DefinePK

Abstract

The article devotes to the problem of controlling discrete nonlinear dynamical systems under non-parametric uncertainty. This term refers to the situation when the model structure of the object under study remains unknown to the parameters. Non-parametric models are considered, in which the difference equation degree of a dynamic process model is refined based on the rule of selection of significant variables. According to this rule, only those variables are included in the non-parametric model, whose optimal coefficient of kernel function is minimal. Non-parametric dual control algorithms are given. Control devices based on these algorithms perform not only the function of controlling the object itself, but also its study. The learning process of the dual control system with active accumulation of information is analyzed. The results of the computational experiment using the non-parametric adaptive dual control algorithm are given in detail. In the simulation the characteristics of objects were described by equations with different degrees of nonlinearity whose form was unknown. Equation form in the process of active accumulation of information was automatically restored on the base of input and output process variables measurements. The above computational experiments confirmed the possibility of using non-parametric algorithms to control nonlinear systems.