• <tr id="yyy80"></tr>
  • <sup id="yyy80"></sup>
  • <tfoot id="yyy80"><noscript id="yyy80"></noscript></tfoot>
  • 99热精品在线国产_美女午夜性视频免费_国产精品国产高清国产av_av欧美777_自拍偷自拍亚洲精品老妇_亚洲熟女精品中文字幕_www日本黄色视频网_国产精品野战在线观看 ?

    Algorithms to Compute the Largest Invariant Set Contained in an Algebraic Set for Continuous-Time and Discrete-Time Nonlinear Systems

    2020-02-29 14:14:16LauraMeniniCorradoPossieriandAntonioTornamb
    IEEE/CAA Journal of Automatica Sinica 2020年1期

    Laura Menini, Corrado Possieri, and Antonio Tornambè

    Abstract—In this paper, some computational tools are proposed to determine the largest invariant set, with respect to either a continuous-time or a discrete-time system, that is contained in an algebraic set. In particular, it is shown that if the vector field governing the dynamics of the system is polynomial and the considered analytic set is a variety, then algorithms from algebraic geometry can be used to solve the considered problem. Examples of applications of the method (spanning from the characterization of the stability to the computation of the zero dynamics) are given all throughout the paper.

    I. INTRODUCTION

    THE “second method” of Lyapunov, which makes use of a function that is monotonically decreasing along the solution of a system, is one of the most widespread and literature pervasive tools to characterize the stability of nonlinear system [1]-[9], such a method has been generalized in order to possibly deal with monotonically non-increasing functions, by exploiting the concept of invariant set. Since these seminal works, the notion of invariant set has been extensively used in order to characterize the properties of nonlinear dynamical systems. For example, in [10], invariant sets are used to characterize the disturbances that let the output of a discrete-time,linear system vanish identically, in [11], controlled invariant sets are used to solve the disturbance decoupling problem, in[12], invariant sets are used to characterize some observability properties, and, in [13], (controlled) invariant sets are used to extend the notion of zero to nonlinear dynamical systems(see also [14] for a comprehensive review of the literature about the output zeroing problem and the zero dynamics of a system).

    Due to the importance of invariant sets in both theoretical research and engineering practice (especially when dealing with stability of nonlinear dynamical system), a lot of research effort has been carried out to characterize such sets [4], [15],[16]-[20].

    The main objective of this paper is to provide computational tools to determine the largest invariant set, with respect to either a continuous-time or a discrete-time nonlinear system,that is contained in an algebraic set. Such a problem is crucial in several control applications, as shown by the subsequent Theorems 1 and 6, that require the solution to such a problem to guarantee asymptotic stability of a given set. The goal of computing such an invariant set is pursued by showing that this set can be determined by solving a set of analytic equations. Furthermore, it is shown that, if the vector field governing the dynamics of the system is polynomial and the algebraic set is an affine variety, the largest invariant set with respect to the system that is contained in the affine variety can be determined through computationally efficient algorithms,which use tools borrowed from algebraic geometry that have already been used to solve control problems, such as observer design [21], [22], generation of algebraic certificates of(possibly, asymptotic) stability [23], solving the equations that arise in the harmonic balancing method [24], multi-objective optimal design of controllers for linear plants [25], checking the controllability of polynomial systems [26], and motion planning of mobile robots [27], [28]. Examples are given all throughout the paper to illustrate and corroborate the theoretical results. Applications of the proposed methods to several control problems are reported as, e.g., stability analysis of nonlinear systems, test of the zero-state observability property, solution to the output zeroing problem,computation of the zero dynamics of a nonlinear plant.

    The results given in this paper are related to the ones concerning the output zeroing of nonlinear autonomous systems [29]-[31]. The main difference between the results given in this paper and the ones available in the literature is that herein algorithmic procedures (ready to be implemented in CAS software, such as Macaulay2 [32]) are proposed to determine the largest invariant set that is contained in an algebraic set with respect to either a continuous-time or a discrete-time analytic (possibly, polynomial) system.

    Organization of the paper: the notation used in this paper is introduced in Section II, together with some preliminary results. The problem of determining the largest invariant set,which is contained in an analytic set, with respect to either a continuous-time or a discrete-time system is addressed in Sections III and IV, respectively. Examples of application of the given methods are reported in Section V. Conclusions are given in Section VI.

    II. NOTATION AND PRELIMINARIES

    Given a multi-index

    is the variety generated byp1,...,p?. Similarly, given an idealinthe variety of the ideal I is the set

    A monomial order ? is a total well ordering relation among the monomialsOnce a monomial order ? is fixed(such as the LEX order defined in [33]), any polynomialcan be rewritten aswherei=1,...,m, andThus,the leading term and the leading coefficient ofpareand LC(p)=c1, respectively. Once a monomial order ? is fixed, the setwith 〈g1,...,gs〉=I,is a Gr?bner basis of an ideal I in R [x] if

    A functionis smooth in its domain of definition(assumed to be open) ifqis C∞at eachIf, a power series can be associated with the smooth functionqlocally aboutx0as follows:

    A smooth functionqis locally analytic inx0if there exists an open neighborhoodofx0such that the series in (2)converges andq(x)=Q(x), for eachx∈B;qis analytic inif it is locally analytic in each

    Givenand smooth functionsanddefine theith directional derivative ofhalongfasandand theith directional successor ofhalongfasand

    Given analytic functionsdefine the analytic set

    wherex=[x1...xn]T, and the vector fieldis analytic inwithbeing an open subset of. Assume,without loss of generality, that the initial time is 0. The response at timetof system (3) from the initial stateis denotedis theCT-flowassociated withfand satisfies (see [34])

    Definition 1:A setis positively (respectively,negatively)f-invariant if, for eachx0∈S , φf(t,x0)∈S,(respectively,A set S isf-invariant if it is both positively and negativelyf-invariant.

    Hence, consider the following problem.

    Problem 1:Let functionsanalytic inbe given. Find the largestf-invariant set contained in

    The availability of tools to determine a solution to Problem 1 is crucial in several control applications and, in particular,for the application of the next two theorems.

    Theorem 1 (Lyapunov stability [4]):Let 0 be an equilibrium of system (3) and assume thatLetV:be a differentiable, radially unbounded, positive definite function such thatLf V(x)≤0, ?x∈Rn. If the largestf-invariant set inis {0} , then 0 is globally asymptotically stable.

    Theorem 2 (Invariance principle [35]):Letbe a compact, positivelyf-invariant set. Letbe analytic in Ω and such thatLf V(x)≤0, ?x∈Ω. Let S ?Ω be the largestf-invariant set in. Every responsex(t) of system (3) starting in Ω approaches S , i.e.,

    III. INVARIANT SETS FOR CONTINUOUS-TIME SYSTEMS Consider the continuous-time, time-invariant system

    In order to apply Theorem 1, one has to guarantee that S={0} is the solution to Problem 1 withh=Lf V(x). On the other hand, in order to apply Theorem 2 to prove attractiveness of the set S, one has to guarantee that S is the largest invariant set that is contained in, i.e., that S is the solution to Problem 1 withh=Lf V(x), and that the set Ωis positively invariant. It is worth noticing that if, in addition to the hypotheses of Theorem 2, the functionVis positive definite and there isc>0, such thatis a subset ofand is compact, then, by classical Lyapunov results [4], the setis positively invariant (and contractive).

