14
This article was downloaded by: [University of Toronto Libraries] On: 22 February 2013, At: 14:02 Publisher: Taylor & Francis Informa Ltd Registered in England and Wales Registered Number: 1072954 Registered office: Mortimer House, 37-41 Mortimer Street, London W1T 3JH, UK Applicable Analysis: An International Journal Publication details, including instructions for authors and subscription information: http://www.tandfonline.com/loi/gapa20 On exponential stability of infinite dimensional linear systems with bounded or unbounded perturbations Peng Li a & N.U. Ahmed b a Department of Electrical Engineering, University of Ottawa, Ottawa, Ontario, Canada b Department of Electrical Engineering and Department of Mathematics, University of Ottawa, Ottawa, Ontario, Canada Version of record first published: 02 May 2007. To cite this article: Peng Li & N.U. Ahmed (1988): On exponential stability of infinite dimensional linear systems with bounded or unbounded perturbations, Applicable Analysis: An International Journal, 30:1-3, 175-187 To link to this article: http://dx.doi.org/10.1080/00036818808839800 PLEASE SCROLL DOWN FOR ARTICLE Full terms and conditions of use: http://www.tandfonline.com/page/terms-and-conditions This article may be used for research, teaching, and private study purposes. Any substantial or systematic reproduction, redistribution, reselling, loan, sub-licensing, systematic supply, or distribution in any form to anyone is expressly forbidden. The publisher does not give any warranty express or implied or make any representation that the contents will be complete or accurate or up to date. The accuracy of any instructions, formulae, and drug doses should be independently verified with primary sources. The publisher shall not be liable for any loss, actions, claims, proceedings, demand, or costs or damages whatsoever or howsoever caused arising directly or indirectly in connection with or arising out of the use of this material.

On exponential stability of infinite dimensional linear systems with bounded or unbounded perturbations

  • Upload
    nu

  • View
    212

  • Download
    0

Embed Size (px)

Citation preview

This article was downloaded by: [University of Toronto Libraries]On: 22 February 2013, At: 14:02Publisher: Taylor & FrancisInforma Ltd Registered in England and Wales Registered Number: 1072954 Registered office: MortimerHouse, 37-41 Mortimer Street, London W1T 3JH, UK

Applicable Analysis: An International JournalPublication details, including instructions for authors and subscription information:http://www.tandfonline.com/loi/gapa20

On exponential stability of infinite dimensionallinear systems with bounded or unboundedperturbationsPeng Li a & N.U. Ahmed ba Department of Electrical Engineering, University of Ottawa, Ottawa, Ontario,Canadab Department of Electrical Engineering and Department of Mathematics, Universityof Ottawa, Ottawa, Ontario, CanadaVersion of record first published: 02 May 2007.

To cite this article: Peng Li & N.U. Ahmed (1988): On exponential stability of infinite dimensional linear systems withbounded or unbounded perturbations, Applicable Analysis: An International Journal, 30:1-3, 175-187

To link to this article: http://dx.doi.org/10.1080/00036818808839800

PLEASE SCROLL DOWN FOR ARTICLE

Full terms and conditions of use: http://www.tandfonline.com/page/terms-and-conditions

This article may be used for research, teaching, and private study purposes. Any substantial orsystematic reproduction, redistribution, reselling, loan, sub-licensing, systematic supply, or distribution inany form to anyone is expressly forbidden.

The publisher does not give any warranty express or implied or make any representation that thecontents will be complete or accurate or up to date. The accuracy of any instructions, formulae, and drugdoses should be independently verified with primary sources. The publisher shall not be liable for anyloss, actions, claims, proceedings, demand, or costs or damages whatsoever or howsoever caused arisingdirectly or indirectly in connection with or arising out of the use of this material.

dpplrtohii .Anoh i i . ~ Lo1 30. pp 175- I87 Rcprlnts d \a~iahle d~rectly from the pubilsher Photocopj~ng pernm~ttcd hq l~ ienre on15

( I988 Gordon and Breach. Sc~ence Publ~sher,, Inc Printed In Great B r ~ t a ~ n

On Exponential Stability of infi- nite Dimensional Linear Systems with Bounded or Unbounded Perturbations Communicated by Z. Nashed

