Advanced Search
Article Contents
Article Contents

Computation of the stochastic basin of attraction by rigorous construction of a Lyapunov function

The research for this paper was supported by the Icelandic Research Fund (Rannís) in the project `Lyapunov Methods and Stochastic Stability' (152429-051), which is gratefully acknowledged

Abstract Full Text(HTML) Figure(2) / Table(3) Related Papers Cited by
  • The γ-basin of attraction of the zero solution of a nonlinear stochastic differential equation can be determined through a pair of a local and a non-local Lyapunov function. In this paper, we construct a non-local Lyapunov function by solving a second-order PDE using meshless collocation. We provide a-posteriori error estimates which guarantee that the constructed function is indeed a non-local Lyapunov function. Combining this method with the computation of a local Lyapunov function for the linearisation around an equilibrium of the stochastic differential equation in question, a problem which is much more manageable than computing a Lyapunov function in a large area containing the equilibrium, we provide a rigorous estimate of the stochastic γ-basin of attraction of the equilibrium.

    Mathematics Subject Classification: Primary: 37B25, 65P40, 93E03, 93E15, 93D30; Secondary: 60H35, 65C30.


    \begin{equation} \\ \end{equation}
  • 加载中
  • Figure 1.  Above: the computed non-local Lyapunov function $ v $ for system (5.1). Below: the function $ Lv $, approximating $ -10^{-3} $

    Figure 2.  Non-local Lyapunov function for system (5.2) with $ \theta = 1 $. The non-local Lyapunov functions looks very similar to the one computed in [3]

    Table 1.  The table shows the Wendland function $ \psi_0(r): = \phi_{8, 6}(cr) $ as well as the related functions $ \psi_1 $ to $ \psi_6 $, defined recursively by $ \psi_{k+1}(r): = \frac{\partial_r \psi_k(r)}{r} $ for $ k = 0, 1, \ldots, 5 $

    $\phi_{8, 6}$
    $\psi_0(r)$ $[46,189(cr)^6+73,206 (cr)^5+54,915(cr)^4+24,500(cr)^3$
    $ +6,755(cr)^2+1,078cr+77]\, (1- cr)^{14}_+$
    $\psi_1(r)$ $-380\, c^2\, [2,431(cr)^5+2,931(cr)^4+1,638(cr)^3+518(cr)^2$
    $+91cr+7]\, (1- cr)^{13}_+$
    $\psi_2(r)$ $12,920\, c^4\, [1,287(cr)^4+1,108(cr)^3+426(cr)^2$
    $+84cr+7]\, (1- cr)^{12}_+$
    $\psi_3(r)$ $-620,160\, c^6\, [429 (cr)^3+239 (cr)^2+55 cr+5]\, (1- cr)^{11}_+$
    $\psi_4(r)$ $112,869,120\, c^8\, [33(cr)^2+10cr+1]\, (1- cr)_+^{10}$
    $\psi_5(r)$ $-4,966,241,280\, c^{10}\, [9cr+1]\, (1- cr)_+^{9}$
    $\psi_6(r)$ $446,961,715,200\, c^{12}\, (1- cr)_+^{8}$
     | Show Table
    DownLoad: CSV

    Table 2.  The table shows the Wendland function $ \psi_0(r): = \phi_{7, 6}(cr) $ as well as the related functions $ \psi_1 $ to $ \psi_6 $, defined recursively by $ \psi_{k+1}(r): = \frac{\partial_r \psi_k(r)}{r} $ for $ k = 0, 1, \ldots, 5 $

    $\phi_{7, 6}$
    $\psi_0(r)$ $[4,096(cr)^6+7,059 (cr)^5+5,751(cr)^4+2,782(cr)^3+830(cr)^2$
    $+143cr+11]\, (1- cr)^{13}_+$
    $\psi_1(r)$ $-38\, c^2\, [2,048(cr)^5+2,697(cr)^4+1,644(cr)^3+566(cr)^2$
    $+108cr+9]\, (1- cr)^{12}_+$
    $\psi_2(r)$ $10,336\, c^4\, [128(cr)^4+121(cr)^3+51(cr)^2+11cr+1]\, (1- cr)^{11}_+$
    $\psi_3(r)$ $-62,016\, c^6\, [320 (cr)^3+197 (cr)^2+50 cr+5]\, (1- cr)^{10}_+$
    $\psi_4(r)$ $3,224,832\, c^8\, [80(cr)^2+27cr+3]\, (1- cr)_+^{9}$
    $\psi_5(r)$ $-354,731,520\, c^{10}\, [8cr+1]\, (1- cr)_+^{8}$
    $\psi_6(r)$ $25,540,669,440\,c^{12}\,(1- cr)_+^{7}$
     | Show Table
    DownLoad: CSV

    Table 3.  The table shows values for $ \psi_{k, i}: = \sup_{r\in[0, \infty)} |\psi_i(r)| r^k $ for the Wendland functions $ \psi_0(r): = \phi_{8, 6}(cr) $ and $ \psi_0(r): = \phi_{7, 6}(cr) $

