Neural network approximation of motion trajectory of multidimensional dynamic object with regularization on the basis of predictive model

The problem of designing multidimensional dynamic object phase trajectory on the basis of noisy measurements is considered. The formalized approach to the regularization principle application on the basis of stabilizer built on the phase trajectory predictive model is suggested. The method of RBF network application for the phase trajectory construction minimizing the proposed regularization criterion based on transformation of initial approximation task to equivalent interpolation task is suggested. A numerical example of drill bit trajectory construction, confirming the regulatory properties of the proposed approach is presented.