On stochastic stabilization of difference equations (Q874390)

The authors study the stochastic difference equation \[ \begin{cases} x_{n+1}=x_n(1+a_nf(x_n)+\sigma_ng(x_n)\xi_{n+1}), &n\geq 0,\\ x_0=\zeta>0,\end{cases}\tag{1} \] where the functions \(f,g:\mathbb R\to \mathbb R\) are continuous, \(\xi_{n+1}\) are independent random variables with \(E\xi_{n+1}=0\), \(E\xi_{n+1}^2=0\), and \(a_n\in \mathbb R\), for all \(n\geq 0\). Under some additional assumptions on \(a_n\), \(\sigma_n\), \(f\) and \(g\), the almost surely (a.s.) asymptotically stability of the solution \(x_n\), \(n\geq 0\), of the problem (1) is proved. The deterministic difference equation \(x_{n+1}=x_n(1+a_nf(x_n))\), \(n\geq 0\), \(x_0=\zeta>0\), with \(a_nf(x)>0\), for all \(x\geq 0\) and \(n\in N\), can be a.s. asymptotically stabilized for sufficiently small \(| a_nf(x)| \), by adding the noise term \(\sigma_ng(x_n)\xi_{n+1}\), where \(\xi_n=1\) with probability \(1/2\) and \(\xi_n=-1\) with probability \(1/2\). The problem (1) with \(a_n=h\) and \(\sigma_n=\sqrt{h}\), \(n\geq 0\), which is a model of an Euler discretization of an Itô type stochastic differential equation, is also investigated from the same point of view, and a bound on the maximum allowable size of \(h\) for a.s. asymptotically stability of the solution is found in terms of the data. The obtained result is finally compared with the similar results in the continuous case.