Exact Feedback Linearization of Stochastic Control Systems

Abstract

This paper studies exact linearization methods for stochastic SISO affine controlled dynamical systems. The systems are defined as vectorfield triplets in Euclidean space. The goal is to find, for a given nonlinear stochastic system, a combination of invertible transformations which transform the system into a controllable linear form. Of course, for most nonlinear systems such transformation does not exist. We are focused on linearization by state coordinate transformation combined with feedback. The difference between Ito and Stratonovich systems is emphasized. Moreover, we define three types of linearity of stochastic systems -- g-linearity, sigma-linearity, and g sigma-linearity. Six variants of the stochastic exact linearization problem are studied. The most useful problem -- the Ito-g sigma linearization is solved using the correcting term, which proved to be a very useful tool for Ito systems. The results are illustrated on a numerical example solved with help of symbolic algebra.