TELK OMNIKA, V ol.16, No .6, December 2018, pp . 3024–3033 ISSN: 1693-6930, accredited First Gr ade b y K emenr istekdikti, Decree No: 21/E/KPT/2018 DOI: 10.12928/TELK OMNIKA.v16i6.10112 3024 T ransf er Function, Stabilizability , and Detectability of Non-A utonomous Riesz-spectral Systems Sutrima *1 , Christiana Rini Indrati 2 , and Lina Ar y ati 3 1 Univ ersitas Sebelas Maret, Ir .Sutami St no .36 A K entingan Sur akar ta, Ind onesia 1,2,3 Univ ersitas Gadjah Mada, Sekip Utar a K otak P os: BLS 21, Y ogy akar ta, Indo nesia * Corresponding author , e-mail: sutr ima@mipa.uns .ac.id, zutr ima@y ahoo .co .id Abstract Stability of a state linear system can be identified b y controllability , obser v ability , stabilizability , detectability , and tr ansf er function. The appro ximate controllability and obser v ability of non-autonomous Riesz-spectr al systems ha v e been estab lished as w ell as non-autonomous Stur m-Liouville systems . As a contin uation of the estab lishments , this paper concer n on the analysis of the tr ansf er function, stabilizability , and detectability of the non-autonomous Riesz-spectr al systems . A str ongly contin uous quasi semig roup approach is implemented. The results sho w that the tr ansf er function, stabilizability , and detectability can be estab lished comprehen siv ely in the non-autonomous Rie sz-spectr al systems . In par ticular , sufficient and necessar y cond itions f or the stabilizability and detectability can be constr ucted. These results are par allel with infinite dimensional of autonomous systems . K e yw or ds: detectability , non-autonomous Riesz-spectr al system, stabilizability , tr ansf er funct ion Cop yright c 2018 Univer sitas Ahmad Dahlan. All rights reser ved. 1. Intr oduction Let X , U , and Y be comple x Hilber t spaces . This paper concer ns on the linear non- autonomous control systems with state x , input u , and output y : _ x ( t ) = A ( t ) x ( t ) + B ( t ) u ( t ) ; x (0) = x 0 ; y ( t ) = C ( t ) x ( t ) ; t 0 ; (1) where A ( t ) is a linear closed oper ator in X with domain D ( A ( t )) = D , independent of t and dense in X ; B ( t ) : U ! X and C ( t ) : X ! Y are bounded oper ato rs such that B ( ) 2 L 1 ( R + ; L s ( U ; X )) and C ( ) 2 L 1 ( R + ; L s ( X ; Y )) , where L s ( V ; W ) and L 1 ( ; W ) denote the space of bounded oper ators from V to W equipped with strong oper ator topology and the space of bounded measurab le functions from to W pro vided with essential suprem um nor m, respectiv ely . The state linear system (1) is denoted b y ( A ( t ) ; B ( t ) ; C ( t )) . Symbol ( A ( t ) ; B ( t ) ; ) and ( A ( t ) ; ; C ( t )) denote the state linear system (1) if C ( t ) = 0 and B ( t ) = 0 , respectiv ely . In par ticular , this paper shall in v estigate stability of the state linear system (1) where each A ( t ) is a gener aliz ed Riesz-spectr al oper ator [1] using tr ansf er function, stabilizability , and detectability . The f ollo wing e xplanations giv e some reasons wh y these in v estigations are impor tant f or the systems . In the autonomous control system ( A; B ; C ) , where A is an infinitesimal gener ator of a C 0 -semig roup T ( t ) , there e xist special relationships among input, state , and output [2]. Controlla- bility map specifies the relationship betw een the input and the state , obser v ability map specifies the relationship betw een an init ial state and the output. The relationship betw een the input and the output can be char acter iz ed b y tr ansf er function, a linear map that specifies relationship betw een the Laplace tr ansf or m of the inputs and outputs . The f ollo wing studies sho w ho w urgency of the tr ansf er fu nction is in the line ar system. In the infinite-dimen sional autonomous system, e xter nal stability is indicated b y boundedness of its tr ansf er function [3]. W eiss [4] ha v e pro v ed a f or m ula f or the tr ansf er function of a regular linear system, which is similar to the f or m ula in the finite di- mensions . As an e xtension of results of [4], Staff ans and W eiss [5] ha v e gener aliz ed the results Receiv ed Ma y 31, 2018; Re vised October 5, 2018; Accepted No v ember 3, 2018 Evaluation Warning : The document was created with Spire.PDF for Python.
