Nonlinear Dynamics and Systems Theory, 5(2) (2005) 157167

Stability of Dynamical Systems in Metric Space

A.A Martynyuk
Stability of Processes Department, S.P.Timoshnko Institute of Mechanics, National Academy of Sciences of Ukraine, 03057, Nesterov str., 3, 03057, Kiev-57, Ukraine Received: October 21, 2004; Revised: March 15, 2005

Abstract: In the paper a new approach is developed for stability analysis of motions of dynamical systems dened on metric space using matrix-valued preserving mappings. These results are applicable to a much larger class of systems then existing results, including dynamical systems that cannot be determined by the usual classical equations and inequalities. We apply our results in the stability analysis of hybrid systems in general and twocomponent hybrid systems. Keywords: Dynamical system; metric space; hybrid system; asymptotic stability; stability matrix-valued preserving mapping. Mathematics Subject Classication (2000): 34G20, 35B35, 37K45, 37L15,

93A15, 93D30.

1 Introduction This paper presents an approach to stability analysis of dynamical systems determined in metric space. The method of analysis of invariant sets of dynamical systems was proposed by Zubov [11] on the basis of generalized direct Liapunov method. In our approach a generalized comparison principle is used together with the idea of multicomponent mapping (cf. matrix-valued Liapunov functions [5, 6]). In the present paper, we rst developed a matrix-valued preserving mapping for stability analysis of general dynamical systems dened on metric space. To accomplish this, we utilize, as in our earlier work (see [7]), stability preserving matrix-valued mappings. We use the above results to establish the principal Lyapunov theorems for dynamical systems on metric space. Finally, we analyze a class of hybrid systems, using some of these results with particular application to two-component hybrid system.

Corresponding author: anmart@stability.kiev.ua 157

c 2005 Informath Publishing Group/1562-8353 (print)/1813-7385 (online)/www.e-ndst.kiev.ua

158

A.A.MARTYNYUK

2 Basic Concepts and Denitions Let X be a set of elements (no matter of what nature) and a measure (x, y) be dened for x, y X . The Denitions 2.1 2.6 presented here follow in the spirit of the works [1, 2, 9, 11] even if some of the formulations are dierent. Denition 2.1 (X, ) is a metric space if the following conditions are fullled for any x, y, z X: (1) (x, y) 0, (2) (x, y) = 0 x = y, (3) (x, y) = (y, x), (4) (x, y) (x, z) + (y, z), and, additionally, for any X0 X, (x, X0 ) = inf (x, y).
yX0

Denition 2.2 A metric space (T, ) is called a temporal space if: (1) T is completely ordered by the ordering <; (2) T has a minimum element tmin T, i.e. tmin < t for any t T, such that t = tmin ; (3) for any t1 , t2 , t3 T such that t1 < t2 < t3 it holds that (t1 , t3 ) = (t1 , t2 ) + (t2 , t3 ); (4) T is unbounded from above; i.e., for any M > 0, there exists t T such that (t, tmin ) > M . Denition 2.3 Let (X, ) be a metric space with a subset A X and let (T, ) be a temporal space with subset T R+ . A mapping p(, a, 0 ) : Ta,0 X is called a motion if p(0 , a, 0 ) = a, where a A, 0 T and Ta,0 = [0 , 1 ) T for 1 > 0 , with 1 being a nite value or innity. Denition 2.4 Let Ta,0 {a} {0 } X denote the set of mappings of Ta,0 {a} {0 } into X, = (a,0 )AT (Ta,0 {a} {0 } X) and S be a family of motions; i.e., S {p( , a, 0 ) : p(0 , a, 0 ) = a}. Then the four-tuple (T, X, A, S) of sets and spaces is called a dynamical system. Note that Denition 2.4 possesses some generality. Specically, (i) if X is a normed linear space and every motion p(, a, 0 ) is assumed to be continuous with respect to , a and 0 , then Denition 2.4 corresponds to the concept of a family of motions in Hahn [3]; (ii) under some additional conditions imposed on p(, a, 0 ) (see [11], pp.183184), Denition 2.4 reduces to the concept of a general system introduced by Zubov. In what follows, we consider dynamical systems satisfying the standard semigroup property p(2 , p(1 , a, 0 )) = p(2 + 1 , a, 0 ) for all a A and any 1 , 2 R+ .