KEY WORDS: Infinite dimensional systems. stabilizability. exponential stability. semigroup, relative bounded perturbations, m- dissipativity, feedback control

( R e w i w / / b r Publication 14 April 1988)

Peng Li Department of Electlical Erlginee~ing, University of Ottawa, Ottawa, On- tario, Canada I i l N G N j

IU.U.11llrned Dcpartnlent of Electrical Engineering and Department of hlathematics, University of O t t ana , Ottawa. Ontalio, Canada I i l K 6x5

Al~st ract In thi, paper. n e stutlj the qi~estion of stal~ilization of uncertain systems governed 11) e ~ o l u t ~ o n ecjuat~ons in IIilbert space. It is shown that exponential stabllit?; can be acliieved h?; choice of suitable state feedback controls even in the presence of l~ountled or relatively bounded(uncertain) perturbations. Our proof is l~ascd on perturbation theory of semigroups. The results are illustrated 115 two examples involving heat equation and wave equation.

This work was supported in part I]?; tlw Natural Science and Engineering Research Council of Canada under Giant No.X7109.

Dow

nloa

ded

by [

Uni

vers

ity o

f T

oron

to L

ibra

ries

] at

14:

02 2

2 Fe

brua

ry 2

013

176 PENG LI AND N. U. AHMED

1. INTRODUCTION

The question of stabilization of uncertain linear systems described by ordinary differential equations,

i ( t ) = ( A + B ( r ) ) x ( t ) + Cu( t )

x (0) = so.

with deterministic feedback control has received considerable attention in cur- rent literature[l-61 where the state r € Rn, A is an n x i z matrix, B ( r ) is the

system uncertainty, C is a matrix of suitable dimension and u is the control. In many scientific and engineering problems the system is modelled by par-

tial differential equations, integral equations or coupled ordinary and partial

differential equations. \I7e know that a large class of linear delay differen- tial equations and partial differential equations can be modelled as abstract evolution equations of the form

k ( t ) = Ax( t ) + Cu( t )

.c (O) = ro

in an appropriate Banach space X.The system operator A may be bounded or

unbounded, the control operator C is bounded and u ( t ) is the control force. The uncertainty considered here is the perturbation of operator governing the

dynamics of the system. That is,

?( t ) = (il + B ( r ) ) z ( t ) + Cu( t )

x ( 0 ) = 5 0

where Bjr ) is a relatively bounded perturbation of A for each r in a suitable

set R; B ( r ) may be bounded or unbounded. The parameter r is not known but

the set R is known. The problem is to design a state feedback linear control

law that assures exponential stability of the zero state uniformly with respect

to r E R. To our knowledge, this problem, involving infinite dimensional

systems with bounded or unbounded perturbation, has not been considered in the literature before. D

ownl

oade

d by

[U

nive

rsity

of

Tor

onto

Lib

rari

es]

at 1

4:02

22

Febr

uary

201

3

INFINITE DIMENSIONAL SYSTEM STABILITY

2. PROBLEM STATEMENT AND ASSUMPTIONS

Consider an abstract system

on a real Hilbert space X where A: D(A) + X is the infinitesimal generator of - a strongly continuous semigroup(Co- semigroup) T( t ) , t 2 0, on X, D(A) = X. C is a bounded linear operator from CJ into X, where U is another real Hilbert space (with dimension(U)finite or infinite), and { B ( r ) , r E R) is a family

of linear operators perturbing the generator A. We introduce the following assumptions for the operators A, B(r) , r E R and C.

Assum~tion 1

The pair {A, C ) is exponentially stabilizable; that is, there exists a Do E

L(X, U) such that 2 - A+ CDo generates an exponentially stable contraction semigroup Tf in X satisfying \/Tf)I < edwt, for some w > 0.