3025 ISSN: 1693-6930 of regular linear systems to w ell-posed linear systems . The y ha v e also introduced the Lax-Phillips semig roup induced b y a w ell-posed linear system. W eiss [6] ha v e studied f ou r tr ansf or mations which lead from one w ell-posed linear system to another : time-in v ersion, flo w-in v ersion, time- flo w-in v ersion, and duality . In par ticular , a w ell-posed linear system is flo w-in v er tib le if and only if the tr ansf er function of the system has a unif or mly bounded in v erse on some r ight half-plane . Finally , P ar tington [7] ha v e giv en simple sufficient condit ions f or a space of f unctions on (0 ; 1 ) such that all shift-in v ar iant oper ators defined on the space are represented b y tr ansf er functions . Controllability and stability are qualitativ e control prob lems which are impor tant aspects of the theor y of control systems . Kalman et al. [8] had initiated the theor y f or the finite dimensional of autonomous systems . Recently , the theor y w as gener aliz ed into controllability and stabilizability of the non-autonomous control systems of v ar ious applications , see; e .g. [9, 10, 11], and the ref erences t herein. The concept of the stabilizability is to find an admissib le cont rol u ( t ) such that the corresponding solution x ( t ) of the system has some required proper ties . If the stabilizabil- ity is identified b y n ull controllability , system (1) is said to be stabilizab le if there e xists a control u ( t ) = F ( t ) x ( t ) such that the z ero solution of the closed-loop system _ x ( t ) = [ A ( t ) + B ( t ) F ( t )] x ( t ) ; t 0 ; is asymptotically stab le in the L y apuno v sense . In this case , u ( t ) = F ( t ) x ( t ) is called the stabili- zing f eedbac k control. In par ticular f or autonomous system ( A; B ; ) , where A is an infinitesimal gener ator of an e xponentially stab le C 0 -semig roup T ( t ) , ( A; B ; ) is stabilizab le if oper ator A + B F is an infinitesimal gener ator of an e xponentially stab le C 0 -semig roup T B F ( t ) [2]. In the finite- dimensional autonomous control system, Kalman et al. [8] and W onham [12] had sho wn that the system is stabilizab le if it is n ull controllab le in a finite time . Bu t, it does not hold f or the con v erse . Fur ther more , if the system is completely stabilizab le , then it is n ull controllab le in a finite time . F or finite-dimensional non-autonomous control systems , Ik eda et al. [13] pro v ed that the system is completely stabilizab le whence it is n ull controllab le . Gener alizations of the results of t he stabilizability f or the finite-dimensional systems into infinite-dimensional systems ha v e been successfully done . Extending the L y apuno v equation in Banach spaces , Phat and Kiet [14] specified the relationship betw een stability and e xact n ull controllability in the autonomous systems . Guo et al. [15] pro v ed the e xistence of the infinitesimal gener ator of the per turbation semig roup . F or neutr al type linear systems in Hilber t spaces , Rabah et al. [16] pro v ed that e xact n ull controllability implies the complete stabilizability . In the paper , unbounded f eedbac k is also in v estigated. In the non-autonomous systems , Hinr ichsen and Pr itchard [17] in v estigat ed r adius stability f or the systems under str uctured non-autonomous per turbations . Niamsup and Phat [18] pro v ed that e xact n ull controllability implies the complete stabilizability f or linear non-autonomous systems in Hilber t spaces . Leiv a and Barcenas [19] ha v e introduced a C 0 -quasi semig roup as a ne w approach to in v estigate the non-autonomous systems . In this conte xt, A ( t ) is an infinitesimal gener ator of a C 0 -quasi semig rou p on a Banach space . Sutr ima et al. [20] and Sutr ima et al. [21] in v estigated the adv anced proper ties and some types of stabilities of the C 0 -quasi semig roups in Banach spaces , respectiv ely . Ev en Barcenas et al. [22] ha v e char acter iz ed the con trollability of the non-autonomous control systems using the quasi semig roup approach, although it is still limited to the autonomous controls . In par ticular , Sutr ima et al. [1] char acter iz ed the controllability and obser v ability of non-autonomous Riesz-spectr al systems . The ref erences e xplain that the tr ansf er function, stabilizability , and detectability of the control systems are impor tant indicators f or the stability . Unf or tunately , there are no studies of these concepts in the non-autonomous Riesz-spectr al systems . Theref ore , this paper concer ns on in v estigations of the tr ansf er function, stabilizability , and detectability of the non-autonomous Riesz-spectr al systems that ha v e not been in v estgated at this time y et. These in v estigations use the C 0 -quasi semig roup approach. 2. Pr oposed Methods and Discussion 2.1. T ransf er Function W e recall the de finition of a non-autonomous Riesz-spectr al system that ref ers to Sutr ima et al. [1]. The definition of a Riesz-spectr al oper ator f ollo ws [2]. TELK OMNIKA V ol. 16, No . 6, December 2018 : 3024 3033 Evaluation Warning : The document was created with Spire.PDF for Python.