NONLINEAR DYNAMICS AND SYSTEMS THEORY, 5(2) (2005) 157167

159

Denition 2.5 A dynamical system (R+ , X, A, S) is called continuous if any of its motions p S is continuous; i.e., any mapping p( , a, 0 ) : Ta,0 X is continuous. Let (X1 , 1 ) and (X2 , 2 ) be metric spaces, and let (R+ , X1 , A1 , S1 ) be a continuous dynamical system. We assume that the space X1 is a Descartes product of spaces X11 , X12 , . . . , X1m , on which the multicomponent mapping (see [7]) U (t, x) : T X11 X12 . . . X1m X2 (1)

is acting. It is assumed that the mapping U : R+ X11 X12 . . .X1m X2 has the following properties: for any motion p( , a, t0 ) S1 , the function q( , b, t0 ) = U ( , p( , a, t0 ), ) with initial value b = U (t0 , a) is another motion for which Ta,t0 = Tb,t0 , b A2 X2 . Let S2 denote the set of motions q determined by initial values a A1 and t0 R+ . Then (R+ , X2 , A2 , S2 ) is a continuous dynamical system. The mapping given by (1) induces a mapping of S1 into S2 , denoted by M; i.e., S2 = M(S1 ). Moreover, we denote by M1 A1 and M2 A2 some sets invariant under S1 and S2 , respectively. The set M2 is then dened by the formula M2 = U (R+ M1 ) = {x2 X2 : x2 = U (t , x1 ) for some x1 M1 and t R+ ,}. (2)

In what follows, we consider continuous dynamical systems (R+ , X1 , A1 , S1 ) and (R+ , X2 , A2 , S2 ) with invariant sets M1 A1 and M2 A2 , respectively. Denition 2.6 Multicomponent mapping (1) U : R+ X11 X12 , . . . , X1m X2 preserves some type of stability of a continuous dynamical system if the sets S2 = M(S1 ) {q( , b, t0 ) : q(t, b, t0 ) = U (t, p(t, a, t0 )), Rm , a A1 , b = U (t0 , a), t0 R+ } (4) (3)

p( , a, t0 ) S1 , Tb,t0 = Ta,t0 ,

and M2 (see formula (2)) satisfy the following conditions: (1) the invariance of (S1 , M1 ) is equivalent to the invariance of (S2 , M2 ); (2) some type of stability of (S1 , M1 ) is equivalent to the same type of stability of (S2 , M2 ).

3 Sucient Conditions for Stability of Dynamical System Note that the mapping U induces a mapping M : S1 S2 , that preserves some types of stability of (S1 , M1 ) and (S2 , U (R+ M1 )).

160

A.A.MARTYNYUK

Theorem 3.1 Let a dynamical system (R+ , X1 , A1 , S1 ) be assigned a comparison system (R+ , X2 , A2 , S2 ) by means of a multicomponent mapping U (t, p) : R+ X1 X2 . Suppose that there exist closed sets Mi Ai , i = 1, 2, and following conditions are fullled: (1) for M(S1 ) and S2 , M(S1 ) = S2 ; (2) there exist constant m m matrix Ai , i = 1, 2, and comparison functions 1 , 2 K such that
T T 1 A1 1 2 (U (t, p), M2 ) 2 A2 2

(5)

