Minimax control of a second order linear system (Q1088640)

The authors deal with a discrete-time second-order linear stochastic system with a random horizon. They assume that the disturbances form a sequence of independent and identically distributed random variables with a distribution function belonging to an exponential family dependent on an unknown parameter \(\lambda\) with finite variance being a quadratic function of the mean. Further, they assume that the control horizon is a random variable with given distribution, independent of disturbances. The loss function is quadratic, non-negatively defined, dependent on state variables, controls and \(\lambda\). The aim of the paper is, first, to present the results for the Bayes control and, further, to study the minimax approach to the problem. At the end of the paper some remarks about higher order linear systems and minimax control for them are given.