Massachusetts Institute of Technology Department of Electrical Engineering and Computer Science 6.243j (Fall 2003): DYNAMICS OF NONLINEAR SYSTEMS by A. Megretski Lecture 7: Finding Lyapunov Functions 1 This lecture gives an introduction into basic methods for flnding Lyapunov functions and storage functions for given dynamical systems. 7.1 Convex search for storage functions The set of all real-valued functions of system state which do not increase along system trajectories is convex, i.e. closed under the operations of addition and multiplication by a positive constant. This serves as a basis for a general procedure of searching for Lyapunov functions or storage functions. 7.1.1 Linearly parameterized storage function candidates Consider a system model given by discrete time state space equations x(t + 1) = f (x(t), w(t)), y(t) = g(x(t), w(t)), (7.1) where x(t) ≤ X ‰ R n is the system state, w(t) ≤ W ‰ R m is system input, y(t) ≤ Y ‰ R k is system output, and f : X ￡ W ∈? X, g : X ￡ W ∈? Y are given functions. A functional V : X ∈? R is a storage function for system (7.1) with supply rate : Y ￡ W ∈? R if V (x(t + 1)) ? V (x(t)) ? (y(t)) (7.2) for every solution of (7.1), i.e. if x, ? x) ? (g(? w), ? x ≤ X, w ≤ W. (7.3)V (f (? w)) ? V (? x, ? w) ? ? ? 1 Version of September 26, 2003 ‰ 2 In particular, when · 0, this yields the deflnition of a Lyapunov function. Finding, for a given supply rate, a valid storage function (or at least proving that one exists) is a major challenge in constructive analysis of nonlinear systems. The most com- mon approach is based on considering a linearly parameterized subset of storage function candidates V deflned by N x) = ? q V q (?V = {V (? x), (7.4) q=1 where {V q } is a flxed set of basis functions, and ? k are parameters to be determined. Here every element of V is considered as a storage function candidate, and one wants to set up an e–cient search for the values of ? k which yield a function V satisfying (7.3). Example 7.1 Consider the flnite state automata deflned by equations (7.1) with value sets X = {1, 2, 3}, W = {0, 1}, Y = {0, 1}, and with dynamics deflned by f (1, 1) = 2, f (2, 1) = 3, f (3, 1) = 1, f (1, 0) = 1, f (2, 0) = 2, f (3, 0) = 2, g(1, 1) = 1, g(? w) = 0 ? (? w) ≡x, ? x, ? = (1, 1). In order to show that the amount of 1’s in the output is never much larger than one third of the amount of 1’s in the input, one can try to flnd a storage function V with supply rate (? w) = w ?3?y, ? ? y. Taking three basis functions V 1 , V 2 , V 3 deflned by 1, x = k, ? V k (?x) = 0, x = k, ? ≡ the conditions imposed on ? 1 , ? 2 , ? 3 can be written as the set of six a–ne inequalities (7.3), two of which (with (? w) = (1, 0) and (? w) = (2, 0)) will be satisfled automatically, while x, ? x, ? the other four are x, ?? 2 ?? 3 ? 1 at (? w) = (3, 0), x, ?? 2 ?? 1 ??2 at (? w) = (1, 1), x, ?? 3 ?? 2 ? 1 at (? w) = (2, 1), x, ?? 1 ?? 3 ? 1 at (? w) = (3, 1). Solutions of this linear program are given by ? 1 = c, ? 2 = c ?2, ? 3 = c ?1, 3 where c ≤ R is arbitrary. It is customary to normalize storage and Lyapunov functions so that their minimum equals zero, which yields c = 2 and ? 1 = 2, ? 2 = 0, ? 3 = 1. Now, summing the inequalities (7.2) from t = 0 to t = T yields T ?1 T ?1 3 y(t) ? V (x(0)) ?V (x(T )) + w(t), t=0 t=0 which is implies the desired relation between the numbers of 1’s in the input and in the output, since V (x(0)) ?V (x(T )) cannot be larger than 2. 