Almost sure stability of some stochastic dynamical systems with memory (Q934149)

Almost sure asymptotic stability of scalar stochastic delay difference and differential equations with nonanticipating memory terms is studied. The first part of the paper deals with implicit difference equations of the form \[ x_{n+1}=x_n+\kappa_n\Phi(x_n)-a_n\Phi(x_{n+1})+f_n(x_n)+g_n((x_l)_{l\leq n})+\sigma_n((x_l)_{l\leq n})\xi_{n+1}, \] which are driven by martingale-differences \((\xi_n)_{n\in{\mathbb N}}\), where \(\Phi,f_n\) are real-valued functions, \(g_n,\sigma_n\) are real-valued functionals, and \(\kappa_n,a_n\) denote nonrandom parameters. Sufficient conditions are deduced such that its solutions converge to zero almost surely. Furthermore, provided it exists, the trivial equilibrium is globally asymptotically stable. A similar result is deduced for the corresponding continuous time Itô-type equation \[ dX(t)=[-\alpha(t)\Phi(X(t))+\beta_0(t)f(X(t))+\beta_1(t)g(X_t)+\lambda(t)]dt+\sigma(t,X_t)dW(t) \] driven by a one-dimensional Wiener process \((W(t))_{t\geq 0}\). In both cases, the proofs are obtained with help of Lyapunov-Krasovskii-type functionals, martingale decomposition and semi-martingale convergence theorems. Moreover, the assumptions are exemplified using various examples. Finally, the authors apply their results as follows: An Itô-type delay differential equation with polynomial nonlinearity is discretized using a so-called partial nonlinear drift-implicit stochastic \(\theta\)-method, in order to reproduce almost sure stability.