TELK OMNIKA ISSN: 1693-6930 3026 Definition 1. The linear non-autonomous system ( A ( t ) ; B ( t ) ; C ( t )) is called a non-autonomous Riesz-spectr al system if A ( t ) is the infinitesimal gener ator of a C 0 -quasi semig roup which has an e xpression A ( t ) = a ( t ) A (2) where A is a Riesz-spectr al oper ator on X and a is a bounded contin uous function such that a ( t ) > 0 f or all t 0 . By e xpression (2), f or e v er y t 0 , w e see that A ( t ) and A ha v e th e common domain and eigen v ectors . In f act, if n , n 2 N , is an eigen v alue of A , then a ( t ) n is the eigen v alue of A ( t ) of (2). Hence , in gener al A ( t ) ma y ha v e the non-simple eigen v alues . In the non-autonomous Riesz-spectr al system ( A ( t ) ; B ( t ) ; C ( t )) , where A ( t ) is an infinitesimal gener ator of a C 0 -quasi semig roup R ( t; s ) , relationships among input, state , and output are deter mined b y R ( t; s ) . The relationships of state-input and state- output of the system ha v e been discussed b y Sutr ima et al. [1]. In this subsection, w e f ocus on char acter izing the relationship betw een the input and output of the system using the tr ansf er function, where B ( t ) 2 L 2 ( U ; X ) and C ( t ) 2 L 2 ( X ; Y ) f or all t 0 . W e define m ultiplicativ e oper ators B : L 2 ( R + ; U ) ! L 2 ( R + ; X ) and C : L 2 ( R + ; X ) ! L 2 ( R + ; Y ) b y: ( B u )( t ) = B ( t ) u ( t ) and ( C x )( t ) = C ( t ) x ( t ) ; t 0 ; respectiv ely . W e see that the oper ators B and C are bounded. The definition of the tr ansf er function of the system ( A ( t ) ; B ( t ) ; C ( t )) f ollo ws the definition f or the autonomous systems of [2]. Definition 2. Let ( A ( t ) ; B ( t ) ; C ( t )) be the non-autonomous Riesz-spectr al system with z ero initial state . If there e xists a real such that ^ y ( s ) = G ( s ) ^ u ( s ) f or Re ( s ) > , where ^ u ( s ) and ^ y ( s ) denote the Laplace tr ansf or ms of u and y , respectiv ely , and G ( s ) is a L ( U ; Y ) -v alued function of comple x v ar iab le defined f or Re ( s ) > , then G is called tr ansf er function of the system ( A ( t ) ; B ( t ) ; C ( t )) . The impulse response h of ( A ( t ) ; B ( t ) ; C ( t )) is defined as the Laplace in v erse tr ansf or m of G . The tr ansf er fun ction of the non-autonomous Riesz-spectr al systems with finite-r ank in- puts and outputs can be stated in eigen v alues and eigen v ectors of the Riesz-spectr al oper ator . Theorem 3. The tr ansf er function G and impulse re sponse h of the non-autonomous Riesz- spectr al system ( A ( t ) ; B ( t ) ; C ( t )) e xist and are giv en b y G ( s ) = C a ( ) ( sI a ( ) A ) 1 B on ( A ( )) ; and h ( t ) = C T ( t ) B ; t 0 0 ; t < 0 ; where T ( t ) is a C 0 -semig roup with the infinitesimal gener ator A . Proof . By definition of Riesz-spectr al oper ator and Theorem 3 of [1], w e ha v e that the resolv ent set of A is connected, so 1 ( A ) = ( A ) . W e shall v er ify the e xistences of the tr ansf er function and impulse response . Let R ( t; s ) be a C 0 -quasi semig roup with infinitesimal gener ator A ( t ) of the f or m (2). F or Re ( s ) > ! 0 a 1 , where ! 0 is the g ro wth bound of a C 0 -semig roup T ( t ) with infinitesimal gener ator A and a 1 := sup t 0 a ( t ) . By Lemma 2.1.11 of [2], w e ha v e C a ( ) ( sI a ( ) A ) 1 B u = C Z 1 0 e s a ( ) T ( ) B ud = Z 1 0 e s a ( ) C T ( ) B ud f or all u 2 L 2 ( R + ; U ) and Re ( s ) > ! 0 a 1 . As in the proof of Lemm a 2.5.6 of [2], these s can be e xtended to all s 2 ( a ( ) A ) . Corollar y 4. Let A ( t ) be an oper ator of the f or m (2), where A is a Riesz-spectr al oper ator with eigen v alues f n 2 C : n 2 N g . If B : R + ! L ( C m ; X ) and C : R + ! L ( X ; C k ) such that T r ansf er Function, Stabilizability , and Detectability of Non-A utonomous ... (Sutr ima) Evaluation Warning : The document was created with Spire.PDF for Python.