7.1.2 Storage functions via cutting plane algorithms The possibility to reduce the search for a valid storage function to convex optimization, as demonstrated by the example above, is a general trend. One general situation in which an e–cient search for a storage function can be performed is when a cheap procedure of checking condition (7.3) (an oracle) is available. Assume that for every given element V ≤V it is possible to flnd out whether condition (7.3) is satisfled, and, in the case when the answer is negative, to produce a pair of vectors x ≤ X, w ≤ W for which the inequality in (7.3) does not hold. Select a su–ciently large ? ? set T 0 (a polytope or an ellipsoid) in the space of parameter vector ? = (? q ) q N =1 (this set will limit the search for a valid storage function). Let ? ? be the “center” of T 0 . Deflne V by the ? ? , and apply the veriflcation “oracle” to it. If V is a valid storage function, the search for storage function ends successfully. Otherwise, the “invalidity certiflcate” x, ?(? w) produced by the oracle yields a hyperplane separating ? ? and the (unknown) set of ? deflning valid storage functions, thus cutting a substantial portion from the search set T 0 , reducing it to a smaller set T 1 . Now re-deflne ? ? as the center of T 1 and repeat the process by constructing a sequence of monotonically decreasing search sets T k , until either a valid storage function is found, or T k shrinks to nothing. With an appropriate selection of a class of search sets T k (ellipsoids or polytopes are most frequently used) and with an adequate deflnition of a “center” (the so-called “analytical center” is used for polytopes), the volume of T k can be made exponentially decreasing, which constitutes fast convergence of the search algorithm. 7.1.3 Completion of squares The success of the search procedure described in the previous section depends heavily on the choice of the basis functions V k . A major di–culty to overcome is veriflcation of (7.3) for a given V . It turns out that the only known large linear space of functionals ? ? 4 F : R n ∈? R which admits e–cient check of non-negativity of its elements is the set of quadratic forms ? ? ? ? ? x x ? ? ? F (?x) = Q , (Q = Q ) 1 1 for which nonnegativity is equivalent to positive semideflniteness of the coe–cient matrix Q. This observation is exploited in the linear-quadratic case, when f, g are linear functions f(? w) = A? w, g(? w) = C ? w, x, ? x + B ? x, ? x + D ? and is a quadratic form ? ? ? ? ? x x? ? (? w) = ? § .x, ? ?w w Then it is natural to consider quadratic storage function candidates V (? x xx) = ? ? P ? only, and (7.3) transforms into the (symmetric) matrix inequality P A + A ? P P B ? §. (7.5) B ? P 0 Since this inequality is linear with respect to its parameters P and §, it can be solved relatively e–ciently even when additional linear constraints are imposed on P and §. Note that a quadratic functional is non-negative if and only if it can be represented as a sum of squares of linear functionals. The idea of checking non-negativity of a functional by trying to represent it as a sum of squares of functions from a given linear set can be used in searching for storage functions of general nonlinear systems as well. Indeed, let ? H : R n ￡ R m ∈? R M and V ? : R n ∈? R N be arbitrary vector-valued functions. For every ? ≤ R N , condition (7.3) with x) = ? ? V ? (?V (? x) is implied by the identity x, ? ? ? x) + ? x, ? H(? w) = (? w) ? ? ?? ? V ? (f(? w)) ? ? V (? H(? w) ? S ? x, ? x, ? x ≤ X, w ≤ W, (7.6) as long as S = S ? ? 0 is a positive semideflnite symmetric matrix. Note that both the storage function candidate parameter ? and the “sum of squares” parameter S = S ? ? 0 enter constraint (7.6) linearly. This, the search for a valid storage function is reduced to semideflnite program. In practice, the scalar components of vector ? H should include enough elements so that identity (7.6) can be achieved for every ? ≤ R N by choosing an appropriate S = S ? (not necessarily positivie semideflnite). For example, if f, g, are polynomials, it may be a good idea to use a polynomial V ? and to deflne ? H as the vector of monomials up to a given degree. 