for all p X1 and t R+ , where 1 = (11 (1 (p, M1 )), . . . , 1m (1 (p, M1 )))T , 2 = (21 (1 (p, M1 )), . . . , 2m (1 (p, M1 )))T . Here, 1 and 2 are metrics dened on X1 and X2 , respectively. If the matrices Ai , i = 1, 2, are positive denite, then the following is true: (1) the invariance of (S2 , M2 ) implies the invariance of (S1 , M1 ); (2) the stability, uniform stability, asymptotic stability, or uniform asymptotic stability of (S2 , M2 ) implies the respective type of stability of (S1 , M1 ); T (3) if in estimate (5) 1 A1 1 = a(1 (p, M1 ))b , where a > 0 and b > 0, then the exponential stability of (S2 , M2 ) implies the exponential stability of (S1 , M1 ). Proof of item (1) Let (S2 , M2 ) be an invariant pair. Then, for any a M1 and any motion p( ; a, t0 ) S1 , we nd that q( ; b, t0 ) = U (t, p( ; a, t0 )) S2 , where b = U (t0 , a). This follows from condition (1) in Theorem 3.1 and from the denition of M(S1 ) by formula (4). Moreover, the invariance of (S2 , M2 ) implies that q(t; b, t0 ) = U (t, p(t; a, t0 )) M2 for all t Tb,t0 = Ta,t0 . Since M1 and M2 are closed and the matrices A1 and A2 are positive denite and satisfy (5), we conclude that p(t; a, t0 ) M1 for all t Ta,t0 . This implies the invariance of (S1 , M1 ). Proof of item (2) Assume that (S2 , M2 ) is stable. Then, by the denition of stability, for any 2 > 0 and t0 R+ , there exists 2 = 2 (t0 , 2 ) > 0 such that 2 (q(t; b, t0 ), M2 ) < 2 for all q( ; b, t0 ) S2 and all t Tb,t0 whenever 2 (b, M2 ) < 2 (t0 , 2 ). Estimates (5) can be transformed into m (A1 )1 (1 (p, M1 )) 2 (U (t, p), M2 ) M (A2 )2 (1 (p, M1 )). (6)

Here m (A1 ) > 0 and M (A2 ) > 0 are the minimum and maximum eigenvalues of the positive denite matrices A1 and A2 , and 1 , 2 K are such that
T 1 (1 (p, M1 ))1 (1 (p, M1 )) 1 (1 (p, M1 ))

and
T 2 (1 (p, M1 ))2 (1 (p, M1 )) 2 (1 (p, M1 )).

NONLINEAR DYNAMICS AND SYSTEMS THEORY, 5(2) (2005) 157167

161

Since (S2 , M2 ) is stable, for any > 0 and any t0 R+ , we choose 2 = m (A1 )1 () 1 and 1 = 1 (A2 )2 (2 ). Assuming that 1 (a, M1 ) < 1 and taking into account (6), M we obtain 2 (b, M2 ) M (A2 )2 (1 (a, M1 )) < M (A2 )2 (1 )
1 = M (A2 )2 (1 (A2 )2 (2 )) = 2 . M

It follows that, for all motions q( ; b, t0 ) S2 , the estimate 2 (q(t; b, t0 ), M2 ) < 2 holds for all t Tb,t0 . Returning to estimates (6), we nd that, for all p( ; a, t0 ) S1 and all t Ta,t0 = Tb,t0 , where b = U (t0 , a), we have
1 1 (p(t; a, t0 ), M1 ) 1 (A1 )1 (2 (V (p(t; a, t0 )), M2 )) m 1 1 1 (A1 )1 (2 (q(t; b, t0 ), M2 )) 1 (A1 )1 (m (A1 )1 ()) = , m m

whenever 1 (a, M1 ) < 1 . It follows that (S1 , M1 ) is stable. It is well known that a system motion is asymptotically stable if it is stable and attracting. Assume that (S2 , M2 ) is attracting. Then, for any t0 R+ there exists 2 = 2 (t0 ) > 0 such that, for all q( ; b, t0 ) S2 , the limit relation
t

lim 2 (q(t; b, t0 ), M2 ) = 0,

holds true whenever 2 (b, M2 ) < 2 . In other words, for any 2 > 0, there exists = (2 , t0 , q) > 0 with q = q( ; b, t0 ) S2 such that 2 (q(t; b, t0 ), M2 ) < 2 for all t Tb,t0 + , whenever 2 (b, M2 ) < 2 . According to condition (1) in Theorem 3.1, for any motion p( ; a, t0 ) S1 , we set b = U (t0 , a). Then q( ; b, t0 ) = U (p( ; a, t0 )) S2 . Fur1 thermore, for any 1 > 0, we choose 2 = m (A1 )1 (1 ) and set 1 = 1 (A2 )2 (A2 ). M For any motion p( ; a, t0 ) S1 , we then have 2 (b, M2 ) M (A2 )2 (1 (a, M1 )) < M (A2 )2 (1 ) = 2 whenever 1 (a, M1 ) < 1 and t Ta,t0 + = Tb,t0 + . Hence, 2 (q(t; a, t0 ), M2 ) < 2 = m (A1 )1 (1 ) for all t Ta,t0 + . Returning to estimate (2), we nd that
1 1 1 (p(t; a, t0 ), M1 ) 1 (A1 )1 (2 (q(t; a, t0 ), M2 )) < 1 (A1 )1 (1 ), m m