3027 ISSN: 1693-6930 B ( t ) 2 L ( C m ; X ) and C ( t ) 2 L ( X ; C k ) , then the tr ansf er function and impulse response of the system ( A ( t ) ; B ( t ) ; C ( t )) are giv en b y G ( s ) = 1 X n =1 a ( ) s a ( ) n C n B   n tr f or s 2 ( A ( )) (3) h ( t ) = 8 < : 1 P n =1 e n a ( ) t C n B   n tr ; t 0 0 ; t < 0 ; (4) where n and   n are the corresponding eigen v ectors of A and A , respectiv ely . Symbol W tr denotes tr anspose of W . Proof . By representation of ( sI a ( ) A ) 1 of Theorem 3 of [1] and f acts that oper ators B ( t ) and C ( t ) are bounded f or all t 0 , then from Theorem 3 f or s 2 ( A ( )) w e ha v e G ( s ) u = C a ( ) ( sI a ( ) A ) 1 B u = C " lim N !1 N X n =1 a ( ) s a ( ) n hB u;   n i n # = 1 X n =1 a ( ) s a ( ) n C n B   n tr : Here , w e use the proper ty h v ; w i C m = w tr v . Thus , e xpression (3) is pro v ed. By a similar argument and condition (c) of Theorem 3 of [1] w e ha v e e xpression (4) f or h ( t ) . The f ollo wing e xample illus- tr ates the tr ansf er function of a non -autonomous Riesz-spectr al system. The e xample is modified from Example 4.3.11 of [2]. Example 5. Consider the controlled non-autonomous heat equation on the inter v al [0 ; 1] , @ x @ t ( t; ) = a ( t ) @ 2 x @ 2 ( t; ) + 2 b ( t ) u ( t ) [ 1 2 ; 1] ( ) ; 0 < < 1 ; t 0 ; @ x @ ( t; 0) = @ x @ ( t; 1) = 0 ; y ( t ) = 2 c ( t ) Z 1 = 2 0 x ( t; ) d ; (5) where a : R + ! R is a boundedly unif or mly contin uous positiv e function and b : R + ! C is a bounded contin uous function. W e v er ify the tr ansf er function and impulse response of the go v er ned system. Setting X = L 2 [0 ; 1] and U = Y = C , the prob lem (5) is a non-autonomous Riesz-spectr al system: _ x ( t ) = A ( t ) x ( t ) + B ( t ) u ( t ) ; y ( t ) = C ( t ) x ( t ) ; t 0 ; (6) where A ( t ) = a ( t ) A with Ax = d 2 x d 2 on domain D = D ( A ) = f x 2 X : x; dx d are absolutely contin uous ; d 2 x d 2 2 X ; dx d (0) = dx d x (1) = 0 g ; B ( t ) u ( t ) = 2 b ( t ) u ( t ) [ 1 2 ; 1] ( ) ; and C ( t ) x ( t ) = 2 c ( t ) Z 1 = 2 0 x ( t; ) d : The eigen v alues and eigen v ectors of A are f 0 ; n 2 2 : n 2 N g and f 1 ; p 2 cos( n ) : n 2 N g , respectiv ely . It is easy to sho w that A is a self-adjoint Riesz-spectr al oper ator with its Riesz basis f 1 ; p 2 cos( n ) : n 2 N g . In this case w e ha v e B ( t ) x = Z 1 0 x [ 1 2 ; 1] ( ) d ; TELK OMNIKA V ol. 16, No . 6, December 2018 : 3024 3033 Evaluation Warning : The document was created with Spire.PDF for Python.
TELK OMNIKA ISSN: 1693-6930 3028 f or all 2 X and t 0 . In vir tue of (3) e v aluating at t , the tr ansf er function of the system (6) is giv en b y G ( s )( t ) = 1 s + 1 X n =1 2 a ( t )[cos( n ) 1] [ s + a ( t ) n 2 2 ]( n ) 2 ; and e v aluating at s in (4), w e ha v e the impulse response h ( t )( s ) = 1 + 1 X n =1 2[cos( n ) 1] ( n ) 2 e ( n ) 2 a ( s ) t ; on t 0 and z ero else where , respectiv ely . 2.2. Stabilizability W e first recall the definition of unif or mly e xponentially stab le C 0 -quasi semig roup that ref ers to [21, 23]. The concept pla y an impor tant role in char acter izing stabilizability of the non- autonomous Riesz-spectr al systems . Definition 6. A C 0 -quasi semig roup R ( t; s ) is said to be: (a) unif or mly e xponentially stab le on a Banach space X if there e xist constants > 0 and N 1 such that k R ( t; s ) x k N e s k x k ; (7) f or all t; s 0 and x 2 X ; (b) -unif or mly e xponentially stab le on a Banach space X if (7) holds f or < . A constant is called deca y r ate and the suprem um o v er all possib le v alues of is called stability margin of R ( t; s ) . Indeed, the stability margin is min us of the unif or m g ro wth bound ! 0 ( R ) defined ! 0 ( R ) = inf t 0 ! 0 ( t ) ; where ! 0 ( t ) = inf s> 0 1 s log k R ( t; s ) k . W e see that R ( t; s ) is -unif or mly e xponentially stab le if its stability margin is at least . W e giv e tw o preliminar y results which are urgent in discussing the stabilizability . Theorem 7. Let R ( t; s ) be a C 0 -quasi semig roup on a Banach space X . The R ( t; s ) is unif or mly e xponentially stab le on X if and only if ! 0 ( R ) < 0 . Proof . By taking log on (7), w e ha v e the asser tion. Theorem 8. Let A ( t ) be an infinitesimal gener ator of C 0 -quasi semig roup R ( t; s ) on a Banach space X . If B ( ) 2 L 1 ( R + ; L s ( X )) , then there e xists a uniquely C 0 -quasi semig roup R B ( t; s ) with its infinitesimal gener ator A ( t ) + B ( t ) such that R B ( r ; t ) x = R ( r ; t ) x + Z t 0 R ( r + s; t s ) B ( r + s ) R B ( r ; s ) xds; (8) f or all t; r ; s 0 with t s and x 2 X . Moreo v er , if k R ( r ; t ) k M ( t ) , then k R B ( r ; t ) k M ( t ) e k B k M ( t ) t : Proof . W e define R 0 ( r ; t ) x = R ( r ; t ) x; R n ( r ; t ) x = Z t 0 R ( r + s; t s ) B ( r + s ) R n 1 ( r ; s ) xds; (9) T r ansf er Function, Stabilizability , and Detectability of Non-A utonomous ... (Sutr ima) Evaluation Warning : The document was created with Spire.PDF for Python.