5 7.2 Storage functions with quadratic supply rates As described in the previous section, one can search for storage functions by considering linearly parameterized sets of storage function candidates. It turns out that storage functions derived for subsystems of a given system can serve as convenient building blocks (i.e. the components V q of V ? ). Indeed, assume that V q = V q (x(t)) are storage functions with supply rates q = q (z(t)). Typically, z(t) includes x(t) as its component, and has some additional elements, such as inputs, outputs, and othe nonlinear combinations of system states and inputs. If the objective is to flnd a storage function V ? with a given supply rate ? , one can search for V ? in the form N V (x(t)) = V q (x(t)), ? q ? 0, (7.7) q=1 where ? q are the search parameters. Note that in this case it is known a-priori that every V ? in (7.7) is a storage function with supply rate N (z(t)) = ? q q (z(t)). (7.8) q=1 Therefore, in order to flnd a storage function with supply rate ? = ? (z(t)), it is su–cient to flnd ? q ? 0 such that N ? 1 q (? z) ? ?z) ? ? (? z. (7.9) q=1 When ? , q are generic functions, even this simplifled task can be di–cult. However, in the important special case when ? and q are quadratic functionals, the search for ? q in (7.9) becomes a semideflnite program. In this section, the use of storage functions with quadratic supply rates is discussed. 7.2.1 Storage functions for LTI systems x) = ? x is a storage function for LTI system A quadratic form V (? x ? P ? x˙ = Ax + Bw (7.10) with quadratic supply rate ? ? ? ? ? x x? ? (? w) = ? §x, ? ?w w if and only if matrix inequality (7.5) is satisfled. The well-known Kalman-Popov-Yakubovich Lemma, or positive real lemma gives useful frequency domain condition for existence of such P = P ? for given A, B, §. 6 Theorem 7.1 Assume that the pair (A, B) is controllable. A symmetric matrix P = P ? satisfying (7.5) exists if and only if ? ? ? ? ? ?x ?w § ?x ?w ? 0 whenever jσ?x = A?x + B ?w for some σ ≤ R. (7.11) Moreover, if there exists a matrix K such that A + BK is a Hurwitz matrix, and ? ? ? ? ? I I § ? 0, K K then all such matrices P = P ? are positive semideflnite. Example 7.2 Let G(s) = C(sI ?A) ?1 B + D be a stable transfer function (i.e. matrix A is a Huewitz matrix) with a controllable pair (A, B). Then |G(jσ)| ? 1 for all σ ≤ R if and only if there exists P = P ? ? 0 such that w| 2 w| 2 x x + B ? x + D ? ? ? ?2? ? P(A? w) ?|? ?|C ? x ≤ R n , w ≤ R m . This can be proven by applying Theorem 7.1 with (? w) = |? x + D ?x, ? w| 2 ?|C? w| 2 and K = 0. 7.2.2 Storage functions for sector nonlinearities Whenever two components v = v(t) and w = w(t) of the system trajectory z = z(t) are related in such a way that the pair (v(t), w(t)) lies in the cone between the two lines w = k 1 v and v = k 2 v, V · 0 is a storage function for (z(t)) = (w(t) ?k 1 v(t))(k 2 v(t) ?w(t)). For example, if w(t) = v(t) 3 then (z(t)) = v(t)w(t). If w(t) = sin(t) sin(v(t)) then 2 (z(t)) = |v(t)| 2 ?|w(t)| . 7.2.3 Storage for scalar memoryless nonlinearity Whenever two components v = v(t) and w = w(t) of the system trajectory z = z(t) are related by w(t) = `(v(t)), where ` : R ∈? R is an integrable function, and v(t) is a component of system state, V (x(t)) = ?(v(t)) is a storage function with supply rate (z(t)) = v˙(t)w(t), where ? y ?(y) = `(?)d?. 0 7 7.3 Implicit storage functions A number of important results in nonlinear system analysis rely on storage functions for which no explicit formula is known. It is frequently su–cient to provide a lower bound for the storage function (for example, to know that it takes only non-negative values), and to have an analytical expression for the supply rate function . In order to work with such “implicit” storage functions, it is helpful to have theorems which guarantee existence of non-negative storage functions for a given supply rate. In this regard, Theorem 7.1 can be considered as an example of such result, stating existence of a storage function for a linear and time invariant system as an implication of a frequency- dependent matrix inequality. In this section we present a number of such statements which can be applied to nonlinear systems. 