An extension of the best linear controller to a polynomial controller for non-Gaussian disturbances
Raphael Sivan · IEEE Transactions on Automatic Control · 1964
The problem of finding the optimal controller (or optimal estimator) in the mean square sense for linear systems which are disturbed by Gaussian additive noise is, by now, a completely solved problem [1], [2]. If fact, the optimal controller turns out to be a linear function of the observations.