Stratonovich type SDE's with normal reflection driven by semimartingales (Q2767487)

From MaRDI portal





scientific article; zbMATH DE number 1697521
Language Label Description Also known as
English
Stratonovich type SDE's with normal reflection driven by semimartingales
scientific article; zbMATH DE number 1697521

    Statements

    15 October 2002
    0 references
    stochastic differential equations with reflection
    0 references
    Markov properties
    0 references
    semimartingales
    0 references
    stability
    0 references
    Stratonovich type SDE's with normal reflection driven by semimartingales (English)
    0 references
    Let \(Z\) be a given \(\mathbb R^k\) semimartingale with respect to some filtration \(\mathcal F_t\), \([Z,Z]^c\) the continuous part of its quadratic variation, \(f\in C^1(\mathbb R^d,\mathbb R^{d\times k})\) with bounded derivative, \(D\subset\mathbb R^d\), and a random variable \(X_0\in \overline D\). Consider the SDE, for \(t\geq 0\), NEWLINE\[NEWLINE\begin{multlined} X_t= X_0 +\int_0^t f(X_{s^-}) dZ_s +\frac{1}{2}\int_0^t f'f(X_{s^-}) d[Z,Z]_s^c + \Phi_t+\\ +\sum_{0\leq s\leq t} \{\varphi(f\Delta Z_s,X_{s^-})-X_{s^-}-f(X_{s^-})\Delta Z_s\}\end{multlined}NEWLINE\]NEWLINE with the conditions: NEWLINENEWLINENEWLINE1) \(X_t\in \overline D\), and \(\Phi_t\) is a bounded variation continuous process, both \(\mathcal F_t\)-adapted, such that NEWLINE\[NEWLINE \Phi_t=\int_0^t \overline\theta(s) d|\Phi|_s,\;\overline\theta(s)\in N(X_s),\;d|\Phi|\text{-a.e.} \quad\text{and}\quad\Phi_t=\int_0^t \int \{X_s\in\partial D\} d\Phi_s , NEWLINE\]NEWLINE with \(N(x)\) the set of inward normal unit vectors at \(x\in\partial D\). NEWLINENEWLINENEWLINE2) \(\varphi(g,x)\) is the solution at time \(t=1\) of NEWLINE\[NEWLINE y_t=x+\int_0^t g(y_s) ds+\kappa_t,\qquad y_t\in\overline D,NEWLINE\]NEWLINE where \(\kappa\) is a deterministic function with the same properties as the paths of \(\Phi\). NEWLINENEWLINENEWLINENEWLINENEWLINENEWLINEThis model introduces reflection at a boundary to the one studied by \textit{T. G. Kurtz, É. Pardoux} and \textit{P. Protter} [Ann. Inst. Henri Poincaré, Probab. Stat. 31, No. 2, 351-377 (1995; Zbl 0823.60046)]. Note that a solution is specified by giving \(X\), \(\Phi\), \(y\), \(\kappa\) (the last two are also random functions since they depend on the jump points of \(Z\)). In case \(\partial D\) is of class \(C^2\), there exists a unique solution to the problem. Moreover, if \(Z\) is a vector of independent Lévy processes, then \(X\) is a strong Markov process. Existence and uniqueness also hold for less regular boundaries, provided the jumps of \(Z\) are summable. Some results on stability of the solution with respect to the data \(f\) and \(Z\) are also obtained.
    0 references

    Identifiers

    0 references
    0 references
    0 references
    0 references