7.3.1 Implicit storage functions for abstract systems Consider a system deflned by behavioral set B = {z} of functions z : [0,?) ∈? R q . As usually, the system can be autonomous, in which case z(t) is the output at time t, or with an input, in which case z(t) = [v(t); w(t)] combines vector input v(t) and vector output w(t). Theorem 7.2 Let : R q ∈? R be a function and let B be a behavioral set, consisting of some functions z : [0,?) ∈? R q . Assume that the composition (z(t)) is integrable over every bounded interval (t 0 ,t 1 ) in R + for all z ≤ B. For t 0 ,t ≤ R + deflne ? t I(z,t 0 ,t) = (z(?))d?. t 0 The following conditions are equivalent: (a) for every z 0 ≤ B and t 0 ≤ R + the set of values I(z,t 0 ,t), taken for all t ? t 0 and for all z ≤ B deflning same state as z 0 at time t 0 , is bounded from below; (b) there exists a non-negative storage function V : B￡R + ∈? R + (such that V(z 1 ,t) = V(z 2 ,t) whenever z 1 and z 2 deflne same state of B at time t) with supply rate . Moreover, when condition (a) is satisfled, a storage function V from (b) can be deflned by V(z 0 (￠),t 0 ) = ?inf I(z,t 0 ,t), (7.12) where the inflmum is taken over all t ? t 0 and over all z ≤ B deflning same state as z 0 at time t 0 . Proof Implication (b)≥(a) follows directly from the deflnition of a storage function, which requires V(z 0 ,t 1 ) ? V(z 0 ,t 0 ) ? I(z,t 0 ,t 1 ) (7.13) ‰ ? 8 for t 1 ? t 0 , z 0 ≤ B. Combining this with V ? 0 yields I(z,t 0 ,t 1 ) ? ?V(z,t 0 ) = ?V(z 0 ,t 0 ) for all z,z 0 deflning same state of B at time t 0 . Now let us assume that (a) is valid. Then a flnite inflmum in (7.12) exists (as an inflmum over a non-empty set bounded from below) and is not positive (since I(z 0 ,t 0 ,t 0 ) = 0). Hence V is correctly deflned and not negative. To flnish the proof, let us show that (7.13) holds. Indeed, if z 1 deflnes same state as z 0 at time t 1 then z 0 (t), t ? t 1 , z 01 (t) = z 1 (t), t > t 1 deflnes same state as z 0 at time t 0 < t 1 (explain why). Hence the inflmum of I(z,t 0 ,t) in the deflnition of V is not larger than the inflmum of integrals of all such z 01 , over intervals of length not smaller than t 1 ? t 0 . These integrals can in turn be decomposed into two integrals I(z 01 ,t 0 ,t) = I(z 0 ,t 0 ,t 1 ) + I(z 1 ,t 1 ,t), which yields the desired inequality. 7.3.2 Storage functions for ODE models As an important special case of Theorem 7.2, consider the ODE model x˙(t) = f(x(t),w(t)), (7.14) deflned by a function f : X ￡ W ∈? R n , where X,W are subsets of R n and R m respectively. Cxonsider the behavior model B consisting of all functions z(t) = [x(t); v(t)] where x : [0,?) ∈? X is a solution of (7.14). In this case, two signals z 1 = [x 1 ; v 1 ] and z 2 = [x 2 ; v 2 ] deflne same state of B at time t 0 if and only if x 1 (t 0 ) = x 2 (t 0 ). Therefore, according to Theorem 7.2, for a given function : X ￡ W ∈? R, existence of a function V : X ￡ R + ∈? R + such that t 2 V(x(t 2 ),t 2 ) ? V(x(t 1 ),t 1 ) ? (x(t),v(t))dt t 1 for all 0 ? t 1 ? t 2 , [x; v] ≤ B is equivalent to flniteness of the inflmum of the integrals ? t (x(?),v(?))d? t 0 over all solutions of (7.14) with a flxed x(t 0 ) = ?x 0 which can be extended to the time interval [0,?). ? ? ? ? 9 In the case when X = R n , and f : R n ∈? R n is such that existence and uniqueness of solutions x : [0, ?) ∈? R n is guaranteed for all locally integrable inputs w : [0, ?) ∈? W and all initial conditions x(t 0 ) = ?x 0 ≤ R n , the inflmum in (7.12) (and hence, the corresponding storage function) do not depend on time. If, in addition, f is continuous and V is continuously difierentiable, the well-known dynamic programming condition x, ? x)f(? w)}0 ? inf { (? w)??V (? x 0 , ? x 0 ≤ R n lim inf { (? w)??V (? x, ? x 0 , ? x 0 )f(? w)} ? ? ??0,?