Assumpt.ion 3,

B(r) Bl(r) + B2(r) , where D(A) C D(Bl( r ) ) C X for all r E R, and for the given R, there exist constants CY E [O,1) and y 2 0 such that, for all r E R,

(x, BI(T)x) L a(x , Ax), for x E D(A)

Assum~tion 3 The linear operator CC* is coercive; that is,

( I , CC*x) 2 x ~ [ x I ( ~ , for some X > 0 and x E X,

where C* is the dual of the operator C.

In majority of engineering problems the stronger form of asymptotic sta- bility(i.e exponential asymptotic stability), is most desirable. In this paper D

ownl

oade

d by

[U

nive

rsity

of

Tor

onto

Lib

rari

es]

at 1

4:02

22

Febr

uary

201

3

178 PENG LI AND N . U . AHMED

we are mainly concerned with this problem. In other words, we wish to find

an operator Dl E C ( X , U ) , giving the feedback control u = D l z , such that

( A + B ( r ) + CD,) is the generator of an exponmtially stable Co-se~nigrorlp in

X for all r E 0.

3. h GFXERAL STABILIZATION RESULT

Our main stability rcqult is given in the follo~r-ing theorem

Theorem 1.

conszder the system (1) tczth lznear feedback control gzven b y

I f the assurnpt~ons 1-3 are satzsfied, then for any znltcal condztzon TO E D ( A ) C X , the null-stclte x = 0 is e.rponent~ully stable unrformly wzth rfcpect to r E R

Some definit~ons and Lemmas useful for the p ~ o o f of the Thtorem are glvcn

below:

Definition 1.

A dissipatice operator 11 for which R(I - A) = S is called in-dissipntice.

It is clear that if A is dissipative so is pr l for all p > 0, therefore if A is

m-dissipatiae then R(XI - A ) = X for every X > 0.

Lemma l .(Paq[7; Theorem 3.2, p.811)

Let A and B be lznear operators In S such that D ( B ) > D ( A ) nnd A + tB 2s d~ss~patzve for 0 5 t 1 1 . I f IIBzIl I. a l /Ax / l + Jllxll for x E D ( A ) where

0 < cr < 1, ,O 2 0 a n d f o r some to E [0 , 11. A + t o B 1s m-dmzpatzce then :I+ t B

zs m-dzsszpat~ve for all t 6 [O, 11.

Lemma 2.

Let T and A be linear operators in X , D ( A ) C D ( T ) C S and D ( A ) = .Y. If ( x , T x ) 1 ( x , A x ) , x E D ( A ) , then l\Txll 1 llAxll, x E D ( A ) . D

ownl

oade

d by

[U

nive

rsity

of

Tor

onto

Lib

rari

es]

at 1

4:02

22

Febr

uary

201

3

INFINITE DIMENSIONAL SYSTEM STABILITY 179

- Since D ( A ) = X it follows that for any 0' E -Y with l I r l t i l = 1, one can fiild a

sequence { T ~ ) c D ( A ) with llsnll = I , n = 1 , 2 . ... such that 77, 4 T ~ ' . S ~ n c e D ( A ) c D ( T ) C X , T x E X for x E D ( A ) and ( v n , T x ) 5 llAx11. taking the

limit we have

( ? ' , T x ) JIAzJl for x E D ( A ) .

Since 0' is an arbitrary element of X satisfying i/r1'1/ = 1, this inequality is

satisfied for all such 7'. Hence

/ ( T x j ( = sup ( q ' , T x ) 5 llAx// for x E D ( A ) . ll7'll=l

Thus if T is form-relatively hounded with respect to '1, that is

- ( x , T x ) < a ( x , A x ) + ?(a , x ) for x E D(A), U(A) = S

then T is operator-relatively bounded with respect to A , tha t is.

Therefore, assumption 2 implies that B ( r ) is relati~ely bounded with re-

spect to A for all r E 0.

Lemma 3. Le t A be t h e genera tor o f C o - s e m l g r o u p Tt. t 2 0, zn X . Then T t , t 2 0. r r

a n exponen tml l y stable contraction scrnzyroup rf and o n l y 1f zts generato? I S

s trzc t ly dzsszpahue, t ha t zs, (x, As) 5 - t ~ j j - c j ( ~ f o r s E D ( A ) . a n d R((1- r c ) I - A ) = X for s o m e w > 0 .

