t or l e f t ) inverse H(z) of G(z). I n the

case m = p = 1 any G(z) has a unique inverse 8(1) whose number of poles at i n f i n i t y i s

equal t o the number of zeros at i n f i n i t y of G(z). I n the m u l t i v a r i a b l e case, i t w i l l

be proved t h a t Z (G) i s a s o r t of lower bound, i n an module t h e o r e t i c sense, f o r P (H). 00 00

More p r e c i s e l y , we have the f o l l o w i n g two pr o p o s i t i o n s .

PROPOSITION 5.1 Let G(z) : AU -> AY be an i n j e c t i v e t r a n s f e r f u n c t i o n and l e t

H(z) : AY •* AU be a l e f t inverse of G(z), i . e . H(z)G(z) = 1 ^ . Then there e x i s t s an

i n j e c t i v e Klz 1-morphism j : Z (G) •> P (H) . OO OO

I

Proof. For any u € fi U such t h a t G(z)u = y belongs t o fi Y, we have H(z)G(z)u = CO CO

= u = H(z)y, hence G (fi Y) C H(fi Y) . This assure the existence of j : Z (G) ->- P (G) 00 CO 00 oo

such t h a t the f o l l o w i n g diagram commutes : i d

o • n u fi u • o 4- 4--1 i n c l 0 <• G (fi Y) + fi U • H (fi Y) + fi U

co co co co

4- « 4-Z (G) - • P (H)

CO CO

Moreover, j i s uniquely determined by the above property and i t . i s easìly seen,

using the snake lemma, t o be i n j e c t i v e .

PROPOSITION 5.2 Let G(z) : AU •* AY be a s u r j e c t i v e t r a n s f e r f u n c t i o n and l e t

H(z) : AY •+ AU be a r i g h t inverse of G(z), i . e . G(z)H(z) = 1 . Then there e x i s t s a AY

s u r j e c t i v e KIZ 1-morphism p : P (H) -*• Z (G) .

Proof. Let u = H(z)y be an element of H(fi^Y}, then G(z)u = G(z)H(z)y = y belongs t o

fi Y and H(fi Y) C G 1 ( f i Y). This assure the existence of p : P (H) •* Z (G) such t h a t

the f o l l o w i n g diagram commutes :

i n c l fi u — - — • — • Ker G + fi U co oo + , 4-i n c l -1 H(fi Y) + fi U G (fi Y) + fi U

P l 4- 4- P 2 p ( H) • Z (G)

CO 00

p i s uniquely determined by the above property.

Let x be an element-of Z (G) , x = p u w i t h G(z)u = y e fi Y. We have y = G(z)H(z)y co 2 0 0

and t h e r e f o r e G(z)(u - H(z)y) =0. This implies t h a t (u - H(z)y) belongs t o Ker G C

C Ker p 2 and t h a t pp^H(z)y - x = p^fHKzJy - u) - 0. As a consequence, x = pj>^H(z)y

and p i s s u r j e c t i v e .

Now, as MacMillan degree H(z) = dim X(H) + dim P (H) and generalized ord. H(z) = K K «>

= dim X(H) + dim X (H) (see [ 2 ] , [ 5 ] , [ 6 i ) , we have the f o l l o w i n g c o r o l l a r y : K K »

COROLLARY 5.3 Let H(z) be a ( r i g h t or l e f t ) inverse of the t r a n s f e r f u n c t i o n G(z).

Then MacMillan degree H(z) = dim X(H) + dim P (H) > dim Z(G) + dim Z (G) and K K » K K »

qeneralized ord. H(z) = dim X(H) 4- dim X (H) = dim X (H) + dim P (H) + (number of cy-3 K K c o K K » c l i c submodules i n d i r e c t sum decomposition of X ( H ) ) > dim Z(G) + dim Z (G) + (number

• » K K of c y c l i c submodules i n d i r e c t sum decomposition of Z (G)).

Proof. By [ 2 ] and by 4.6, 5.1, 5.2.

REMARK 5.4 We remark t h a t using the same techniques, w i t h the obvious m o d i f i c a t i o n s ,

as i n [ 3 ] 3.6 and 3.9 i t i s possible t o construct r i g h t or l e f t inverses such t h a t

j o r , r e s p e c t i v e l y , p are isomorphism.

- 4 4 -

CONCLUSION Twc abstract algebraic objects associateci w i t h any t r a n s f e r f u n c t i o n G(z), namely

the i n f i n i t e zero module Z (G) and the i n f i n i t e pole module P (G), have been i n t r o d u ­co co

ced. I t has been shown t h a t they describe the zero/pole s t r u c t u r e a t i n f i n i t y o f G(z)

and t h a t there e x i s t s a canonical r e l a t i o n between Z (G) and P (H) where H(z) i s a CO oo

( r i g h t or l e f t ) inverse of G(z). More p r e c i s e l y , Z (G) i s contained, i n a s u i t a b l e

sense, i n P (H). oo

These r e s u l t s complete the algebraic theory of the ( f i n i t e ) zero and pole module i n

the sense of B.Wyman and M.Sain [ 8 ] .

Moreover, together w i t h the r e a l i z a t i o n theory f o r non proper r a t i o n a l t r a n s f e r

f u n c t i o n s developed i n [ 2 ] , they give a b e t t e r understanding of the problems involved

i n the c o n s t r u c t i o n of the minimal inverse of a given G(z), as shown i n 5.3. Further

i n v e s t i g a t i o n s on t h i s subject w i t h the aid of the algebraic t o o l s described here w i l l

be the argument of a forthcoming pàper.

REFERENCES [ 1 ] C.Commault and J.M.Dion - Structure a t i n f i n i t y of l i n e a r m u l t i v a r i a b l e systems :

a geometrie approach - 20th IEEE Conf. on Decision and Control (1981)

[ 2 ] G.Conte and A.Perdon - Generalized s t a t e space r e a l i z a t i o n of non proper r a t i o n a l t r a n s f e r functions - System & Control L e t t e r s 1 (1982)

f 3 ] G.Conte and A.Perdon - An algebraic notion of zeros f o r systems over r i n g s - MTNS 1983 Conf., Beer Sheva (1983)

[ 4 ] j.Bammer and M.Heymann - Causai f a e t o r l z u t i o n an l i n e a r feedback - SIAM J. Control Opt. 19 (1981)

[ 5 ] T.Kailath - Linear Systems - Prentice H a l l (1980) [ 6 ] H.Rosenbrock - S t r u c t u r a l p r o p e r t i e s of l i n e a r dynamical systems - I n t . J. Control

20 (1974) [ 7 ] B.Wyman - Linear systems over commutative rings - Lecture Notes, Stanford Univ.

(1972) t 8 ] B.Wyman and M.Sain - The zero modula and e s s e n t i a l inverse systems - IEEE Trans.

C i r c u i t and Systems CAS-28 (1901)