Stability of equilibria of randomly perturbed maps (Q523998)

The idea of stabilizing unstable equilibria of dynamical systems by noise originated from the engineering work of \textit{R. Khasminskii} on stochastic stability [Stochastic stability of differential equations. With contributions by G. N. Milstein and M. B. Nevelson. 2nd completely revised and enlarged ed. Berlin: Springer (2012; Zbl 1241.60002)]. In this paper, the authors investigate the following difference equation in \(\mathbb{R}^{d}\), \[ x_{n+1}=(A+B_{n})x_{n}+q(x_{n}), \qquad n\in \mathbb{N},\tag{1} \] where \(q(x)=O(|x^{2}|)\) is a smooth function, \(A\) is a \(d\times d\) deterministic matrix and \(B_{n}\) are independent copies of a random matrix \(B\). In the second section of the paper they consider an initial value problem for the following difference equation \[ x_{n}=M_{n}x_{n-1}q(x_{n-1}), \qquad n\in \mathbb{N}, \tag{2} \] where \(M_{n}\) are independent copies of a \(d\times d\) random matrix \(M\) with \(\mathbb{E}(\left\| M\right\|)<\infty\), \(q: \mathbb{R}^{d}\to \mathbb{R}^{d}\) is a continuous \[ \left| q(x)\right|\leq C_{1}\left\|x\right\|^{2}, \qquad x\in \mathbb{R}^{d}, \, \left\| x\right\| \leq \delta \] for some \(C_{1}, \delta>0\). Further, the initial condition \(x_{0}\) is assumed to the deterministic. In this setting they prove the following (Theorem 2.2). {Theorem.} Suppose that \[ 0< \lambda =-\mathbb{E} \left(\log \left(\left\| M \right\| \right)\right)<\infty. \] Then the equilibrium at the origin of \((2)\) is stable in probability. For arbitrary \(n\in \mathbb{N}\), let \(B_{n}(\epsilon)\) be independent copies of random \(d\times d\) matrices of \(B(\epsilon)\). In the third part of the paper, the authors' aim is to describe a class of mean-zero stabilizing matrices \(B(\epsilon)\) such that the origin becomes a stable equilibrium (with high probability) of the randomly perturbed map \[ x_{n+1}= \left(A(\epsilon)+B_{n}(\epsilon)\right)x_{n}+q(x_{n}), \qquad n\in \mathbb{N}\cup \left\{ 0\right\}.\tag{3} \] The construction of stabilizing perturbation for (3) relies on a class of random matrices (see Lemma 3.2 and Theorem 3.3). Finally, the last section of the paper contains two numerical examples, that illustrate the analysis of the stabilization.