>0 w→W,? x 0 ) w→W? x→B ? (? ? (7.15) will be satisfled. However, using (7.15) requires a lot of caution in most cases, since, even for very smooth f, , the resulting storage function V does not have to be difierentiable. 7.3.3 Zames-Falb quadratic supply rate A non-trivial and powerful case of an implicitly deflned storage function with a quadratic supply rate was introduced in late 60-s by G. Zames and P. Falb. Theorem 7.3 Let A, B, C be matrices such that A is a Hurwitz matrix, and |Ce At B|dt < 1. 0 Let ` : R ∈? R be a monotonic odd function such that ? w) ? | ? ?0 ? w`( ? w| 2 ? w ≤ R. Then for all < 1 system x˙(t) = Ax(t) + Bw(t) has a non-negative storage function with supply rate x, ? w ? `( ? w ? C ? + (? w) = ( ? w))( ? x), and system x˙(t) = Ax(t) + B(w(t) ? `(w(t)) has a non-negative storage function with supply rate x, ? w ? `( ? x) ? ? (? w) = ( ? w) ? C ? w. The proof of Theorem 7.3 begins with establishing that, for every function h : R ∈? R with L1 norm not exceeding 1, and for every square integrable function w : R ∈? R the integral (w(t) ? `(w(t)))y(t)dt, ?? where y = h ? w, is non-negative. This verifles that the assumptions of Theorem 7.2 are satisfled, and proves existence of the corresponding storage function without actually flnding it. Combining the Zames-Falb supply rate with the statement of the Kalman- Yakubovich-Popov lemma yields the following stability criterion. ? ? ? ? 10 Theorem 7.4 Assume that matrices A p , B p , C p are such that A p is a Hurwitz matrix, and there exists ? > 0 such that Re(1 ? G(jσ))(1 ? H(jσ)) ? ? ? σ ≤ R, where H is a Fourier transform of a function with L1 norm not exceeding 1, and G(s) = C p (sI ? A p ) ?1 B p . Then system x˙(t) = A p x(t) + B p `(Cx(t) + v(t)) has flnite L2 gain, in the sense that there exists > 0 such that |x(t)| 2 dt ? (|x(0)| 2 + |v(t)| 2 dt 0 0 for all solutions. 7.4 Example with cubic nonlinearity and delay Consider the following system of difierential equations 2 with an uncertain constant delay parameter ?: x˙ 1 (t) = ?x 1 (t) 3 ? x 2 (t ? ?) 3 (7.16) x˙ 2 (t) = x 1 (t) ? x 2 (t) (7.17) Analysis of this system is easy when ? = 0, and becomes more di–cult when ? is an arbitrary constant in the interval [0, ? 0 ]. The system is not exponentially stable for any value of ?. Our objective is to show that, despite the absence of exponential stability, the method of storage functions with quadratic supply rates works. The case ? = 0 For ? = 0, we begin with describing (7.16),(7.17) by the behavior set Z = {z = [x 1 ; x 2 ; w 1 ; w 2 ]}, where 3 3 w 1 = x 1 , w 2 = x 2 , x˙ 1 = ?w 1 ? w 2 , x˙ 2 = x 1 ? x 2 . Quadratic supply rates for which follow from the linear equations of Z are given by ? ? ? ? ? x 1 ?w 1 ? w 2 LT I (z) = 2 P x 1 ? x 2 , x 2 2 Suggested by Petar Kokotovich 11 where P = P ? is an arbitrary symmetric 2-by-2 matrix deflning storage function V LTI (z(￠),t) = x(t) ? Px(t). Among the non-trivial quadratic supply rates valid for Z, the simplest are deflned by NL (z) = d 1 x 1 w 1 + d 2 x 2 w 2 + q 1 w 1 (?w 1 ? w 2 ) + q 2 w 2 (x 1 ? x 2 ), with the storage function V NL (z(￠),t) = 0.25(q 1 x 1 (t) 4 + q 2 x 2 (t) 4 ), where d k ? 0. It turns out (and is easy to verify) that the only convex combinations of these supply rates which yield ? 0 are the ones that make = LTI + NL = 0, for example ? ? 0.5 0 P = , d 1 = d 2 = q 2 = 1, q 1 = 0. 0 0 The absence of strictly negative deflnite supply rates corresponds to the fact that the system is not exponentially stable. Nevertheless, a Lyapunov function candidate can be constructed from the given solution: 4 4 2 4 V(x) = x ? Px+ 0.25(q 1 x 1 + q 2 x 2 ) = 0.5x 1 + 0.25x 2 . This Lyapunov function can be used along the standard lines to prove global asymptotic stability of the equilibrium x = 0 in system (7.16),(7.17). 