Optimal Control over Networks with Long Random Delays

This paper studies the effects of stochastic time delays on automatic control systems which uses communication networks. We assume a linear process to be controlled and known delay probability distributions.The contribution of this paper is to extend the theory in(Nilsson 1998) to delays that may be longer than one sample period.Using a quadratic cost we find the optimal full-state-information controller. We also show that the standard Kalman filter is an optimal observer, and that the separation principle holds.