3029 ISSN: 1693-6930 f or all t; r ; s 0 with t s , x 2 X , and n 2 N , and R B ( r ; t ) = 1 X n =0 R n ( r ; t ) ; (10) F ollo wing the proof of Theorem 2.4 of [19], w e obtain the asser tions . The in v estigations of stabiliz- ability and detectability of the non-autonomous Riesz-spectr al systems are gener alizations of the concepts of stabilizability and detectability f or the autonomous systems that had been de v eloped b y Cur tain and Zw ar t [2]. Definition 9. The non-autonomous Riesz-spectr al system ( A ( t ) ; B ( t ) ; C ( t )) is said to be: (a) stabilizab le if there e xsits an oper ator F 2 L 1 ( R + ; L s ( X ; U )) such that A ( t ) + B ( t ) F ( t ) , t 0 , is an infinitesima l gener ator of a unif or mly e xponentially stab le C 0 -quasi semig roup R B F ( t; s ) . The oper ator F is called a stabilizing f eedbac k oper ator ; (b) detectab le if there e xists an oper ator K 2 L 1 ( R + ; L s ( Y ; X )) such that A ( t ) + K ( t ) C ( t ) , t 0 , is an infinitesimal gener ator of a unif or mly e xponentially stab le C 0 -quasi semig roup R K C ( t; s ) . The oper ator K is called an output injection oper ator . If the quasi semig roup R B F ( t; s ) is -unif or mly e xponentially stab le , w e sa y that the system ( A ( t ) ; B ( t ) ; ) is -stab lizab le . If R K C ( t; s ) is -unif or mly e xponentially stab le , w e sa y that the system ( A ( t ) ; ; C ( t )) is -detectab le . F or 2 R , w e can decompose the spectr um of A in comple x plane into tw o distinct par ts: + ( A ) := ( A ) \ C + ; C + = f 2 C : Re ( ) > g ( A ) := ( A ) \ C ; C = f 2 C : Re ( ) < g : In the autonomous case , if B has finite-r ank and the system ( A; B ; ) is stabilizab le , then w e can decompose the spectr um of A into a -stab le par t and a -unstab le par t which compr ises eigen v alues with finite m ultiplicity . In other w ord, A has at most finitely man y eigen v alues in C + . W e shall apply the decom position of the spectr um to the non-autonomous Riesz-spectr al systems . Definition 10. An oper ator A is said to be satisfying the spectr um decomposition assumption at if + ( A ) is bounded and separ ated from ( A ) in such w a y that a rectifiab le , simple , closed cur v e , , can be dr a wn so as to enclose an open set containing + ( A ) in its inter ior and ( A ) in its e xter ior . According to Definition 10, w e ha v e that classes of Riesz-spectr al o per ators with a pure point spectr um and only finitely man y eigen v alues in + ( A ) satisfy the spectr um decomposition assumption. F or e v er y t 0 w e define the spectr al projection P ( t ) on X b y P ( t ) x = 1 2 i Z ( I A ( t )) 1 xd; (11) f or all x 2 X , where is tr a v ersed once in the positiv e direction (countercloc kwise). By this oper ator , w e can decompose an y Hilber t space X to be: X = X + X ; where X + := P ( t ) X and X := ( I P ( t )) X : (12) By this decomposition, w e denote A ( t ) = A + ( t ) 0 0 A ( t ) ; R ( t; s ) = R + ( t; s ) 0 0 R ( t; s ) (13) B ( t ) = B + ( t ) B ( t ) ; C ( t ) = C + ( t ) C ( t ) ; (14) TELK OMNIKA V ol. 16, No . 6, December 2018 : 3024 3033 Evaluation Warning : The document was created with Spire.PDF for Python.
TELK OMNIKA ISSN: 1693-6930 3030 where B + ( t ) = P ( t ) B ( t ) 2 L ( U ; X + ) , B ( t ) = ( I P ( t )) B ( t ) 2 L ( U ; X ) , C + ( t ) = C ( t ) P ( t ) 2 L ( X + ; Y ) , and C ( t ) = C ( t )( I P ( t )) 2 L ( X ; Y ) . In vir tue of the decomposition, w e can e x- press the system ( A ( t ) ; B ( t ) ; C ( t )) as the v ector sum of the tw o subsystems: ( A + ( t ) ; B + ( t ) ; C + ( t )) on X + and ( A ( t ) ; B ( t ) ; C ( t )) on X . In par ticular , if the input oper ator B ( ) has finite-r ank, then the subsystem ( A + ( t ) ; B + ( t ) ; ) has a finite dimension. Theorem 11. Let ( A ( t ) ; B ( t ) ; ) be a non-autonomous Riesz-spectr al system on the state space X where B ( ) has a finite r ank. If A satisfies the spectr um decomp osition assumption at , X + has a finite dimension, R ( t; s ) is -unif or mly e xponent ially stab le , and subsystem ( A + ( t ) ; B + ( t ) ; ) is controllab le , then the system ( A ( t ) ; B ( t ) ; ) -stabilizab le . In this case , a -stabilizing f eed- bac k oper ator is giv en b y F ( t ) = F 0 ( t ) P ( t ) , where F 0 is a -stabilizing f eedbac k oper ator f or ( A + ( t ) ; B + ( t ) ; ) . Proof . Since ( A + ( t ) ; B + ( t ) ; ) is controllab le , there e xists a f eedbac k op er ator balik F 0 ( ) 2 L 1 ( R + ; L s ( X + ; U ) such that the spectr um of A + ( t ) + B + ( t ) F 0 ( t ) in C f or a ll t 0 . W e choose a f eedbac k oper ator F ( ) = ( F 0 ( ) ; 0) 2 L 1 ( R + ; L s ( X ; U )) f or the system ( A ( t ) ; B ( t ) ; ) . According to Theorem 8, the per turbed oper ator A ( t ) + B ( t ) F ( t ) = A + ( t ) + B + ( t ) F 0 ( t ) 0 B ( t ) F 0 ( t ) A ( t ) is the