i.e., (S1 , M1 ) is an attractive pair. Thus, if (S2 , M2 ) is asymptotically stable, then (S1 , M1 ) is asymptotically stable as well. The statements on uniform stability and uniform asymptotic stability are proved following the same scheme, but 2 and 2 are chosen to be independent of t0 R+ . Let us prove statement (3) of the theorem. Assume that (S2 , M2 ) is exponentially stable. Then there exists 2 > 0 and, for any 2 > 0, there exists 2 = 2 (2 ) > 0 such that for any motion q( ; b, t0 ) S2 and all t Tb,t0 2 (q(t; b, t0 ), M2 ) < 2 e2 (tt0 ) whenever 2 (b, M2 ) < 2 . According to condition (1) in Theorem 3.1, for any motion p( ; a, t0 ) S1 , there exists a motion q( ; b, t0 ) = U (p( ; a, t0 )) S2 , where b =

162

A.A.MARTYNYUK

U (t0 , a). Furthermore, for any 1 > 0, we choose 2 = ab . Let 1 = 2 /b and 1 1 1 = 1 (A2 )2 (2 ). For p(t; a, t0 ) M1 with 1 (a, M1 ) < 1 , in view of (6), we M obtain 2 (b, M2 ) M (A2 )2 (1 (a, M1 )) < M (A2 )2 (1 ) = 2 . Consequently, 2 (q(t; b, t0 ), M2 ) < 2 e2 (tt0 ) for all t Tb,t0 . According to the hypothesis of Theorem 3.1, we have to set
T 1 A1 1 = a(1 (p, M1 ))b

in (1.6.6). It is easy to see that 1 (p(t; a, t0 ), M1 ) < 2 a

1/b

2 b

(tt0 )

= 1 e1 (tt0 )

for all t Ta,t0 . Thus, (S1 , M1 ) is exponentially stable.

4 Stability Analysis of Hybrid System Many physical and technical problems of real world are modelled by mixed systems of equations and correlations. For example, in motion control theory the feedback consists of several interconnected blocks. These blocks are described by equations of dierent types. Such systems are called hybrid (see [9]). Under certain assumptions real hybrid system can correspond to the dynamical system (T, X, A, S) in metric space. Assume that (X, ) and (Xi , i ), i = 1, 2, . . . , m, are metric spaces. Let X = X1 X2 . . . Xm and there exist constants a1 , a2 > 0 such that
m

a1 (x, y)
i=1

i (xi , yi ) a2 (x, y)

(7)

for all x, y X, where x = (x1 , . . . , xm )T , y = (y1 , . . . , ym )T , xi Xi , yi Xi , i = 1, 2, . . . , m. Further on we will assume that

m

(x, y) =
i=1

i (xi , yi ).

(8)

Denition 4.1 (cf. [9] ) Dynamical system (T, X, A, S) is hybrid, if its metric space (X, ) consists of metric spaces (Xi , i ), i = 1, 2, . . . , m, where Xi are nontrivial unsplitted with metrics i (xi , yi ), and if there exist at least two metric spaces Xi and Xj , 1 i = j m, which are not isometric. The proposition below is necessary when the multicomponent mapping is made by matrix-valued functional.