Proof: Applying Lumer-Phillips Theorem (Pazy[7; Theorem 4.3, p.141) to the linear D

ownl

oade

d by

[U

nive

rsity

of

Tor

onto

Lib

rari

es]

at 1

4:02

22

Febr

uary

201

3

180 PENG LI AND N. U. AHMED

operator B = A + w I , the result follows upon noting that B generates the

Co-semigtuup St = eu"Tt of contractions given that {Tt , t 2 0) is the Co- semigroup generated by A if and only if B is rn-dissipatiue. rn

IVitll feedback control (2) and subject to the assumptions 1-3, the system

(1) becomes

nllerc Ck B l ( r ) + n C D o . C l f B2(r ) - (*I/X)CC* ale linear operators in S. LVith these preparations we can now give a proof of our main result.

Proof of Theorem 1:

In ordel to prove the Theorem, one must show that (a) gi\en any initial condition x(0) = s o E D(A), there is a t least one classical

solution of (3).

(b) A + C, is the generator of an exponentially stable contraction semigroup with exponent independent of r E R. For (a) i t suffices to show that A -+ C, generates a Co-semigroup for every

r E n. This, however, follows from (b). Hence we concentrate on proving (b).

Since C and Do are bounded operators, ~ ( 2 ) = D ( A $ CDo) = D(A) and - ~ ( 2 ) are both in X and D ( A ) = X. By assumption 1 and Lemma 3, we know

that 2 is strictly dissipative satisfying

Define A, = A + w I , then by definition A, is rn-dissipatice and D(C,) > D(A,). LVe prove that A, + tCr is dissipative for 0 5 t 5 1. Indeed by assumptions 2 and 3

< t a ( x , Ax) + tcu(x, CDox) t ty(x, x) - t(y/X)(x, CC*x)

Dow

nloa

ded

by [

Uni

vers

ity o

f T

oron

to L

ibra

ries

] at

14:

02 2

2 Fe

brua

ry 2

013

INFINITE DIMENSIONAL SYSTEM STABILITY

Hence A, + C, is dissipative and A + Cr is strictly dissipative.

Utilizing assumption 2, we obtain

= a ( x , Ax)

5 a ( x , A , x ) for x E D ( A , ) = D ( A ) ,

and hence it follows from Lemma 2 that

Since C E L(U, X) it follows from assumption 2 that there exists a constant

P = P(7, A, IlCll) > 0 such that

As A, is rn-dissipative and CI is relatively bounded, it follows from Lemma

1 that A, + Cr is m-dissipative uniformly with respect to r 6 R. Thus by

Lemma 3 A + C, is the generator of an exponentially stable contration semi-

group with exponent w which is indepent of r € 0. This completes the proof

of Theorem 1. W

We have shown in Theorem 1 that if the perturbing operator B ( r ) , r E

R, in system (I), is uniformly form-relatively bounded then the system is

exponentially stable. Next we show that if B ( r ) , r E R, is operator-relatively

bounded then the same result follows under certain additional assumptions.

For this purpose we shall use the following Lemma.

Lemma 4.(I<ato[9; Theorem 4.12, p.2921) Let A be selfadjoint and nonnegative and T symmetn'c with D(T) > D(A) and

Dow

nloa

ded

by [

Uni

vers

ity o

f T

oron

to L

ibra

ries

] at

14:

02 2

2 Fe

brua

ry 2

013

182 PENG L1 AND N. U. AHMED

Then

( T u , u ) < ( A u , u ) for u E D ( A )

iVe replace the assumption 2 by

Assumption 2'

B ( r ) r B l ( r ) + B z ( r ) , where D ( A ) C D ( B l ( r ) ) C X for all r E a, B l ( r ) is

symmetric and for the given R, there exist a E [ O , l ) and y 2 0 such that

IIB(r)x1I < CY\IAXI( + Y / ( x ~ / for x E D ( A ) and for all r E 0.

Under the assumptions 1, 2' and 3 it is easy to prove the following result.

Corollan 1.