    Remark 1:The LaSalle invariance principle recalled in Theorem 2 has already been used in [36], [37] to determine estimates of the region of attraction of an equilibrium point via polynomial optimization techniques, such as sum of squares tools. The main difference between such techniques and the ones given in this paper is that the former allow one to determine positivelyf-invariant sets that constitute inner estimates of the region of attraction of an asymptotically stable equilibrium point, whereas the latter allow to determine algebraically the largestf-invariant set S contained in a variety (such as, e.g.,i.e., under the hypotheses of Theorem 2, they allow one to characterize the set to which the trajectories of system (3) converge rather than its region of attraction.

    In the following Section III-A, a method to determine the solution to Problem 1 in the analytic case is proposed. An algorithm, which uses the algebraic geometry tools recalled in Section II, is given in Section III-B to apply such a technique in the polynomial case.

    A. Solution to Problem 1 in the Analytic Case

    The main objective of this subsection is to provide a solution to Problem 1 whenfandh1,...,hmare analytic functions, but not necessarily polynomial. To this end,consider the following remark.

    Remark 2:Letbe analytic in some intervalpossibly coincident with R. Letbe any open subinterval ofOne has that α(t) is equal tocfor allfor some constantif and only if α(t) is equal tocfor allTherefore, for anyt0in the interior of, α(t) is equal tocfor allif and only if α(t0)=candk>0.

    Following Remark 2, the next lemma provides necessary and sufficient conditions ensuring that a functionhvanishes identically along the CT-flow φf(t,x0).

    Lemma 1:Letf:andbe analytic inFor eachone hash(φf(t,x0))=0 ,if and only if?i∈Z≥0.

    Proof:Sincefandhare analytic infor any, one has thath(φf(t,x0)) is an analytic function oftinHence,sincefor alltif and only ifh(φf(t,x0))=0 for allwhereis a sufficiently small neighborhood of 0. Since

    one has thath(φf(t,x0)) vanishes identically if and only if

    The results given in Lemma 1 are used in the following lemma to prove that the largest positively (respectively,negatively)f-invariant subset ofis the largestfinvariant subset of

    Lemma 2:Ifis the largest positively (respectively,negatively)f-invariant set contained in, then it is the largestf-invariant set contained in

    Proof:Letx0∈S be fixed. Thus, letbe the maximal interval of existence of φf(t,x0) . Since S is the largest positivelyf-invariant set contained inthere existssuch thatqj(φf(t,x0))=0 for allt∈[0,T],j=1,2,.... By Lemma 1, if there exists a responsex(t) of system (3) such thatqj(x(t))=0, ?t∈[0,T], for somethenTherefore, sincex0can be chosen arbitrarily in S and there does not existsuch that, for alland(otherwisewould be in S) , S is the largestf-invariant set contained in

    By using Lemma 2, the following theorem shows how to determine the solution to Problem 1 in the analytic case.

    Theorem 3:Letandbe analytic functions in U. The largestf-invariant set contained inis

    The following example shows how Theorem 3 can be used to determine the solution to Problem 1.

    Example 1:Consider the following system

    whereqi(x) is a globally analytic function for each(namely,q0(x)=1,q1(x)=cos(x1),q2=sin2(x2)+cos(x1)×cos(x2), and so on). Therefore, sincethe largest invariant w.r.t. system (4) contained inis

    In order to apply Theorem 3, one has to compute the intersection of infinitely many sets, i.e.

    Proof:By Lemma 2, the largestf-invariant set contained inis

    By Lemma 1,hj(φf(t,x))=0 for allif and only if

    and leth(x)=sin(x1). It can be easily derived that

    Hence, consider the following proposition, which show how these computations can be simplified.

    Proposition 1:Assume that there exist a set {q1,...,q?} of analytic functions insuch that

    Fig. 1. Phase plot of system (4) (blue), algebraic set (green), and invariant set S (red).

    for some αk,χ,k=1,...,? , being analytic in. Hence, there exist functionsbeing analytic insuch that=1,...,? ,i∈Z≥0. Furthermore, ifthenqi(φf(t,x0))=0,i=1,...,?, for allt∈Tx0.

    Proof:The proof is carried out by induction. By assumption, there exist analytic functions αj,ksuch that

    Thus, the statement holds fori=1. Assume now that, fori∈Z≥0, there exist analytic functions βi,j,ksuch that. Therefore,

    which is a linear combination ofq1,...,q?, with analytic coefficients in U. Thus, ifqj(x0)=0 for allj=1,...,?, thenfor allHence, the proof follows by Lemma 1. ■

    By Proposition 1, if there existssuch that each

    Therefore, if such an assumption holds, then the computations to be carried out so to find a solution to Problem 1 involve just a finite number of intersections, as shown in the next example.

    Example 2:Consider the following system (p. 214 of [38]):

    Therefore, since there exists an analytic function α such thatLf h(x)=α(x)h(x), by Proposition 1, the largestf-invariant set contained inFig. 2 depicts the phase plot of system (7) and

    Fig. 2. Phase plot of system (7) (blue curves), algebraic set (green),and f-invariant set S (red).

    Example 3:Consider the system given in Ex. 8, Ch. 9 of[38]

    it is immediate to see that it cannot be expressed as α(x)h(x),for some α(x) analytic on the whole R2. Similarly, it is not trivial to determine whether

    can be expressed as a linear combination ofhandLf h.However, algebraic geometry can be used to solve this problem, as suggested in Chapter 6 of [33]. Introduce the auxiliary single variablesc1,c2,s1,s2, and fix inthe LEX order ?, withConsider the polynomial -c1c2obtained by replacing cos(x1) and cos(x2) inLf h(x) withc1andc2, respectively; consider the ideal J1generated byx2,c1c2and by thei=1,2, where the polynomialarises from the equality sin2(xi)+cos2(xi)=1, with similar substitutions. In the same way, consider the polynomialwhich arises fromBy computingp2%J1=c2s2≠0, one obtains thatp2cannot be expressed as a polynomial sum of the generators ofwhencecannot be expressed as a linear combination ofhandLf h. Thus, computeand its associated polynomialHence, by letting

    one obtains thatwhencecan be expressed as a linear combination ofThus, by Proposition 1,the solution to Problem 1 is given by

    Fig. 3 shows the phase plot of system (8),and S.

    Fig. 3. Phase plot of system (8) (blue), algebraic set (green), and invariant set S (red).

    Note that, although for many analytic systems there isN∈Z≥0such thatcan be expressed as a linear combination of...,m,i=0,...,N, this need not always hold, as shown in the following counterexample,which shows that the solution to Problem 1 need not be computable as a finite intersection, as in (6).