7.4.1 The general case Now consider the case when ? ≤ [0,0.2] is an uncertain parameter. To show that the delayed system (7.16),(7.17) remains stable when ? ? 0.2, (7.16),(7.17) can be represented by a more elaborate behavior set Z = {z(￠)} with z = [x 1 ; x 2 ; w 1 ; w 2 ; w 3 ; w 4 ; w 5 ; w 6 ] ≤ R 8 , satisfying LTI relations x˙ 1 = ?w 1 ? w 2 + w 3 , x˙ 2 = x 1 ? x 2 and the nonlinear/inflnite dimensional relations 3 3 3 w 1 (t) = x 1 , w 2 = x 2 , w 3 = x 2 ? (x 2 + w 4 ) 3 , 3 w 4 (t) = x 2 (t? ?) ? x 2 (t), w 5 = w 4 , w 6 = (x 1 ? x 2 ) 3 . Some additional supply rates/storage functions are needed to bound the new variables. These will be selected using the perspective of a small gain argument. Note that the ? ? fl fl fl fl ? ? 12 perturbation w 4 can easily be bounded in terms of ˙x 2 = x 1 ? x 2 . In fact, the LTI system with transfer function (exp(??s) ? 1)/s has a small gain (in almost any sense) when ? is small. Hence a small gain argument would be applicable provided that the gain “from w 4 to ˙x 2 ” could be bounded as well. It turns out that the L 2 -induced gain from w 4 to x˙ 2 is unbounded. Instead, we can use the L 4 norms. Indeed, the last two components w 5 ,w 6 of w were introduced in order to handle L 4 norms within the framework of quadratic supply rates. More speciflcally, in addition to the usual supply rate ? ? ? ? ? x 1 ?w 1 ? w 2 + w 3 LT I (z) = 2 P x 1 ? x 2 , x 2 the set Z has supply rates (z) =d 1 x 1 w 1 + d 2 x 2 w 2 + q 1 w 1 (?w 1 ? w 2 + w 3 ) + q 2 w 2 (x 1 ? x 2 ) + d 3 [0.99(x 1 w 1 + x 2 w 2 ) ? x 1 w 3 + 2.5 4 w 4 w 5 ? 0.5 4 (x 1 ? x 2 )w 6 ] + q 3 [0.2 4 (x 1 ? x 2 )w 6 ? w 4 w 5 ], d i ? 0. Here the supply rates with coe–cients d 1 ,d 2 ,q 1 ,q 2 are same as before. The term with d 3 , based on a zero storage function, follows from the inequality ? 4 ? 4 x 1 ? x 2 4 4 3 0.99(x 1 + x 2 ) ? x 1 (x 2 ? (x 2 + w 4 ) 3 ) + 5w 4 ? ? 0 2 2 (which is satisfled for all real numbers x 1 ,x 2 ,w 4 , and can be checked numerically). The term with q 3 follows from a gain bound on the transfer function G ? (s) = (exp(??s)? 1)/s from x 1 ? x 2 to w 4 . It is easy to verify that the L 1 norm of its impulse response equals ?, and hence the L 4 induced gain of the causal LTI system with transfer function G ? will not exceed 1. Consider the function fl fl 4 ? ? ? t V d (v(￠),T) = ? inf 0.2 4 |v 1 (t)| 4 ? fl v 1 (r)dr fl dt, (7.18) T t?? where the inflmum is taken over all functions v 1 which are square integrable on (0,?) and such that v 1 (t) = v(t) for t ? T. Because of the L 4 gain bound of G ? with ? ≤ [0,0.2] does not exceed 0.2, the inflmum in (7.18) is bounded. Since we can always use v 1 (t) = 0 for t > T, the inflmum is non-positive, and hence V d is non-negative. The IQC deflned by the “q 3 ” term holds with V = q 3 V d (x 1 ? x 2 ,t). Let 4 4 0 (z) = ?0.01(x 1 w 1 + x 2 w 2 ) = ?0.01(x 1 + x 2 ), which re ects our intention to show that x 1 ,x 2 will be integrable with fourth power over (0,?). Using 0.5 0 P = , d 1 = d 2 = 0.01, d 3 = q 2 = 1, q 1 = 0, q 3 = 2.5 4 0 0 13 yields a Lyapunov function V(x e (t)) = 0.5x 1 (t) 2 + 0.25x 2 (t) 4 + 2.5 4 V d (x 1 ?x 2 ,t), where x e is the “total state” of the system (in this case, x e (T) = [x(T); v T (￠)], where v T (￠) ≤ L 2 (0,?) denotes the signal v(t) = x 1 (T ?? + t) ?x 2 (T ?? + t) restricted to the interval t ≤ (0,?)). It follows that dV(x e (t)) ??0.01(x 1 (t) 4 + x 2 (t) 4 ). dt On the other hand, we saw previously that V(x e (t)) ? 0 is bounded from below. There- fore, x 1 (￠),x 2 (￠) ≤ ? 4 (fourth powers of x 1 ,x 2 are integrable over (0,?)) as long as the initial conditions are bounded. Thus, the equilibrium x = 0 in system (7.16),(7.17) is stable for 0 ? ? ? 0.2.