    $\psi_{k, i}$ $\phi_{8, 6}$$\phi_{7, 6}$
    $\psi_{6, 6}$ $3.148511062 \cdot 10^7\cdot c^6$$3.240130299 \cdot 10^6\cdot c^6$
    $\psi_{5, 5}$ $2.363249538\cdot 10^6\cdot c^5$ $2.588617377\cdot 10^5\cdot c^5$
    $\psi_{5, 4}$ $6.409097287\cdot 10^6\cdot c^6$ $6.534280933\cdot 10^5\cdot c^6$
    $\psi_{4, 4}$ $1.947997580\cdot 10^5\cdot c^4$ $2.262550039\cdot 10^4\cdot c^4$
    $\psi_{4, 3}$ $6.000016519\cdot 10^5 \cdot c^5$ $6.515237949\cdot 10^4 \cdot c^5$
    $\psi_{4, 2}$ $2.215560450\cdot 10^6\cdot c^6$ $2.237953342\cdot 10^5\cdot c^6$
    $\psi_{3, 3}$ $1.807542870\cdot 10^4\cdot c^3$ $2.219149087\cdot 10^3\cdot c^3$
    $\psi_{3, 2}$ $6.618581621\cdot 10^4\cdot c^4$ $7.625999381\cdot 10^3\cdot c^4$
    $\psi_{3, 1}$ $3.172360616 \cdot 10^5 \cdot c^5$ $3.414789975 \cdot 10^4 \cdot c^5$
    $\psi_{3, 0}$ $ 3.1008\cdot 10^6\cdot c^6$ $ 3.1008\cdot 10^5\cdot c^6$
    $\psi_{2, 2}$ $1.970990855\cdot 10^3\cdot c^2$ $2.550970282\cdot 10^2\cdot c^2$
    $\psi_{2, 1}$ $9.418422390\cdot 10^3\cdot c^3$ $1.147899628\cdot 10^3\cdot c^3$
    $\psi_{2, 0}$ $9.044\cdot 10^4\cdot c^4$ $1.0336\cdot 10^4\cdot c^4$
    $\psi_{1, 1}$ $2.767275907\cdot 10^2\cdot c$ $3.766803387\cdot 10^1\cdot c$
    $\psi_{1, 0}$ $2.66\cdot 10^3\cdot c^2$ $3.42\cdot 10^2\cdot c^2$
     | Show Table
    DownLoad: CSV
  • [1] H. Björnsson, P. Giesl, S. Gudmundsson and S. Hafstein, Local Lyapunov functions for nonlinear stochastic differential equations by linearization, In Proceedings of the 15th International Conference on Informatics in Control, Automation and Robotics (ICINCO 2018) - Volume 1, 2018,579–586, .
    [2] M. Buhmann, Radial Basis Functions: Theory and Implementations, volume 12 of Cambridge Monographs on Applied and Computational Mathematics, Cambridge University Press, Cambridge, 2003. doi: 10.1017/CBO9780511543241.
    [3] F. Camilli and L. Grüne, Characterizing attraction probabilities via the stochastic Zubov equation, Discrete Contin. Dyn. Syst. Ser. B, 3 (2003), 457-468.  doi: 10.3934/dcdsb.2003.3.457.
    [4] P. Giesl, Construction of Global Lyapunov functions using Radial Basis Functions, volume 1904 of Lecture Notes in Mathematics, Springer, Berlin, 2007.
    [5] P. Giesl and S. Hafstein, Review of computational methods for Lyapunov functions, Discrete Contin. Dyn. Syst. Ser. B, 20 (2015), 2291-2331.  doi: 10.3934/dcdsb.2015.20.2291.
    [6] P. Giesl and N. Mohammed, Verification estimates for the construction of Lyapunov functions using meshfree collocation, Discrete Contin. Dyn. Syst. Ser. B, in press.
    [7] P. Giesl and H. Wendland, Meshless collocation: Error estimates with application to dynamical systems, SIAM J. Numer. Anal., 45 92007), 1723–1741. doi: 10.1137/060658813.
    [8] S. Gudmundsson and S. Hafstein, Probabilistic basin of attraction and its estimation using two Lyapunov functions, Complexity, 2018 (2018), Article ID 2895658, 9 pages. doi: 10.1155/2018/2895658.
    [9] S. HafsteinS. GudmundssonP. Giesl and E. Scalas, Lyapunov function computation for autonomous linear stochastic differential equations using sum-of-squares programming, Discrete Contin. Dyn. Syst. Ser. B, 2 (2018), 939-956.  doi: 10.3934/dcdsb.2018049.
    [10] N. Mohammed, Grid Refinement and Verification Estimates for the RBF Construction Method of Lyapunov Functions, PhD thesis, University of Sussex, 2016.
    [11] M. J. D. Powell, The theory of radial basis function approximation in 1990, In Advances in Numerical Analysis, Vol. Ⅱ (Lancaster, 1990), Oxford Sci. Publ., pages 105-210. Oxford Univ. Press, New York, 1992.
    [12] R. Schaback and H. Wendland, Kernel techniques: From machine learning to meshless methods, Acta Numer., 15 (2006), 543-639.  doi: 10.1017/S0962492906270016.
    [13] H. Wendland, Error estimates for interpolation by compactly supported radial basis functions of minimal degree, J. Approx. Theory, 93 (1998), 258-272.  doi: 10.1006/jath.1997.3137.
    [14] H. Wendland, Scattered Data Approximation, volume 17 of Cambridge Monographs on Applied and Computational Mathematics, Cambridge University Press, Cambridge, 2005.
  • 加载中




Article Metrics

HTML views(612) PDF downloads(305) Cited by(0)

Access History



    DownLoad:  Full-Size Img  PowerPoint