    Example 4:Consider the system

    By [41], the functionh:R →R is analytic in R,andfori=1,2,..., andj=1,2,3,.... Since it results thatandh(N+1)=0,...,=0,there is nosuch thatcan be expressed as a linear combination of. Namely, for each

    whereas the largestf-invariant set contained inis

    i.e., the set S cannot be determined with a finite number of intersections.

    In the following subsection, it is shown that, if bothfandhare polynomial functions, then there always existsN∈Z≥0such thatcan be expressed as a weighted sum ofwith polynomial weights, thus allowing the design of an algorithm to solve Problem 1.

    B. Solution to Problem 1 in the Polynomial Case

    The main objective of this subsection is to show that the computations required to determine the solution to Problem 1 are much simpler when the entries offandh1,...,hmare inwhich implies that

    Consider the following lemma.

    Lemma 3:Let a countable (possibly, infinite) sequence of polynomialsp1,p2,... inbe given. There exists a finite number of polynomialssuch that:

    Proof:LetandSinceis an ascending chain of ideals inby Theorem 7 at p. 80 of [33], there isN≥1 such thatBy Theorem 4 at p. 190 of[33],

    Lemma 3 suggests that, in order to determine a solution to Problem 1 by using the expression given in (5), if both the entries offandh1,...,hmare polynomials, then S can be obtained by computing the intersection of a finite number of varieties. In the remainder of this subsection, it is shown how such computations can be carried out by using the algebraic geometry tools recalled in Section II.

    Let a monomial order ? be fixed. Given an ideal I in R[x],letbe the reduced Gr?bner basis of I with respect to ?. The directional derivativeLfI of I alongfis the ideal defined as follows:

    Given a polynomialpin some ideal I, the next lemma states that the directional derivative alongfofpbelongs to the sum of I and of its directional derivativeLfI.

    Lemma 4:Iff∈Rn[x] and I is an ideal in R[x], thenp∈I implies that

    Proof:Ifp∈I, then, letting 〈p1,...,p?〉=:I, there exist α1,...,α?∈R[x] such thatThus, one has

    The following statement extends the results of Lemma 4 to theith directional derivative of a polynomial.

    Lemma 5:Iff∈Rn[x] and I is an ideal in R[x], define the sequence of ideals I0,I1,I2,... iteratively as

    p∈I

    Hence, if,then

    Proof:The lemma is proved by induction. The base of the induction is proved by Lemma 4 by lettingi=1. As for the inductive step, assume that for somei∈Z≥0, ifp∈I0, thenSinceand Ii+1=Ii+LfIi, by Lemma 4, one has that

    Consider the ideals I0,I1,I2,... defined as in (10). The following lemma states that there is an element INof such a sequence such thatfor allj∈Z≥0.

    Lemma 6:Iff∈Rn[x] and I is an ideal in R[x], define I0,I1,I2,... as in (10). Hence, there isN∈Z≥0, such thatfor alli∈Z≥0and allj∈{1,...,m}.

    Proof:Since I0?I1?I2?... is an ascending chain of ideals in R[x], by Theorem 7 at p. 80 of [33], there existsN∈Z≥0such that IN+j=INfor allj∈Z≥0. By Lemma 5,, for alli∈Z≥0.

    By combining the results given in Theorem 3 and Lemmas 3 and 6, the following lemma shows that if I0=〈h1,...,hm〉,thenis the solution to Problem 1.

    Lemma 7:Let I =〈h1,...,hm〉 and define I0,I1,I2... as in Lemma 6. Hence, the solution to Problem 1 is V (IN).

    The following example shows how to use the results given in Lemma 7 to determine the solution to Problem 1.

    Example 5:The objective of this example is to study the stability of the origin of

    Consider the candidate Lyapunov function

    Thus, the largestf-invariant set contained in V(Lf V(x)) isand, by Theorem 1, the origin is globally asymptotically stable.

    As shown by Example 5, in order to apply the method proposed in Lemma 7 to solve Problem 1, one has to compute a large number of reduced Gr?bner bases. In the following, it will be shown that such computations can be reduced by exploiting the concept of the radical of an ideal. In fact, letting I0,I1,I2,... be defined as in (10), letBy Lemma 6, there issuch thatj∈Z≥0. Thus, consider the next lemma, whose proof follows by Lem. 5 at p. 182 of [33].

    Lemma 8:Assume thatf∈Rn[x] andh1,...,hm∈R[x]. Let I=〈h1,...,hm〉, define I0,I1,I2,... as in (10), and letThere issuch thatfor alli∈Z≥0andj∈{1,...,m}.

    The following lemma shows how to determine an idealcontainingfori∈Z≥0andj∈{1,...,m}.

    Lemma 9:Assume thatandThus,define the sequence of idealsas

    Proof:By induction, it is proved thatfor alli∈Z≥0and allj∈{1,...,m}. Leti=0. Clearly,hj∈〈h1,...,hm〉, whence by Lemma 5 at p. 182 of [33],Now, for anyi∈Z≥0, assume thatBy Lemma 4, since one has thatwhence

    by Lemma 5 at p. 182 of [33]. Sinceis an ascending chain of ideals inby Theorem 7 at p. 80 of[33], there existssuch thatfor allj∈Z≥0.

    In the following theorem, whose proof follows directly from Theorem 3 and Lemma 9, a method is proposed to determine a solution to Problem 1 in the polynomial case.

    Theorem 4:Assume thatf∈Rn[x] andh1,...,hm∈R[x],and defineas in (11). Lettingbe such thatfor allthe largestf-invariant set contained inis

    The following Algorithm 1 summarizes the computations that have to be carried out to solve Problem 1 in the polynomial case by using Theorem 4.

    Algorithm 1 Solution to Problem 1

    Note that in order to apply Algorithm 1, just the knowledge of the polynomialfandh1,...,hmis required. Once a monomial order is fixed, Step 2 of Algorithm 1 can be carried out by comparing the reduced Gr?bner basesandof the respective idealsandAs a matter of fact, by Theorem 5 at p. 120 of [33], for a given monomial order, each ideal has a unique reduced Gr?bner basis, whenceif and only ifNote that, by Lemma 9, such an algorithm entails a finite number of iterations. The following Fig. 4 illustrates Algorithm 1.

    Fig. 4. Illustration of Algorithm 1.

    The following example is an application of Algorithm 1.

    Example 6:Consider again the system and the Lyapunov functionVgiven in Example 5. By using Algorithm 1 with inputfandLf V, one obtains thatandThus, sincethe algorithm terminates by returningNote that, in order to obtain the same result by using Lemma 6, one has to compute the reduced Gr?bner bases of 9 ideals, whereas, by using Algorithm 1, just 3 reduced Gr?bner bases have to be computed.

    The following remark details how to compute the restriction of the dynamics (3) to anf-invariant set S .

    Remark 3:By [42], if the set S isf-invariant andis not a critical point, thenfis tangent to S inthus allowing to easily determine the restriction of the dynamics of system(3) to the set S. As a matter of fact, define

    and the tangent space of S inas

    By [33], if S=〈p1,...,p?〉, thenwhich is a translation of a linear subspace of Rn, i.e., there existb1,...,bs∈Rn(dependent onsuch thatTherefore, for each non-criticalone hasi.e., there existc1,...,cs∈R

    (dependent onsuch that

    This parametrization constitutes the restriction of the dynamics (3) to S.