NONLINEAR DYNAMICS AND SYSTEMS THEORY, 5(2) (2005) 157167

163

Proposition 4.1 Let multicomponent mapping U (t, x) : T X X2 be performed by matrix-valued functional U (t, x) = [vij (t, x)], i, j = 1, 2, . . . , m, for the elements of which: (a) vii C(R+ X, R+ ), i = 1, 2, . . . , m, vij C(R+ X, R) for all i = j and for all x X and t R+ ; (b) there exist comparison functions i1 , i2 of class K, positive constants cii > 0, cii > 0 and arbitrary constants cij R, cij R for i = j such that cii 2 (i (xi , Mi )) vii (t, x) cii 2 (i (xi , Mi )), i1 i2 cij i1 (i (xi , Mi ))j1 (j (xj , Mj )) vij (t, x) cij i2 (i (xi , Mi ))j2 (j (xj , Mj )) for all xi Xi , x X and t R+ . Then for the functional v(t, x, ) = T U (t, x), the bilateral inequality uT ((x, M ))H T CHu1 ((x, M )) v(t, x, ) 1 uT ((x, M ))H T CHu2 ((x, M )) 2 holds for all x X and t R+ , where H = diag (1 , 2 , . . . , m ), C = [cij ], C = [cij ], i, j = 1, 2, . . . , m, u1 () = (i1 (1 (x1 , M1 )), . . . , m1 (m (xm , Mm )))T , u2 () = (i2 (1 (x1 , M1 )), . . . , m2 (m (xm , Mm )))T . Proof Estimate (10) is obtained by direct substitution by estimates (b) of Proposition 4.1 in the expression
m m m R+ ,

(9)

i > 0,

(10)

v(t, x, ) =
i=1 j=1

vij (t, x)i j .

Theorem 4.1 Assume that behaviour of the hybrid system is correctly described by the dynamical system (T, X, A, S), where T = R+ , X = X1 . . . Xm and Xi are subspaces with metrics i , i = 1, 2, . . . , m. Let Mi Xi and M = M1 M2 . . . Mm be an invariant set. If (1) there exist functionals vij (t, x) mentioned in Proposition 4.1; m (2) given functionals vij (t, x) and a vector R+ , > 0, there exist bounded for all x X functions ij (x, ), i, j = 1, 2, . . . , m, and comparison functions i3 of class K such that D+ v(t, x, )|(S) uT (x, )u3 3 on system of motions S for all x X and t R+ , where u3 ((x, M )) = (13 (1 (x1 , M1 )), . . . , m3 (m (xm , Mm )))T .

164

A.A.MARTYNYUK

Then (a) If matrices B1 = H T CH, B2 = H T CH are positive denite and constant mm 1 matrix (T (x, ) + (x, )) for all x X is negative semidenite, then 2 the couple (S, M ) is uniformly stable. (b) If matrices B1 and B2 are positive denite and matrix is negative denite, then the couple (S, M ) is uniformly asymptotically stable. (c) If matrices B1 and B2 are positive denite, matrix is negative semidenite, the set M is bounded and the comparison functions i1 , i2 KR class i = 1, 2, . . . , m, then the family of motion S is uniformly bounded. (d) If in condition (c) the matrix is negative denite, then the family of motions S is uniform-ultimately bounded and the couple (S, M ) is uniformly asymptotically stable in the whole. (e) If there exist constants a1 , a2 , b, c such that a1 rb uT ((x, M ))H T Cu1 ((x, M )), 1 uT ((x, M ))H T Cu2 ((x, M )) a2 rb , 2 T 3 crb 3 for all r R+ , then the couple (S, M ) is exponentially stable in the whole. Proof Let us prove statement (a) of Theorem 4.1. Under condition (1) of Theorem 4.1 the functional v(t, x, ) is positive denite and decreascent because matrices B1 and B2 are positive denite. Under condition (2) of Theorem 4.1 the functional D+ v(t, x, ) on the system of motions S is negative semidenite due to restrictions on matrix . In this case the functional v(t, x, ) is nonincreasing for all t 0 along the system of motions S. Further, given > 0, we compute = inf v(t, x, ) for (x, M ) = . Because of
t0

estimate (9) we can nd by value the value > 0 such that for (x, M ) < the estimate v(t, x, ) < holds for all t 0. Now we show that the obtained value > 0 corresponds to the given > 0, i.e. for (x, M ) < the inequality (q(t; a, t0 ), M ) < holds for all t 0. Assume on the contrary, let there exist a motion q(t; a, t0 ) S such that for some value t R+ the inequality (q(t ; a, t0 ), M ) = takes place. Then we get v(t, q(t ; a, t0 ), ) , but due to condition (a) of Theorem 4.1 the functional v(t, x, ) is nonincreascent along the system of motions S. Therefore v(t, q(t; a, t0 ), ) v(t, x, ) < for any q(t; a, t0 ) S. The contradiction obtained shows that the system of motions S of the hybrid system is uniformly (S, M ) stable. The proof of statements (b) (e) of Theorem 4.1 is similar to that of statement (a) following the Liapunov (, )-technique.

