Domain Recurrence and Probabilistic Analysis of Residence Time of Stochastic Systems and Domain Aiming Control (1807.03009v2)

Abstract: The problem of domain aiming control is formulated for controlled stochastic nonlinear systems. This issue involves regularity of the solution to the resulting closed-loop stochastic system. To begin with, an extended existence and uniqueness theorem for stochastic differential equation with local Lipschitz coefficients is proven by using a Lyapunov-type function. A Lyapunov-based sufficient condition is also given under which there is no regularity of the solution for a class of stochastic differential equations. The notions of domain recurrence and residence time for stochastic nonlinear systems are introduced, and various criteria for the recurrence and non-recurrence relative to a bounded open domain or an unbounded domain are provided. Furthermore, upper bounds of either the expectation or the moment-generating function of the residence time are derived. In particular, a connection between the mean residence time and a Dirichlet problem is investigated and illustrated with a numerical example. Finally, the problem of domain aiming control is considered for certain types of nonlinear and linear stochastic systems. Several examples are provided to illustrate the theoretical results.