Optimum Control of Non-Gaussian Linear Stochastic Systems with Inaccessible State Variables

Abstract

This article presents a new result in the optimum control of linear systems with respect to a quadratic performance criterion. It is assumed that the system is subject to additive random disturbance and that some state variables cannot be measured or can only be measured with additive noise. It is well known that when the disturbances and noise are Gaussian random variables, the optimum controller is a certain linear function of the mean of the posteriori distribution of state variables. It is shown here that this result holds without qualification.