NONLINEAR DYNAMICS AND SYSTEMS THEORY, 5(2) (2005) 157167

165

5 Stability Analysis of Two-Component Systems We consider a hybrid two-component system [4] dx = X(t, x(t)) + g1 (t, z, x(t), w(t, z)), x(t0 ) = x0 , dt w = L(t, x, /z)w + g2 (t, z, x(t), w(t, z)), t where w(t0 , z) = w0 (z), M (t, z, /z)w| = w1 (t, s), L : B1 B2 , s , Rk , (11) (12)

X : T0 U R n ,

M : B1 B3 ,

w0 B4 ,

L, M are some dierential operators and B1 , . . . , B4 are Banach spaces. A hybrid system (11) and (12) consists of the independent subsystems dx = X(t, x(t)), dt w = L(t, z, /z)w t and interconnection functions between them g1 = g1 (t, z, x, w) : T0 H Q Rn , g2 = g2 (t, z, x, w) : T0 H Q Rm . Let us introduce the assumptions on subsystems (13), (14) and interconnection functions between them. Assumption 5.1 There exist functions vij C(R+ H Q, R), i, j = 1, 2, vij (t, x, w) is locally Lipschitzian in x and w, functions of comparison i , i K, i = 1, 2, and positive constants ii , ii > 0, i = 1, 2, and arbitrary constants 12 , 12 such that 11 2 ( x ) v11 (t, x, w) 11 2 ( x ); 1 2
2 2 22 1 ( x ) v22 (t, x, w) 22 2 ( x );

(13) (14)

12 1 ( x )1 ( x ) v12 (t, x, w) 12 2 ( x )2 ( x ) for all x H, w Q and t 0. Lemma 5.1 If all conditions of Assumption 5.1 are fullled and the matrices A1 = A2 = 11 21 11 21 12 22 12 22 , , 12 = 21 , 12 = 21 ,

166

A.A.MARTYNYUK

are positive denite, then the function v(t, x, w) = T U (t, x, w), where = (1 , 2 ) , i > 0, is positive denite and decreasing. Proof We introduce the notations q = (2 ( x ), 2 ( w ))T , B= 1 0 0 2 .
T

(15)

r = (1 ( x ), 1 ( w ))T ,

Under the conditions of Assumption 5.1 for the function (15) the bilateral estimation rT B T A1 Br T U (t, x, w) q T B T A2 Bq (16) holds. By virtue of conditions of Lemma 5.1 it follows from the estimation (16) that the function v(t, x, w) is positive denite and decreasing. Assumption 5.2 There exist: (1) functions v11 (t, x), v22 (t, w) and functions v12 (t, x, w) = v21 (t, x, w); (2) constants ik , i = 1, 2, k = 1, . . . , 8, and functions 1 = 1 ( x ) and 2 = 2 ( w ) of the K-class such that
+ + 2 (a) Dt v11 (t, x) + Dx v11 (t, x)|X 11 1 ; + 2 (b) Dx v11 (t, x)|g1 12 1 + 13 1 3 ; + + 2 (c) Dt v22 (t, w) + Dw v22 (t, w)|L 21 2 ; + 2 (d) Dw v22 (t, w)|g2 22 2 + 23 1 2 ; + + 2 (e) Dt v12(t,x,w) + Dx v12 (t, x, w)|X 14 1 + 15 1 2 ; + 2 (f) Dw v12 (t, x, w)|L 24 1 + 25 1 2 ; + 2 2 (g) Dx v12 (t, x, w)|g1 16 1 + 17 1 2 + 18 2 ; + 2 2 (h) Dw v12 (t, x, w)|g2 26 1 + 27 1 2 + 28 2 .