infinitesimal gener ator of a C 0 -quasi semig roup R ( t; s ) . Moreo v er , if 1 is the g ro wth bound of R ( t; s ) , then 1 is the maxim um of that of the quasi semig roups gener ated b y A + ( t ) + B + ( t ) F 0 ( t ) and A ( t ) . Theref ore , 1 < , i.e . the system ( A ( t ) ; B ( t ) ; ) is -stabilizab le . By the dual proper ty betw een the stabilizability and detectability , Theorem 11 pro vides the sufficiency f or the -detectability . Theorem 12. Let ( A ( t ) ; ; C ( t )) be the non-autonomous Riesz-spectr al system on the state space X where C ( ) has a finite r ank. If A satisfies the spectr um decomposition assumption at , X + has a finite dimension, R ( t; s ) is -unif or mly e xponentially stab le , and ( A + ( t ) ; ; C + ( t )) is obser v ab le , then ( A ( t ) ; ; C ( t )) is -detectab le . In this case , -stabilizing output injection oper ator is giv en b y K ( t ) = i ( t ) K 0 ( t ) , where K 0 is an oper ator such that A + ( t ) + K 0 ( t ) C + ( t ) is the infinitesimal gener ator of the -unif or mly e xponentially stab le C 0 -quasi semig roup and i ( t ) is an injection oper ator from X + to X . Proof . By the dual concept, w e ha v e that the system ( A ( t ) ; ; C ( t )) is -detectability if and only if ( A ( t ) ; C ( t ) ; ) is -stabilizability . F rom Theorem 11, A satisfies the spectr um decomposition assumption at . The corresponding spectr al projection is giv en b y P ( t ) x = 1 2 i Z ( I A ( t )) 1 xd; f or all x 2 X . W e can choose such that is symmetr ic with respect to the real axis . Hence , the decomposition of ( A ( t ) ; C ( t ) ; ) is the adjoint of the decomposition of ( A ( t ) ; ; C ( t )) . By the dual argument, w e ha v e the required results . Example 13. Consider the non-autonomous Riesz-spectr al system in Example 5. Using all of the ag reement there , w e sho w that the system is stabilizab le and detectab le . W e ha v e the spectr al decomposition of A ( t ) : A ( t ) x = a ( t ) 1 X n =1 ( n ) 2 h x; n i n f or x 2 D ; where n ( ) = p 2 cos( n ) , and the f amily of the oper ators gener ates a C 0 -quasi semig roup R ( t; s ) giv en b y R ( t; s ) x = h x; 1 i 1 + 1 X n =1 e ( n ) 2 ( g ( t + s ) g ( t )) h x; n i n ; T r ansf er Function, Stabilizability , and Detectability of Non-A utonomous ... (Sutr ima) Evaluation Warning : The document was created with Spire.PDF for Python.
3031 ISSN: 1693-6930 where g ( t ) = R t 0 a ( s ) ds . Since the set of eigen v alues of A has an upper bound, then A satisfies the spectr um assumption at an y real . Suppose w e choose = 2 . In this case w e ha v e + 2 ( A ) = f 0 g and so there e xists a closed simple cur v e 2 that encloses the eigen v alue 0 . F or an y x 2 X and t 0 , Cauch y’ s Theorem giv es P 2 ( t ) x = 1 2 i Z 2 ( I a ( t ) A ) 1 xd = 1 2 i Z 2 1 h x; 1 i d = h x; 1 i : In vir tue of (12), w e ha v e that X + 2 has dimension 1 . Moreo v er , in the subspace w e ha v e ( A + 2 ( t ) ; B + 2 ( t ) ; C + 2 ( t )) = (0 ; I ; I ) which are controllab le and obser v ab le . There e xists F 0 ( ) 2 L 1 ( R + ; L s ( X + 2 ; U )) such that the spectr um of A + 2 ( t ) + B + 2 ( t ) F 0 ( t ) in C 2 f or all t 0 . W e can choose F 0 ( t ) x = 3 h x; 1 i f or all x 2 X + 2 . Theorem 11 concludes that ( A ( t ) ; B ( t ) ; C ( t )) is ( 2) - stabilizab le with the f eedbac k oper ator u = F ( t ) z , where F ( t ) = F 0 ( t ) 0 = 3 I 0 . By the duality , Theorem 12 sho ws that the system ( A ( t ) ; ; C ( t )) is ( 2) -detectab le wit h output injection oper ator K ( t ) = 3 I 0 such tha t K ( t ) y = 3 y 1 . The deca y constants of the quasi semig roup gener ated b y A ( t ) + B ( t ) F ( t ) and A ( t ) + L ( t ) C ( t ) are 3 . The f ollo wing theorem is a similar result with Theorem 13 of [1] f or controllability and obser v ability of the non-autonomous Riesz-spectr al systems on Hilber t spaces . Theorem 14. Let ( A ( t ) ; B ( t ) ; C ( t )) be a non-autonomous Riesz-spectr al system. Necessar y and sufficient conditions f or ( A ( t ) ; B ( t ) ; ) is -stabilizab le are that ther e e xists an > 0 such that + ( A ) compr ises at most finitely man y eigen v alues and rank ( h b 1 ( t ) ; ' n i ; : : : ; h b m ( t ) ; ' n i ) = 1 ; (15) f or all n such that n 2 + ( A ) and t 0 . Necessar y and sufficient conditions f or ( A ( t ) ; ; C ( t )) is -detectab le are that there e xists an > 0 such that + ( A ) compr ises at most finitely man y eigen v alues and rank ( h n ; c 1 ( t ) i ; : : : ; h n ; c k ( t ) i ) = 1 ; (16) f or all n such that n 2 + ( A ) and t 0 . Proof . W e only need to pro v e necessar y and sufficient conditions f or -stabilizability . F or - detectability f ollo ws the dual argument of the system. Sufficiency f or -stabilizability . Since + ( A ) only contains at most finitely man y eigen v alues of A , then X + has a finite dimension and A satisfies the spectr um decomposition assumption at . By condition (a) of Theorem 3 of [1] and Cauch y’ s Theorem, f or an y t 0 w e ha v e P ( t ) x = 1 a ( t ) X n 2 + h x;   n i n : Since A ( t ) and A ha v e common eigen v ectors f or e v er y t 0 , again Theorem 3 of [1] giv es X + = span n 2 + f n g ; X = span n 2 f n g ; R ( t; s ) = X n 2 e n ( g ( t + s ) g ( t ) h x;   n i n ; A + ( t ) x = a ( t ) X n 2 + n h x;   n i n ; and B + ( t ) u = X n 2 + h B ( t ) u;   n i n ; (17) where g ( t ) = R t 0 a ( s ) ds . This result sho ws that R ( t; s ) is a C 0 -quasi semig roup correspond- ing to the Riesz-spectr al oper ator A on X . Consequently , ! 