    IV. INVARIANT SETS FOR DISCRETE-TIME SYSTEMS Consider the discrete-time, time-invariant system

    Definition 2:The setis

    ● positivelyf-invariantiff(S)?S;

    ● negativelyf-invariantiff(S)?S;

    ●f-invariantiff(S)=S.

    The main goal of this section, formalized in the following problem, is to determine the largest positivelyf-invariant set contained in an analytic set.

    Problem 2:Let functionsbe analytic in U. Find the largest positivelyf-invariant set contained in V (h1,...,hm).

    Note that, for discrete-time systems, the attention can be focused just on positivelyf-invariant sets, when considering the following two theorems.

    Theorem 5 (Invariance principle [5], [43]):Let Ω be a subset of U and letV:Ω →R be a continuous function such that Δf V(x)-V(x)≤0 for allx∈Ω. Let S be the largest positivelyf-invariant set contained inwhereis the closure of Ω . If ψf(t,x0)∈Ω for allt∈Z≥0and ψf(t,x0) is bounded, then there existssuch thatast→∞.

    Theorem 6 (Lyapunov stability [44]):Letand letbe differentiable, radially unbounded, and positive definite. If Δf V(x)-V(x)≤0 for allx∈Rn, then any response of system (12) tends to the largest positivelyf-invariant set contained in

    Note that, in order to apply both Theorems 5 and 6, one has to determine the largest positivelyf-invariant set contained inf

    In the following Section IV-A, a method to determine the solution to Problem 2 in the analytic case is proposed. An algorithm, which uses the algebraic geometry tools recalled in Section II, is given in Section IV-B to apply such a technique in the polynomial case.

    A. Solution to Problem 2 in the Analytic Case

    The main objective of this section is to provide a solution to Problem 2 when both the entries offandh1,...,hmare analytic in U. Consider the following lemma.

    Lemma 10 (Existence of responses [45]):The DT-flow ψf(t,x0) is well defined for each

    Note that the statement of Lemma 10 holds just fort∈Z≥0.On the other hand, uniqueness of the responses of system (12)need not hold backward in time, sincefneed not be bijective.Consider the following theorem, which shows how to determine the solution to Problem 2 in the analytic case.

    Theorem 7:Letandj=1,...,m, be analytic functions in U. The largestf-invariant set contained inis

    Proof:By the definition of the DT-flow ψf(·,·) and by Lemma 10, for allandone has that

    Therefore, since the largest positivelyf-invariant set contained in V (h1,...,hs) is

    this set equals the set given in (13).

    In order to apply the technique proposed in Theorem 7 to solve Problem 2, one has to compute the intersection of infinitely many sets, i.e.,

    As for continuous-time systems, such computations can be simplified, as shown in the following proposition, whose proof is similar to the proof of Proposition 1.

    Proposition 2:Assume that there exist a set {q1,...,q?} of analytic functions in U such that

    for some αk,χ,k=1,...,? , being analytic in U. Hence, there exist βi,j,kbeing analytic in U such that, forj=1,...,?,

    Moreover, ifq1(x0)=···=q?(x0)=0 thenq1(x(k))=···=q?(x(k)), for allk∈Z≥0.

    By Proposition 2, if there isN∈Z≥0such thatcan be expressed as a linear combination of,...,m,i=0,...,N, with analytic coefficients, then

    i.e., as for continuous-time systems, just the intersection of a finite number of sets needs to be determined.

    Example 7:Consider the system [43]

    By Theorem 6, trajectories of (16) tend to S. Fig. 5 depicts some trajectories of system (16).

    Fig. 5. Set V (h) (green), positively invariant set S (red), and trajectories of system (16).

    In the following subsection, it is shown that, if bothfandhare polynomial functions, then there existsN∈Z≥0such thatcan be expressed as a linear combination ofthus allowing the design of an algorithm to solve Problem 2.

    B. Solution to Problem 2 in the Polynomial Case

    The main objective of this subsection is to show that the computations required to solve Problem 2 are simpler in case of polynomial systems, for which U=Rn. By Lemma 3, iff∈Rn[x] andh1,...,hm∈R[x], then the set S given in (14)can be determined by computing a finite number of intersections. In the following, it is shown how such computations can be carried out by using the tools recalled in Section II.

    Let a monomial order ? be fixed. Given an ideal I in R [x],letbe the reduced Gr?bner basis of I. The directional successor of I alongfis

    Given a polynomialpin some ideal I, the next lemma states that the directional increment alongfofpbelongs to the sum of I and of its directional increment ΔfI.

    f∈Rn[x] I R[x]p∈I

    Lemma 11:If and is an ideal in , then implies that

    In view of Lemma 11, consider the following lemma.

    Lemma 12:Iff∈Rn[x] andh1,...,hm∈R[x], then define the sequence of ideals I0,I1,I2,... as

    There existsN∈Z≥0such that

    Furthermore, the solution to Problem 2 is V (IN).

    Proof:Note that the sequence I0,I1,... is such that

    The following example shows how the tool given in Lemma 12 can be used to solve Problem 2.

    Example 8:Consider the following system

    Consider the candidate Lyapunov functionIt can be easily seen that

    which is negative semi-definite about the origin. Now, one can compute the largest positivelyf-invariant set contained in V(Δf V(x)-V(x))by using the sequence of ideals defined in Lemma 12. Thus, letI1=I0+ΔfI0and I2=I1+ΔfI1. By computing the reduced Gr?bner basesandof I1and I2, respectively, one haswhence I1=I2. In particular, the reduced Gr?bner basis of I1iswhere:

    Thus, by Lemma 12, the largest positivelyf-invariant set contained inis

    As for continuous-time systems, the concept of radical of an ideal can be used to reduce the number of Gr?bner bases to be computed, as shown in the following theorem.

    Theorem 8:Assume thatf∈Rn[x] andh∈Rm[x]. Define the sequence of ideals K0,K1,K2,... as

    There existssuch thatj∈{1,...,m} , and such that the largest positivelyf-invariant set contained in

    Proof:First, it is proved, by induction, that

    Since, by Lemma 5 at p. 182 of [33],one has that (21) holds fori=0. Assume now that (21) holds for someHence, by Lemma 11,

    thus concluding the induction.

    Secondly, it is proved that there existssuch thatfor allj∈Z≥0. By considering that

    is an ascending chain of ideals in R [x], by Theorem 7 at p. 80 of [33], there existssuch thatfor allj∈Z≥0. The proof is concluded by the fact that, by Theorem 7, the largest positivelyf-invariant set contained in V(h1,...,hm) is given by V (h1,...,hm,Δf h1,...).

    The following Algorithm 2 resumes the computations to be carried out to solve Problem 2 in the polynomial case by using the method outlined in Theorem 8.

    Note that, as for Algorithm 1, in order to apply Algorithm 2,just the knowledge of the polynomialfandh1,...,hmis required. The following Fig. 6 illustrates Algorithm 2.

    Fig. 6. Illustration of Algorithm 2.

