An averaging principle for stochastic evolution equations with jumps and random time delays (Q2131431)
From MaRDI portal
scientific article
Language | Label | Description | Also known as |
---|---|---|---|
English | An averaging principle for stochastic evolution equations with jumps and random time delays |
scientific article |
Statements
An averaging principle for stochastic evolution equations with jumps and random time delays (English)
0 references
26 April 2022
0 references
Let \(\mathbb H\) and \(\mathbb K\) be separable Hilbert spaces, \(U\) a Borel subset of \(\mathbb K\setminus \{0\}\), \(\mathbb Q\) and \(\widehat{\mathbb Q}\) two \((n\times n)\)-matrices generating continuous-time Markov chains, let \(\mathbb Q\) be weakly irreducible with a stationary measure \(\eta\in\mathbb R^n\), define \(\mathbb Q^\varepsilon\) as \[\mathbb Q^\varepsilon:=\frac{\mathbb Q}{\varepsilon}+\widehat{\mathbb Q}, \] let \(r^\varepsilon\) be a continuous-time Markov chain in a finite state space \(\{r_1<\dots<r_n\}\subseteq[0,\infty)\) with the generator \(\mathbb Q^\varepsilon\), let \(\widetilde N\) be a compensated Poisson measure associated to a \(\mathbb K\)-valued Poisson point process with a \(\sigma\)-finite Lévy measure \(\nu\), let \(A\) be an infinitesimal generator of an exponentially decaying analytic semigroup on \(\mathbb H\), let \(f:\mathbb H\times\mathbb H\to\mathbb H\) and \(h:\mathbb H\to L^q(U,\,d\nu;\mathbb H)\) be Lipschitz continuous mappings (for some \(q\ge 2\)), let \(\psi\) be an almost surely bounded initial condition with values in \(D([-r_n,0];\mathbb H)\) such that \[ \mathbb E\,\sup_{-r_n\le s\le 0}\|\psi(s)\|^2<\infty\quad\text{and}\quad\|\psi(t)-\psi(s)\|\le K|t-s| \] hold for some \(K>0\) and every \(s,t\in[-r_n,0]\), and let \(\psi(0)\in\operatorname{Dom}D(-A)^\theta\) for some \(\theta\in(0,\frac{1}{2})\). It is proved that the equation \begin{align*} dX^\varepsilon&=(AX^\varepsilon(t)+f(X^\varepsilon(t),X^\varepsilon(t-r^\varepsilon)))\,dt+\int_Uh(X^\varepsilon(t),z)\tilde N(dt,dz) \\ dX^\varepsilon(0)&=\psi \end{align*} has a unique solution \(X^\varepsilon\) which converges, as \(\varepsilon\to 0+\), to a process \(\overline X\) satisfying \[ d\overline X=(A\overline X(t)+\sum_{i=1}^nf(\overline X(t),\overline X(t-r_i))\eta_i)\,dt+\int_Uh(\overline{X}(t),z)\widetilde N(dt,dz), \] and if \(0<\rho<\beta<1\) and \(p\in[1,\frac{1}{2\theta})\) then \[\mathbb E\,\sup_{t\in[0,T]}\|X^\varepsilon(t)-\overline X(t)\|_{\mathbb H}^{2p}\le C_{p,n}(\varepsilon^{2p\rho\theta}+\varepsilon^{\frac{\beta-\rho}{2}}) \] holds as \(\varepsilon\to 0+\).
0 references
averaging principle
0 references
stochastic evolution equation
0 references
random time delays
0 references
two-time-scale Markov switching processes
0 references
0 references
0 references
0 references
0 references