0 ( R ) < 0 and Theorem 7 states that R ( t; s ) is -unif or mly e xponentially stab le . W e need pro v e that ( A + ( t ) ; B + ( t ) ; ) is controllab le . The reachibility subspace of ( A + ( t ) ; B + ( t ) ; ) is the smallest A + -in v ar iant TELK OMNIKA V ol. 16, No . 6, December 2018 : 3024 3033 Evaluation Warning : The document was created with Spire.PDF for Python.
TELK OMNIKA ISSN: 1693-6930 3032 subspace of X + which contains ran B + ( t ) . In vir tue of Lemma 2.5.6 of [2], this subspace is spanned b y the eigen v ectors of A + . Hence , if this subspace does not equal with the state space X + , then there e xists a j 2 + ( A ) such that j = 2 X + . Theref ore , its bior thogonal element   j is or thogonal to the reachibility subspace , in par ticular h B ( t ) u;   j i = 0 f or e v er y u 2 C m and t 0 . This contr adicts to (15), and so ( A + ( t ) ; B + ( t ) ; ) is controllab le . Thus , Theorem 11 sho ws that the system ( A ( t ) ; B ( t ) ; ) is -stabilizab le . Necessity f or -stabilizability . F rom Def- inition 6 and Definition 9, if ( A ( t ) ; B ( t ) ; ) is -stabilizab le , then there e xists > 0 such that the system is also ( ) -stabilizab le . In vir tue of Theorem 11, A satisfies the spectr um decompo- sition assumption at . Moreo v er , the subspace X + is R ( t; s ) -in v ar iant. Lemma 2.5.8 of [2] giv es X + = span n 2 J f n g : Since X + has a finite dimension, then J contains at most finitely man y elements . So the spec- tr um of A + = A j X + is contained in C + and the spectr um of A = A j X is contained in C . This concludes that the inde x set J equals with the set f n 2 N : n 2 + ( A ) g . Thus , + ( A ) g compr ises at most finitely man y eigen v alues . By (17) and Theorem 11, w e conclude that ( A + ( t ) ; B + ( t ) ; ) is controllab le . Suppose that the condition (15) does not hold. There e xists n 2 + ( A ) such tha t h B ( t ) u;   n i = 0 f or all u 2 C m . This states that the reachibility subspace of ( A + ( t ) ; B + ( t ) ; ) does not equal to X + , that is ( A + ( t ) ; B + ( t ) ; ) is not con- trollab le . This giv es a contr adiction. In pr actice , Theorem 14 is more applicab le than Theorem 11 and 12 in char acter izing the stabilizability and detectability of the non-autonomous Riesz-spectr al systems . W e retur n to Example 13. F or = 2 , w e can choose = 1 such that + 3 ( A ) = f 0 g and the corresponding eigen v ector is 0 ( ) = 1 . In this case , it is ob vious that Z 1 0 b ( t ) 0 ( ) d 6 = 0 and Z 1 0 c ( t ) 0 ( ) d 6 = 0 ; f or all t 0 , i.e . the conditions (15) and (16) are confir med. Hence , the system ( A ( t ) ; B ( t ) ; C ( t )) are stabilizab le and detectab le . 3. Conc lusion The concepts of the tr ansf er function, stabilizability , and detectability of the autonomous systems can be gener aliz ed to the non-autonomous Riesz-spectr al systems . The results are alter nativ e consider ations in analyzing the related control prob lems . There are oppor tunities to gener aliz e these results to the gener ally non-autonomous systems inclu ding the time-dependent domain. Ac kno wledg ement The authors are g r ateful to Research Institution and Comm unity Ser vice of Univ ersitas Sebelas Maret Sur akar ta f or funding and to the re vie w ers f or helpful comments . Ref erences [1] Sutr i ma, Indr ati CR, and Ar y ati L. Controllability and Obser v ability of Non-A utonomous Riesz- Spectr al Systems . Abstr act and Applied Analysis . 2018; Ar ticel ID 4210135, https://doi.- org/10.1155/2018/4210135, 10 pages . [2] Cur tai n RF and Zw ar t HJ . Introduction to Infinite-Dimensional Linear Systems Theor y . Ne w Y or k: Spr inger , 1995. [3] Chicone C and Lat ushkin Y . Ev olution Semig roups in Dy amical Systems and Diff erential Equations . Rhode Island: Amer icans Mathematical Society , 1999. [4] W eiss G. T r ansf er Functions of Regular Linear Systems . P ar t I: Char acter izations of Regular ity . T r ans- actions of The Amer ican Mathenatical Society . 1994; 342: 827–854. T r ansf er Function, Stabilizability , and Detectability of Non-A utonomous ... (Sutr ima) Evaluation Warning : The document was created with Spire.PDF for Python.