    The following remark details how to determine the restriction of the dynamics (12) to S in a particular case.

    Algorithm 2 Solution to Problem 2.

    Remark 4:If the variety S is rational (i.e., there exists∈Z≥0and rational mappingssuch thatfor allx∈S andfor all σ ∈Rs), then the restriction of the dynamics (12) to S can be easily determined. Namely, lettings∈Z≥0be the dimension of S, the dynamics of system (12) restricted to S are

    V. EXAMPLES OF APPLICATION OF THE GIVEN METHODS

    In this section, several examples are given in order to illustrate the application of the proposed methods.

    The following example shows how Algorithm 1 can be used to determine the set of initial conditions that make the output of a linear system be identically zero.

    Example 9:Consider the linear system obtained either from(3) or from (12) by lettingf(x)=Ax, with

    Consider the linear functionh(x)=Cx, where

    (iv) if (a1-a2)c2=0,c1=0, one has three cases:

    (iv.a) if, then S =V(x2);

    a1≠a2c2=0 S=R2

    (iv.b) if , , then ;

    a1=a2c2=0 S=R2

    (iv.c) if , , then .

    The following example illustrates how the proposed methods can be used to study the stability of a mechanical system by using its energy as Lyapunov function.

    Example 10:Consider the mechanical system depicted in Fig. 7, whose dynamics are given by, where

    Fig. 7. A mechanical system.

    x(t)∈R4represents the positions and velocities of the two bodies having massm,kis the stiffness of the springs anddis the damping coefficient of the damper.

    Consider the candidate Lyapunov function

    which is the total energy of the system. It can be easily derived thatLf V=-d(x2-x4)2≤0. Therefore, the origin of the considered system is stable (actually, the sub-level set{x∈R4:V(x)≤c} is positively invariant for anyc∈R≥0). By using Algorithm 1 with inputf(x)=Axandh(x)=Lf V(x),one obtainsand K2=K1. Thus, the largest invariant set contained in V(Lf V)is S=V(x2-x4,x1-x3). Thus, by Theorem 2, every solution ofxapproaches S ast→∞.

    The methods given in this paper can be used even if in the largestf-invariant set contained in V(h1,...,hm) there is a limit cycle, as shown in the next example.

    Example11:Consider system (3) withwhich has been obtained by using [27, Alg. 1] and imposingas limit cycle. Lettingnote thatthus implying that sub-level sets ofVare positively invariant. By using Algorithm 1 with inputf(x) andh(x)=Lf V(x), one obtainsandThus, the largestf-invariant set contained in V (Lf V) is

    The following example shows how the given tools can be used to study the behavior of an averaging algorithm.

    Example 12:Consider the equal-neighbor averaging model[46]x(t+1)=Ax(t), with

    Consider the candidate Lyapunov function

    By using the “complete the squares” procedure [23], one obtains that Δf V-Vcan be written as

    thus guaranteeing that the sub-level sets ofVare positively invariant. By using Algorithm 2 with inputf(x)=Axandh(x)=Δf V(x)-V(x), one obtains K0=〈h〉 andThus, the largest positively invariant set with respect tox(k+1)=Ax(k) contained in V(h)is S =V(x1-x2,x2-x3). Hence, by Theorem 5, each bounded solution of the considered averaging algorithm tends to S.

    The following example shows how Algorithm 1 can be used, with minor modifications, to solve the output zeroing problem and to determine the zero dynamics of a nonlinear controlled system.

    Example 13:Consider the system given in Example 6.1.6 of[42],=f(x)+g(x)u,y=η(x), where

    andy(t)∈R2are the control and output vectors, respectively. The objective of this example is to determine the set of initial conditions and control inputs that make the output of the system be identically zero. By the same reasoning used to prove Lemma 1, one has thaty(t)=0 for all admissiblet∈R if and only if all of its time derivatives vanish. Thus, by computing the time derivative of the output up to the third order, one obtains a set of polynomials in the state of the system and the time derivatives of the input, up to the second order. By computing the reduced Gr?bner basis of the corresponding ideal, one obtains

    where

    Thus, by lettingand lettingh(x)=[x1x2x4x5]T, and by using Algorithm 1 with inputandh, one obtains that the set S =V(h) is the largest invariant set with respect tocontained in V(h).Therefore, the control inputand the set S solve the output zeroing problem. Thus, the method given in Remark 3 can be used to determine the zero dynamics of the system. In fact,Txˉ(S)=V(x1,x2,x4,x5) for eachand the restriction off*(x) to S is

    whence the zero dynamics of the system areNote that the same result has been obtained in [42], by using different arguments.

    It is worth noticing that the procedure outlined in Example 13 does not require neither that the system is input affine nor that the number of inputs is equal to the number of outputs, as shown in the following example.

    Example 14:Consider the plant

    By computing the time derivatives of the output, the reduced Gr?bner basis of the corresponding ideal is that the largest invariant set with respect to the system

    Thus, by lettingandh(x)=η(x), one obtainscontained in V(h) is S=V(h). Therefore,and S solve the output zeroing problem. The method given in Remark 3 can be used to determine the zero dynamics of the system. In fact, one has thatfor eachand the restriction of the vector fieldf*(x) to S isThus, the zero dynamics of the system are

    The following example shows how Algorithm 2 can be used to characterize the observability of system (12) from the output. Consider the discrete-time system

    System (23) is zero-state observable ify(t)=0,t∈Z≥0,impliesx(t)=0,k∈Z≥0[44].

    Example 15:Consider system (23), with

    By using Algorithm 2 with inputfandh, one obtains thatandThus, the algorithm returns as outputi.e., the largest positivelyf-invariant set such thaty(t)=0 for allt∈Z≥0is {0}, whence the system is zero-state observable.

    VI. CONCLUSIONS

    In this paper, some computational tools have been proposed to determine the largest invariant set contained in an analytic set, in the case of both continuous-time and discrete-time nonlinear systems. In particular, it has been shown that, if the vector field governing the dynamics of the system is analytic,such a set can be computed by determining the set of all the solutions of a system of (possibly, infinite) analytic equalities.On the other hand, if the vector field is polynomial and the algebraic set is a variety, then this set can be determined with a finite number of steps. Several applications to control problems have been reported. In fact, as shown in Section V,the techniques proposed in this paper can be used to determine the set of initial conditions that make the output of a linear system be identically zero, to study the stability of dynamical systems, to characterize the behavior of averaging algorithms,to solve the output zeroing problem, to determine the zero dynamics of a nonlinear controlled system, and to characterize the zero-state observability of autonomous nonlinear systems.

    Future developments of the techniques given in this paper will deal with the extension of the proposed algorithms to the hybrid case, i.e., to systems presenting both continuous-time and discrete-time dynamics.

    Note that Algorithms 1 and 2 are computationally tractable in many cases of practical interest. In fact, although computing Gr?bner bases is, in the worst case, an EXPSPACE-complete problem [48], in the generic case,Gr?bner bases can be computed efficiently by the most recent algorithms and hence the algorithms given herein can be actually employed (see [49], [50] for further details).

