Combined procedure with randomized controls for the parameters' confidence region of linear plant under external arbitrary noise

The new algorithm is proposed for the estimating of linear plant's unknown parameters in the case of observations with arbitrary external noises. It is based on adding of randomized inputs (test perturbations) through the feedback channel. The assumptions about the noise are reduced to a minimum: it can virtually be arbitrary but independently of it the user must be able to add test perturbations. We combine the previous result about asymptotic properties of randomized control strategy with the new one which is followed by a non-asymptotic approach of LSCR (Leave-out Sign-dominant Correlation Regions) method. The new algorithm gives confidence regions for series of finite sets of observations. These regions shrink to the true values of an unknown parameters when number of observations tents to infinity while the algorithm complexity does not increases.