Stability of motion with respect to some of the variables for stochastic systems (Q794613)

Certain types of n-dimensional Itô stochastic differential equations \(dx(t)=A(t,x)dt+\sigma(t,x)dW_ t\) with a trivial solution, i.e., \(A(0,x)\equiv 0\) and \(\sigma(0,x)\equiv 0\), are investigated with respect to the (sometimes asymptotic) stability in probability of some of the x- variables, say \((x_ 1,...,x_ m), m<n\). For linear time-homogeneous systems a method of reduction to a lower dimensional linear homogeneous system is presented by combining some of the non-interesting variables to define new ones, using (possible) linear dependencies among them. A similar result for an appropriate class of nonlinear systems is obtained by linearization and using Lyapunov functions. In both cases the desired stability result then follows from global stability of the reduced system. For stochastic control systems the results give a criterion for a feedback control to stabilize the \((x_ 1,...,x_ m)\)-variables. For a class of (nonlinear drift/linear diffusion) systems, \(m=1\) and a single input acting only on one variable \(x_ s\), \(s>1\), a stabilizing feedback control (depending on all variables!) is explicitely constructed. Several nontrivial examples are provided.