    Therefore, although the main limitation of the proposed approach is the computation of Gr?bner bases of the idealsit may be used to solve problems of practical interest. For instance, an implementation of Algorithm 1 in Macaulay2 has been used to solve Problem 1 withfandhbeing polynomials in 10 variables of degree 6 within reasonable computing time.

    欧美乱码精品一区二区三区| 亚洲精华国产精华精| 久久久水蜜桃国产精品网| 日韩三级视频一区二区三区| 国产精品一区二区在线观看99| 性高湖久久久久久久久免费观看| 啪啪无遮挡十八禁网站| 国产日韩欧美在线精品| 久久狼人影院| kizo精华| 国产亚洲欧美在线一区二区| 久久人妻熟女aⅴ| 亚洲美女黄色视频免费看| 91国产中文字幕| 国产黄色免费在线视频| 视频区图区小说| 久9热在线精品视频| 99久久国产精品久久久| 国产精品免费视频内射| 久久99热这里只频精品6学生| 桃花免费在线播放| 久9热在线精品视频| 国产区一区二久久| 亚洲综合色网址| 一本—道久久a久久精品蜜桃钙片| 成人国语在线视频| 午夜成年电影在线免费观看| 久久久久久久精品精品| 欧美乱码精品一区二区三区| 岛国在线观看网站| 久久人人爽人人片av| 女性生殖器流出的白浆| 99国产精品免费福利视频| 亚洲欧美精品自产自拍| 久久性视频一级片| 久久香蕉激情| 国产成人啪精品午夜网站| 制服人妻中文乱码| 91精品国产国语对白视频| 色播在线永久视频| 色视频在线一区二区三区| 看免费av毛片| 夜夜骑夜夜射夜夜干| 亚洲欧美日韩另类电影网站| 99国产极品粉嫩在线观看| 成人国产一区最新在线观看| 老司机影院毛片| 日本91视频免费播放| 精品人妻1区二区| 亚洲av成人不卡在线观看播放网 | 午夜影院在线不卡| 免费观看av网站的网址| √禁漫天堂资源中文www| 黄色视频在线播放观看不卡| 国产一卡二卡三卡精品| 国产熟女午夜一区二区三区| 黄色a级毛片大全视频| 性色av一级| 精品乱码久久久久久99久播| 18禁黄网站禁片午夜丰满| 亚洲精品av麻豆狂野| 欧美日韩亚洲高清精品| 亚洲性夜色夜夜综合| 女人精品久久久久毛片| 水蜜桃什么品种好| 国产av一区二区精品久久| 人成视频在线观看免费观看| 欧美大码av| 亚洲三区欧美一区| 美女高潮到喷水免费观看| 一区二区av电影网| 亚洲va日本ⅴa欧美va伊人久久 | 亚洲人成电影免费在线| 国产在视频线精品| 日本91视频免费播放| 欧美xxⅹ黑人| 制服人妻中文乱码| 欧美日韩黄片免| 一二三四社区在线视频社区8| 欧美人与性动交α欧美精品济南到| 亚洲精品久久午夜乱码| av在线app专区| 欧美日韩亚洲综合一区二区三区_| 国产区一区二久久| 手机成人av网站| 国产精品久久久人人做人人爽| 久久精品成人免费网站| 亚洲情色 制服丝袜| 999久久久精品免费观看国产| 老汉色av国产亚洲站长工具| 国产一区二区在线观看av| 黑人操中国人逼视频| 国产野战对白在线观看| 亚洲av国产av综合av卡| 一级毛片精品| 国产精品麻豆人妻色哟哟久久| 国产免费福利视频在线观看| 国产人伦9x9x在线观看| 狠狠精品人妻久久久久久综合| av有码第一页| 老司机午夜福利在线观看视频 | 免费在线观看影片大全网站| av有码第一页| 色综合欧美亚洲国产小说| 九色亚洲精品在线播放| 国产在线视频一区二区| 成人国产av品久久久| 精品高清国产在线一区| 精品福利观看| 国产亚洲精品第一综合不卡| 亚洲精品美女久久av网站| 一个人免费在线观看的高清视频 | 少妇的丰满在线观看| 黄片播放在线免费| 一本久久精品| 国产欧美日韩一区二区三区在线| 欧美午夜高清在线| 性高湖久久久久久久久免费观看| 国内毛片毛片毛片毛片毛片| 国产三级黄色录像| 曰老女人黄片| 久久久久国内视频| 久久中文看片网| 亚洲 国产 在线| 国产野战对白在线观看| 欧美日韩中文字幕国产精品一区二区三区 | 一级黄色大片毛片| 日韩一区二区三区影片| 视频区欧美日本亚洲| 啪啪无遮挡十八禁网站| 国产精品秋霞免费鲁丝片| 日日夜夜操网爽| 人人妻,人人澡人人爽秒播| 香蕉丝袜av| 免费久久久久久久精品成人欧美视频| 热re99久久精品国产66热6| 国产欧美亚洲国产| 菩萨蛮人人尽说江南好唐韦庄| 一本综合久久免费| 欧美午夜高清在线| 亚洲av男天堂| 亚洲五月色婷婷综合| 久热这里只有精品99| 狠狠婷婷综合久久久久久88av| 欧美日韩成人在线一区二区| av网站免费在线观看视频| 最黄视频免费看| 丰满迷人的少妇在线观看| 亚洲精品在线美女| 男人舔女人的私密视频| 中文字幕制服av| 99国产极品粉嫩在线观看| 国产精品二区激情视频| 亚洲精品国产精品久久久不卡| 亚洲精品中文字幕一二三四区 | 三级毛片av免费| 国产99久久九九免费精品| 超碰成人久久| 丰满少妇做爰视频| 老司机影院成人| 制服诱惑二区| 亚洲激情五月婷婷啪啪| 在线观看舔阴道视频| 少妇的丰满在线观看| 精品久久蜜臀av无| 精品一区二区三区四区五区乱码| 国产一区有黄有色的免费视频| h视频一区二区三区| 国产精品久久久人人做人人爽| 久久中文看片网| 最近最新中文字幕大全免费视频| 人人妻人人爽人人添夜夜欢视频| 日韩熟女老妇一区二区性免费视频| 亚洲一码二码三码区别大吗| 99精品久久久久人妻精品| 精品国产乱子伦一区二区三区 | 亚洲熟女毛片儿| 国产视频一区二区在线看| av天堂在线播放| 久久国产精品男人的天堂亚洲| 热re99久久国产66热| 亚洲精品国产av成人精品| 秋霞在线观看毛片| 欧美97在线视频| 久久人妻熟女aⅴ| e午夜精品久久久久久久| 国产精品一区二区免费欧美 | 日韩欧美免费精品| 