Description: In this paper it is shown, how can be synthesized models of the optimal control on the basis of samples or precedents. The proposed BOMD approach is based on empirical induction and directed to obtaining regularities in the form of empirical optimization models which are synthesized in analytical form. We follow the Kolmogorov idea about regularity as non-randomness. This allows us to estimate the probability of non-random model selection from the set of admissible models which are consistent to the sample or to given initial data. The proposed methods and algorithms can be applied to solve wide range of tasks of intelligent control, in particular, in Robotics.