Strong convergence of Wong-Zakai approximations of reflected SDEs in a multidimensional general domain (Q462309): Difference between revisions

The author considers a reflected stochastic ordinary differential equation \[ dX(t)=\sigma(X(t))\circ W(t)+b(X(t))+dL(t), \] \[ |L|(t)=\int_0^tI_{\partial D}(X(s))\,d|L|(s) \] and reflected random ordinary differential equations \[ \dot X^n(t)=\sigma(X^n(t))\dot W^n(t)+b(X^n(t))+\dot L^n(t), \] \[ |L^n|(t)=\int_0^tI_{\partial D}(X^n(s))\,d|L^n|(s) \] with initial conditions \(X(0)=X^n(0)=x_0\) and with a constraint \(X(t),X^n(t)\in\overline{D}\) for \(t\geq 0\). Here \(D\) is a bounded domain in \(\mathbb R^d\), \(\nu(x)\subseteq\mathbb S^{d-1}\) for \(x\in\partial D\) is a non-empty collection of reflecting directions, \(D\) and \(\nu\) are assumed to satisfy a set of geometric, regularity and admissibility assumptions. \(W^n\) is the \(n\)-dyadic adapted piecewise linear interpolation of an \(m\)-dimensional Brownian motion \(W\), \(\sigma\in C^1(\overline D,\mathbb R^d\otimes\mathbb R^m)\) is such that \(\sigma^\prime\) is Lipschitz continuous and \(b:\overline D\to\mathbb R^d\) is a Lipschitz continuous function. The random variables \(|L|(t)\) and \(|L^n|(t)\) stand for the total variation of \(L\) and \(L^n\), respectively, on \([0,t]\). The pairs \((X,L)\) and \((X^n,L^n)\) are the unknowns, \(L\) is assumed to be adapted and continuous with bounded variation on every finite interval and \(dL(t)/d|L|(t)\in\nu(X(t))\) is assumed to hold \(dL(t)\)-a.e. (the definition of \((X^n,L^n)\) being analogous). While it has been known that the solutions \((X,L)\) and \((X^n,L^n)\) under the assumptions of the paper exist, the author proves that, for every \(p>0\) and \(T>0\), \[ \lim_{n\to\infty}\mathbb E\,\left[\sup_{0\leq t\leq T}|X^n(t)-X(t)|^p\right]=0. \]