一区二区三区四区激情视频| 国产真人三级小视频在线观看| 中文欧美无线码| 搡老熟女国产l中国老女人| 久热这里只有精品99| 在线观看免费午夜福利视频| 777久久人妻少妇嫩草av网站| 亚洲久久久国产精品| av片东京热男人的天堂| 亚洲中文字幕日韩| 欧美在线一区亚洲| 亚洲精品国产一区二区精华液| 高清欧美精品videossex| a 毛片基地| 99国产极品粉嫩在线观看| 国产99久久九九免费精品| 老司机深夜福利视频在线观看 | 成人手机av| 精品卡一卡二卡四卡免费| 啦啦啦免费观看视频1| 香蕉国产在线看| 亚洲精品国产av成人精品| a级片在线免费高清观看视频| 亚洲国产av影院在线观看| 9热在线视频观看99| 捣出白浆h1v1| 一级毛片电影观看| 波多野结衣av一区二区av| 十八禁人妻一区二区| 天天躁日日躁夜夜躁夜夜| 欧美变态另类bdsm刘玥| 日本wwww免费看| 99热全是精品| 国产免费福利视频在线观看| 免费在线观看日本一区| 91麻豆精品激情在线观看国产 | 精品一品国产午夜福利视频| av片东京热男人的天堂| 在线亚洲精品国产二区图片欧美| 亚洲av国产av综合av卡| 色老头精品视频在线观看| 国产成人欧美在线观看 | 丰满迷人的少妇在线观看| 国精品久久久久久国模美| 国产老妇伦熟女老妇高清| 涩涩av久久男人的天堂| 岛国毛片在线播放| 午夜免费成人在线视频| 国产亚洲欧美在线一区二区| 最新在线观看一区二区三区| 亚洲av男天堂| 99久久99久久久精品蜜桃| 欧美97在线视频| 亚洲激情五月婷婷啪啪| 搡老乐熟女国产| 久久久久久久大尺度免费视频| 国产在线视频一区二区| 嫁个100分男人电影在线观看| 亚洲免费av在线视频| 国产精品一区二区精品视频观看| 久热爱精品视频在线9| 日韩三级视频一区二区三区| 伊人久久大香线蕉亚洲五| 高清av免费在线| 久久av网站| 国产99久久九九免费精品| 香蕉国产在线看| 天天影视国产精品| 亚洲欧美一区二区三区久久| cao死你这个sao货| 亚洲精品一卡2卡三卡4卡5卡 | 捣出白浆h1v1| 五月天丁香电影| 99久久国产精品久久久| 午夜福利免费观看在线| 欧美日韩亚洲高清精品| www.熟女人妻精品国产| 成年美女黄网站色视频大全免费| 成人国产一区最新在线观看| 国产成人av激情在线播放| 亚洲专区国产一区二区| 久久国产精品男人的天堂亚洲| 久久久久国内视频| 国产成人精品久久二区二区免费| avwww免费| a级毛片在线看网站| 日本91视频免费播放| 中文字幕av电影在线播放| 香蕉国产在线看| 在线永久观看黄色视频| 又黄又粗又硬又大视频| 婷婷丁香在线五月| 亚洲精品国产精品久久久不卡| 亚洲国产成人一精品久久久| 国产人伦9x9x在线观看| 久久久久久久久久久久大奶| 无遮挡黄片免费观看| 国精品久久久久久国模美| 嫁个100分男人电影在线观看| 欧美黑人精品巨大| 曰老女人黄片| 各种免费的搞黄视频| 俄罗斯特黄特色一大片| 日韩制服丝袜自拍偷拍| 亚洲精品国产av成人精品| 多毛熟女@视频| 久久天躁狠狠躁夜夜2o2o| 国产精品久久久人人做人人爽| 亚洲精品第二区| 多毛熟女@视频| 国产欧美日韩一区二区精品| 97精品久久久久久久久久精品| tocl精华| 亚洲精品久久成人aⅴ小说| 欧美日韩av久久| 多毛熟女@视频| 色精品久久人妻99蜜桃| 午夜福利一区二区在线看| 亚洲欧美精品综合一区二区三区| 久久久精品94久久精品| 99国产精品一区二区蜜桃av | 久久国产精品影院| 亚洲欧洲精品一区二区精品久久久| 久久免费观看电影| 国产主播在线观看一区二区| 亚洲熟女毛片儿| 高清在线国产一区| 久久香蕉激情| 精品国产一区二区三区久久久樱花| 成年av动漫网址| 我的亚洲天堂| 国产成人av教育| 欧美日韩国产mv在线观看视频| 久热这里只有精品99| 两性夫妻黄色片| 久久九九热精品免费| 电影成人av| 亚洲伊人久久精品综合| 亚洲欧美日韩高清在线视频 | 欧美久久黑人一区二区| 午夜福利影视在线免费观看| 国产成人欧美在线观看 | 精品亚洲成国产av| 久久久久久人人人人人| 欧美变态另类bdsm刘玥| 99国产精品一区二区三区| 免费观看a级毛片全部| 久久中文看片网| 国产亚洲精品第一综合不卡| 免费一级毛片在线播放高清视频 | 亚洲 国产 在线| 不卡av一区二区三区| 伦理电影免费视频| 免费人妻精品一区二区三区视频| 国产在线免费精品| 午夜福利,免费看| 99久久人妻综合| av不卡在线播放| 久久久久久久国产电影| 亚洲国产精品一区二区三区在线| 青春草亚洲视频在线观看| 9色porny在线观看| 亚洲国产av新网站| 天堂中文最新版在线下载| 亚洲精品美女久久av网站| 日韩 欧美 亚洲 中文字幕| 嫩草影视91久久| 国产91精品成人一区二区三区 | 久久精品aⅴ一区二区三区四区| 老熟妇乱子伦视频在线观看 | 国产成人免费观看mmmm| 国产黄色免费在线视频| 国产在线免费精品| 国产精品国产av在线观看| 性色av一级| av有码第一页| 精品国产一区二区三区四区第35| 国产高清视频在线播放一区 | 狂野欧美激情性bbbbbb| 精品国产乱子伦一区二区三区 | 免费黄频网站在线观看国产| 少妇粗大呻吟视频| 国产精品一区二区精品视频观看| 91老司机精品| 久久久久视频综合| 中文欧美无线码| 欧美日韩成人在线一区二区| 1024视频免费在线观看| 久久天躁狠狠躁夜夜2o2o| 国产成人精品无人区| 国产精品.久久久| 80岁老熟妇乱子伦牲交| 伊人久久大香线蕉亚洲五| 999精品在线视频| 欧美日本中文国产一区发布| 欧美激情 高清一区二区三区| 亚洲成人免费av在线播放| 91字幕亚洲| 三上悠亚av全集在线观看| 亚洲精品在线美女| 不卡一级毛片| 日韩一区二区三区影片| 少妇 在线观看| 一区二区三区乱码不卡18| 国产免费一区二区三区四区乱码| 窝窝影院91人妻| 精品高清国产在线一区| 中文字幕最新亚洲高清| 亚洲第一欧美日韩一区二区三区 | 91老司机精品| 欧美精品高潮呻吟av久久| 亚洲熟女精品中文字幕| 亚洲精品久久午夜乱码| 欧美另类一区| 久久久久久免费高清国产稀缺| www.999成人在线观看| 国产欧美日韩综合在线一区二区| 欧美 亚洲 国产 日韩一| 久久久欧美国产精品| 久9热在线精品视频| 中文字幕色久视频| 熟女少妇亚洲综合色aaa.