Conszder the system (1) where A zs a selfadjoznt posztzve definzte h e a r operator

wzth feedback control zi = D l x where D l zs as dejined 211 equatzon (2). If the

assumptions 1 , 2 ' and 3 are satzsjied then x = 0 zs exponentzally stable.

In case of bounded perturbation, B ( r ) = B 2 ( r ) , we have the following

result.

Corollarv 2.

Conszder the system (1) and suppose the assumptrons 1 and 3 hold wzth the

ptrturbzng operator B(r) bounded such that lJB(r) j j 5 7 for a l l r E R. Then

wzth control u = ( D O - ( y /X )C t ) x , the system (1 ) is exponentrally stable wzth

respect t o the null state.

4. FINITE DIhIEIiSIOn'.4L COKTROL

In practical situations, one usually considers the system

i ( t ) = ( A + B ( r ) ) x ( t ) + C u ( t )

= ( A + B ( r ) ) x ( t ) + C:,c,u,(t), x ( 0 ) = xo E D ( A ) ( 5 )

Dow

nloa

ded

by [

Uni

vers

ity o

f T

oron

to L

ibra

ries

] at

14:

02 2

2 Fe

brua

ry 2

013

INFINITE DIMENSIONAL SYSTEM STABILITY 183

where c, E X, u, E L?'(O, w). This is a special case of (1) corresponding

to finite dimensional control action, u E Rm. Under the assumptions 1-3,

Theorem 1 is also true for the system ( 5 ) . In this case, C : Rm i X. Let us

consider the system with a different constraint on C. Instead of CC' being a

coercive operator we may assume that the following condition holds:

Assumption 2"

B ( r ) = B1 ( r ) + B % ( r ) , where D ( A ) C D ( B ( r ) ) c X for all r E 0, and there

exist a E [0, l) , yl > 0 such that for all r E f2

(x, B1 ( r ) ~ ) 5 Q ( X , A x ) for x E D ( A ) ,

and Ba(r ) _< l l C C i for some yl > 0 and all r E R.

Theorem 2.

Consider the system (5) with linear feedback control given by

and D o zs as gzaen zn ccssumptzon 1. If the assumptzons 1 and 2 " are sa t~$ied ,

then for any admzsszble uncertaznty r E Cl , and any znitzal condztzon xo E

D ( A ) c X , the null state x = 0 zs exponentzally stable.

Proof: Identifying A = A 4 C D o and C: = B l ( r ) + a C D o it is clear from the proof of

Theorem 1 that A+c: is strictly dissipative satisfying ( ( ~ + C : ) X , s) .< - w ~ l z ~ ~ 2

for all r E fl and x E D ( A ) . Therefore it suffices to verify that the remaining

part C: r B Z ( r ) - r l C C i i s bounded and dissipative. Clearly C: is bounded

and it follows from assumption 2" that C: is dissipative. . In fact, Theorem 2 can also be applied to infinite dimensional control

problem where C E L ( U , X ) and u E U . See example 2.

Dow

nloa

ded

by [

Uni

vers

ity o

f T

oron

to L

ibra

ries

] at

14:

02 2

2 Fe

brua

ry 2

013

PENG LI AND N. U. AHMED

E x a m ~ l e 1.

Consider the controlled diffusion equation

z(0 , t ) = Z(1,t) = 0,

z(x, 0) = sin x x .

LVhere r is the uncertain parameter taking values from Q {[ E R : - N < - E 5 w.

iVe write this in the canonical form as follows:

t = Az + B ( r ) z + CU, z(0) = sin x x and r 6 R

with the state space X = L2(0, 1). The operator A is given by

a? a2 z A = - D(A) = {z E L2(0, 1); - E L2(0, 1) and z(0 , t ) = z (1 , t ) = 01,

8x2 ' 8x2

the operator B ( r ) is given by

B( r )z = r 2 z and C = Identity.

It is easy to verify that the assumptions 1-3 of Theorem 1 are satisfied. In

fact, the Co-semigroup T ( t ) generated by A satisfies

thus is exponentially stable (Curtain and Pritchard[lO; p.2151). Hence, in this

case we can take Do = 0. The simulation result is shown when N = 111 =3.3 for R.

