Closed-loop optimal control law for discrete-time LQG problems with a mean-variance objective

Abstract

The classical linear-quadratic Gaussian problem is reexamined in this paper with a purpose to better control the dispersion of its randomly valued performance index. A closed-loop optimal control law is derived for discrete time linear-quadratic Gaussian problems with a mean-variance objective, by developing a nested form of the variance and using nonlinear mathematical programming as a solution scheme.