| 性色av一级| 久久女婷五月综合色啪小说| 国产高清国产精品国产三级| 亚洲精品一区蜜桃| 美女大奶头黄色视频| 色播在线永久视频| 日韩精品免费视频一区二区三区| 久久九九热精品免费| 日本vs欧美在线观看视频| 国产男人的电影天堂91| 美女国产高潮福利片在线看| h视频一区二区三区| 中国美女看黄片| 亚洲成人手机| 18禁国产床啪视频网站| 18禁裸乳无遮挡动漫免费视频| 最新的欧美精品一区二区| 黄片大片在线免费观看| 国产熟女午夜一区二区三区| 国产av又大| 99热国产这里只有精品6| 老司机午夜福利在线观看视频 | 国产高清国产精品国产三级| av在线app专区| 中文字幕人妻丝袜制服| 久久国产精品人妻蜜桃| 真人做人爱边吃奶动态| 国产免费一区二区三区四区乱码| 一边摸一边抽搐一进一出视频| 777久久人妻少妇嫩草av网站| a级毛片黄视频| 婷婷成人精品国产| 亚洲 国产 在线| av在线老鸭窝| 欧美精品高潮呻吟av久久| 欧美变态另类bdsm刘玥| 久久久久精品人妻al黑| 黑丝袜美女国产一区| 交换朋友夫妻互换小说| 久久精品成人免费网站| 亚洲精品日韩在线中文字幕| 99香蕉大伊视频| 国产亚洲欧美精品永久| 黄片播放在线免费| 久久 成人 亚洲| 9热在线视频观看99| 国产野战对白在线观看| 国产精品香港三级国产av潘金莲| 一个人免费在线观看的高清视频 | 久热爱精品视频在线9| 少妇猛男粗大的猛烈进出视频| 女性生殖器流出的白浆| 中文字幕最新亚洲高清| 国产免费av片在线观看野外av| 午夜激情久久久久久久| av在线老鸭窝| 日本撒尿小便嘘嘘汇集6| 不卡一级毛片| 欧美日韩成人在线一区二区| 丝袜脚勾引网站| 亚洲国产日韩一区二区| 丰满人妻熟妇乱又伦精品不卡| 无遮挡黄片免费观看| 精品一区二区三区四区五区乱码| 超碰97精品在线观看| 这个男人来自地球电影免费观看| 丝袜脚勾引网站| 亚洲七黄色美女视频| 无限看片的www在线观看| 午夜福利免费观看在线| 看免费av毛片| 国产精品久久久人人做人人爽| 精品欧美一区二区三区在线| 女人被躁到高潮嗷嗷叫费观| 国产一区二区三区综合在线观看| 国产一区二区 视频在线| 美女午夜性视频免费| 99久久精品国产亚洲精品| 国产无遮挡羞羞视频在线观看| 色精品久久人妻99蜜桃| 老司机深夜福利视频在线观看 | 欧美黄色淫秽网站| 亚洲精品一区蜜桃| 制服诱惑二区| 黄色毛片三级朝国网站| 国产又爽黄色视频| 女人精品久久久久毛片| 老司机在亚洲福利影院| 亚洲国产日韩一区二区| 国产97色在线日韩免费| 久久久久国产精品人妻一区二区| 国产成人精品久久二区二区91| 脱女人内裤的视频| av视频免费观看在线观看| 亚洲成av片中文字幕在线观看| 在线看a的网站| 99热国产这里只有精品6| 青青草视频在线视频观看| 在线 av 中文字幕| 婷婷成人精品国产| 亚洲第一av免费看| 黄频高清免费视频| 视频区图区小说| 精品一区在线观看国产| 一本综合久久免费| 中文字幕人妻丝袜制服| 老司机深夜福利视频在线观看 | www.999成人在线观看| 免费久久久久久久精品成人欧美视频| 国产精品久久久av美女十八| 免费不卡黄色视频| 精品高清国产在线一区| 国产成人系列免费观看| 精品久久蜜臀av无| 一区福利在线观看| 成人影院久久| 91精品三级在线观看| 国产成人免费无遮挡视频| 黑丝袜美女国产一区| 不卡av一区二区三区| 欧美精品人与动牲交sv欧美| 久久国产精品影院| 亚洲精华国产精华精| 精品少妇一区二区三区视频日本电影| 视频在线观看一区二区三区| 亚洲熟女毛片儿| 国产精品偷伦视频观看了| 亚洲精品中文字幕在线视频| 女警被强在线播放| 91成人精品电影| 久久久国产一区二区| 免费黄频网站在线观看国产| 日韩大片免费观看网站| 国产真人三级小视频在线观看| 欧美少妇被猛烈插入视频| 久久中文看片网| 日日爽夜夜爽网站| 狠狠婷婷综合久久久久久88av| 曰老女人黄片| 18在线观看网站| 中文字幕制服av| 亚洲免费av在线视频| 欧美黑人精品巨大| 国产成人啪精品午夜网站| 久久亚洲精品不卡| 久久久精品94久久精品| 久热这里只有精品99| 伦理电影免费视频| 男女床上黄色一级片免费看| 在线观看免费高清a一片| 欧美日韩亚洲综合一区二区三区_| 国产成人精品久久二区二区免费| 午夜福利在线免费观看网站| 国产成人精品久久二区二区免费| 欧美xxⅹ黑人| 一边摸一边抽搐一进一出视频| 亚洲国产精品一区二区三区在线| 777久久人妻少妇嫩草av网站| 久久精品成人免费网站| 国产精品久久久久久精品古装| 午夜老司机福利片| 伊人久久大香线蕉亚洲五| 精品久久久久久电影网| 天天躁狠狠躁夜夜躁狠狠躁| 大片电影免费在线观看免费| 少妇精品久久久久久久| 精品久久久久久电影网| 精品一区二区三卡| 久久久国产一区二区| 久久亚洲精品不卡| 久久久久精品国产欧美久久久 | 91av网站免费观看| 午夜激情av网站| 18在线观看网站| 国产成人欧美| 久久久精品国产亚洲av高清涩受| 女人高潮潮喷娇喘18禁视频| 日本猛色少妇xxxxx猛交久久| 他把我摸到了高潮在线观看 | 国产99久久九九免费精品| 久久女婷五月综合色啪小说| 高清欧美精品videossex| 电影成人av| 久久久精品区二区三区| 国产成人一区二区三区免费视频网站| 午夜影院在线不卡| 国产片内射在线| 91精品三级在线观看| 丰满少妇做爰视频| 欧美精品人与动牲交sv欧美| 婷婷色av中文字幕| 桃花免费在线播放| 两个人免费观看高清视频| 黄色怎么调成土黄色| 午夜福利在线免费观看网站| 欧美人与性动交α欧美软件| 麻豆av在线久日| 老熟妇仑乱视频hdxx| e午夜精品久久久久久久| 国产xxxxx性猛交| 久久综合国产亚洲精品| 国产男女超爽视频在线观看| 色婷婷久久久亚洲欧美| 青草久久国产| 国产日韩欧美亚洲二区|