Dow

nloa

ded

by [

Uni

vers

ity o

f T

oron

to L

ibra

ries

] at

14:

02 2

2 Fe

brua

ry 2

013

INFINITE DIMENSIONAL SYSTEM STABILITY

CONTROL

FIGURE 1 Diffusion equation (example 1)

Example 2. While considering distributed controls for vibrating systems it is required that every initial state of finite energy be transferred to a position of rest (zero state).

Consider the scalar wave equation [ll]

with initial state given by

2 z (s , 0) = x - x, x E (0 , l ) and zi(z,O) = s insx , x E (0 , l ) .

We may associate with this equation an abstract evolution equation of the form

t = Az + B(r)z + Cu, z(0) =

Dow

nloa

ded

by [

Uni

vers

ity o

f T

oron

to L

ibra

ries

] at

14:

02 2

2 Fe

brua

ry 2

013

186 PENG LI AND N . U. AHMED

where r E fl = {r E R : -N 5 r 5 M}, 0 5 ,V, 12f < w. In this case the state

space is given by X = lI1(O. 1) x L2(Ol l) , with D ( I ) = H2(0, 1) x H i ( 0 . 1 ) ~

A = (A A) , B(r) = (: , and C = (0, l ) ' .

The operator A generates a Co-semigroup T ( t ) and there exists Do =

(-1, -5) such that A + CDo generates an exponentially stable contraction

semigroup. Further B( r ) = CE(r)C* where E( r ) = r 2 , C * = (0, l ) . IIence,

the assumptions 1 and 2" are satisfied. For r2 5 8.05, the simulation result is

shown in the Fig.2. for N = 2 . 3 and ,I1 =2.5.

For numerical computation we have used finite difference method and the IMSL library subroutine.

FIGURE 2 Wave equation (example 2) Dow

nloa

ded

by [

Uni

vers

ity o

f T

oron

to L

ibra

ries

] at

14:

02 2

2 Fe

brua

ry 2

013

INFINITE DIMENSIONAL SYSTEM STABILITY

REFERENCES

[ l ] G.Leitmann, Guaranteed asymptotic stability for some linear systems with bounded uncertainties ,.Journal of Dynaniic Systems, Measurement and Control, VOL.lO1, 1\70.3,1979.

[2] G.Leitmann, On the efficacy of nonlinear control in uncertain linear sys- tems, Journal of Dynamic Systems, i\leasureinent and Control, VOL.102, N0.2,lgSl.

[3] B.R.Barmish and G.Leitmann, On ultimate boundedness control of unccr- tain systems in the absence of matching conditions,IE'EE Transactions on Automatic Control, VOL.AC-27,1982.

[4] B.R.Barmis11,I.R.Petersen and AFeuer, Linear ultimate boundedness con- trol of uncertain dynamical systems. Autornatica, L70L.19,19S3.

[J] J.R.Petersen, Structural stabilization of uncertain systems: necessity of the matching condition, SMI\I J. Confrol and Optim., L'OL.23,N0.2,1983.

(61 P.Li and N.U.Ahmed, Ol~servcr based stabilizing controllers for uncertain system with incomplete state information, Submitted.

[7] A.Pazy, Semigroups of Linear Opernfors and Applications to Partial Dif- ferential Equations, Springer-I,'erlag, 1983.

[8] G.R.Buis, Lyapunov stability for partial differential equations, Part I, N A S A Contractor Report, CR-1100, \Vashington, D.C., 1968.

[9] T.Kato, Perturbation Theory for Linear Operuiors, Springer -Lrerlag, 1980. [lo] R.F.Curtain and X.J.Pritchard, Functional Annlysis in Aloderii Applied

Mafhematics, Academic Press, 1977. [ l l ] A.T.Bharucha-Reid, Randorrl Integrnl Equations, Acadernic Press, 1972.

Dow

nloa

ded

by [

Uni

vers

ity o

f T

oron

to L

ibra

ries

] at

14:

02 2

2 Fe

brua

ry 2

013