scientific article; zbMATH DE number 1409853

The author considers the stochastic differential equation \[ {\dot a}(t)=a(t)Ax(t)+Bx(t-h)+\sigma(t,x(t)){\dot\xi}(t), \quad x_0=\varphi_0\in H, \tag{1} \] where \(H\) is the set of piecewise continuous measurable functions \(\varphi(s)\in R^{n}\), \(s\leq 0\), with the norm \(\|\varphi\|= \sup_{s\leq 0}(E|\varphi(s)|^2)^{1/2}\); \(f_{t}\), \(t\geq 0\), is a current of \(\sigma\)-algebras; \(\xi(t)\in R^{N}\) is an \(f_{t}\)-measurable standard Wiener process. A solution to equation (1) is called mean square stable if for any \(\varepsilon>0\) there exists a \(\delta>0\) such that \(E|x(t)|^2<\varepsilon\) if \(\|\varphi_0\|^2< \delta.\) If in addition \(\lim_{t\to\infty} E|x(t)|^2=0\), then the zero solution to equation (1) is called asymptotically mean square stable. The author supposes that \(a(t)\) is an \(f_{t}\)-measurable scalar Markov process, independent on the process \(\xi(t)\), with the denumerable set of states \((a_1,a_2,\ldots)\), \(p_{ij}= P\{ a(\tau+t)=a_{j}\mid a(\tau)=a_{i}\}\), \(t,\tau\geq 0\), \(\lambda_{ij}=\lim_{t\to 0}{p_{ij}(t)/t}.\) The author finds sufficient conditions of asymptotic mean square stability for equation (1). For example the following theorem is proved: Let the states of the process \(a(t)\) satisfy the inequalities \(|a_1|\geq |a_2|\geq\ldots \geq c>0\), and let the matrices \[ R_{k}=a_{k}|a_{k}|(A+A')+ ({\widetilde f}_{k}+(|a_{k}|+|a_1|) |B|+\sigma_0^2 |a_1|)I \] be negative definite uniformly with respect to \(k=1,2,\ldots,\) \( {\widetilde f}_{k}= \sum_{i\not=k}(|a_{i}|+|a_{k}|) \lambda_{ki}\). Then the zero solution to equation (1) is asymptotically mean square stable.

Recommendations

Cited in

(68)

This page was built for publication:

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q4940263)