Lemma 5.2 If all conditions of Assumption 5.2 are fullled and the matrix C= with the elements
2 c11 = 1 (11 + 12 ) + 21 2 (14 + 16 + 26 ), 2 c22 = 2 (21 + 22 ) + 21 2 (18 + 24 + 28 ), 1 2 2 c12 = (1 13 + 2 23 ) + 1 2 (15 + 25 + 17 + 27 ) 2 is negative denite, then the derivative

c11 c21

c12 c22

c12 = c21 ,

D+ v(t, x, w) = T D+ U (t, x, w) of the function v(t, x, w) is a negative denite function by virtue of the system (11), (12). Proof By virtue of the estimations (a) (d) of Assumption 5.2 the estimation D+ v(t, x, w) pT Cp holds, where p = (1 ( x ), 2 ( x ))T . A denite negativeness of the derivative follows from the condition of Lemma 5.2.

NONLINEAR DYNAMICS AND SYSTEMS THEORY, 5(2) (2005) 157167

167

Theorem 5.1 If the two-component system (11), (12) is such that all conditions of Lemmas 5.1 and 5.2 are fullled, then the state of equilibrium x = 0, w = 0 of the system is uniform asymptotically stable. If in Assumption 5.1 Nx = Rk , Nw = Q, functions i , i , i belong to the KR-class and conditions of Lemmas 5.1, 5.2 are fullled, then the state of equilibrium x = 0, w = 0 of the system (11), (12) is uniform asymptotically stable in the whole. Proof Under the enumerated conditions the function v(t, x, w) and its full derivative satisfy all conditions of Theorem 4.1. It proves the statement of Theorem 5.1. Remark 5.1 If in estimations (a) (d) of Assumption 5.2 we change the sign of the inequality for the opposite one and leave in the inequalities of Assumption 5.1 only estimation from below, then it isnt dicult to dene conditions of instability of the state x = 0, w = 0 of the system (11), (12). 6 Concluding Remarks Similar to Theorem 3.1 in the paper [7] the theorem was proved for discontinuous dynamical system. The mappings preserving stability in metric space were rst considered by Thomas [10] and Hahn [3]. In the papers [8] and the book [9] and other mappings of the type were studied in the stability analysis of large-scale systems. The application of multicomponent mapping U (t, p) : R+ X1 X2 adds more exibility to the approach to stability analysis of dynamical system in metric space, because this mapping admits a wider class of components for its elements vij (t, p),. References
[1] Bhatia, N.P. and Szeg, G.P. Stability Theory of Dynamical Systems. Springer-Verlag, o Berlin, 2002. [2] Birkho, G.G. Dynamical Systems. GITTL, MoscowLeningrad, 1941. [Russian]. [3] Hahn, W. Uber Stabilitatserhaltende Abbildungen and Ljapunovsche Funktionen. J. fur Reine und Andewandte Math. 228 (1967) 189192. [4] Martynyuk, A.A. The matrix-function of Liapunov and hybrid systems stability. Int. Appl. Mech. 21(4) (1985) 8996. [5] Martynyuk, A.A. Stability by Liapunovs Matrix Function Method with Applications. Marcel Dekker, Inc., New York, 1998. [6] Martynyuk, A.A. Matrix Liapunov functions and stability analysis of dynamical systems. In: Advances in Stability Theory at the End of XXth Century. (Ed.: A.A. Martynyuk). Taylor and Francis, London, 2003, P.135151. [7] Martynyuk, A.A. Stability of motion of discontinuous dynamical systems. Dokl. Akad. Nauk 397(3) (2004) 308312. [Russian]. [8] Michel A.N., Miller R.K. and Tang W. Lyapunov stability of interconnected systems: Decomposition into strongly connected subsystems. IEEE Trans. on Circuits and Systems CAS25 (1978) 799810. [9] Michel, A.N., Wang, K. and Hu, B. Qualitative Theory of Dynamical Systems. The Role of Stability Preserving Mappings. Marcel Dekker, Inc., New York, 2001. [10] Thomas J. Uber die Invarianz der Stabilitat bei einem Phasenraum-Homoomorphismus. J. fur Reine und Angewandte Math. 213 (1964) 147150. [11] Zubov, V.I. Methods of A. M. Liapunov and its Applications. Izdat. Leningrad. Gos. Universitet, Leningrad, 1957. [Russian]