3033 ISSN: 1693-6930 [5] Staff ans O and W eiss G. T r ansf er Functions of Regular Linear Systems . P ar t II: The System Oper ator and The Lax-Phillips Semig roup . T r ansactions of The Amer ican Mathenatical Society . 2002; 354: 3229– 3262. [6] Staff ans O and W eiss G. T r ansf er Functions of Regular Linear Systems . P ar t III: In v ersion and Duality . Integ r al Equations and Oper ator Theor y . 2004; 49: 517—558. [7] P ar ting ton JR and ¨ Unalmis ¸ B . On the Representation of Shift-In v ar iant Oper ators b y T r ansf er Functions . Systems and Control Letters . 1998; 33: 25—30. [8] Kalman RE, Ho YC , and Narenda KS . Controllability of Linear Dynamical Systems . Diff errential Equa- tions . 1963; 1: 189–213. [9] Phat VN. W eak Asymptotic Stabilizability of Discrete-Time Inclusion Giv en b y Set-V alued Oper ator . Jour- nal of Mathematical Analysis and Applicationss . 1996; 202: 353–369. [10] Phat VN. Global Stabilization f or Linear Contin uous Time -v ar ying Systems . Apllied Mathematics and Computation . 2006; 175: 1730–1743. [11] Phat VN and Ha QP . Ne w Char acter ization of Controllability via Stabilizability and Riccati Equation f or L TV Systems . IMA Jour nal of Mathematical Control and Inf or mation . 2008; 25: 419–429. [12] W onham WM. On P ole Assignment in Multi- Input Contor llab le Linear Systems . IEEE T r ansactions on A utomatic Control . 1967; 12: 660–665. [13] Ik eda M, Maeda H, and K odama S . Stabilization of Linear Systems . SIAM Jour nal on Control . 1972; 10: 716–729. [14] Phat VN and Kiet TT . On the L y apuno v equation in Banach Spaces and Applications to Control Prob- lems . Inter national Jour nal of Mathematics and Mathematical Science . 2002; 29: 155–166. [15] Guo F , Zhang Q, Huang F . W ell-P ose dness and Admissib le Stabilizability f or Pr itchard-Salamon Sys- tems . Applied Mathematica Letters . 2003; 16: 65–70. [16] Rabah R, Skly ar G, and Bar kha y e v P . Exact Null Controllability , Complete Stability and Contin uous Final Obser v ability of Neutr al T ype Systems . Inter national Jour nal of Applied Mathematics and Computer Science . 2017; 27(3): 489—499. [17] Hinr ichsen D and Pr itchard AJ . Rob ust Stability of Linear Ev olution Oper ator on Ba nach Spaces . SIAM Jour nal of Control and Optimization . 1994; 32: 1503–1541. [18] Niamsup P and Phat VN. Linear Ti me-V ar ying Systems in Hilber t Spaces: Exact Controllability Implies Complete Stabilizability . Thai Jour nal of Mathematics . 2009; 7: 189–199. [19] Leiv a H and Bar cenas D . Quasi-semig roups , Ev olut ion Equation and Controllability . Notas de Matem- aticas . 1991; 109: 1–29. [20] Sutr ima, Indr at i CR, Ar y ati L, and Mardiy ana. The Fundamental Proper ties of Quasi-Semig roups . Jour- nal of Ph ysics: Conf erence Ser ies . 2017; Ar ticel ID 855(2017)012052: 9 pages . [21] Sutr ima, Indr ati CR, and Ar y ati L. Stability of C 0 -quasi semig rou ps in Banach spaces . Jour nal of Ph ysics: Conf erence Ser ies . 2017; Ar ticel ID 943(2017)012021: 14 pages . [22] Barcenas D , Leiv a H, and Mo y a A T . The Dual Quasi-Semig roup and Controllability of Ev olut ion Equa- tions . Jour nal of Mathematical Analysis and Applications . 2006; 320: 691–702. [23] Megan M and Cuc V . On Exponential Stability of C 0 -Quasi-Semig roups in Banach Spaces . Le Matem- atiche . 1999; Liv (1999): 229–241. TELK OMNIKA V ol. 16, No . 6, December 2018 : 3024 3033 Evaluation Warning : The document was created with Spire.PDF for Python.