Combined Method of Neurocontrol for Nonlinear Non-Stationary Object

EasyChair Preprint 4435

Abstract

Problem of finding of optimal turning parameters for neurocontrol of nonlinear non-stationary object in a presence of random disturbances is formulated. Combined method of neurocontrol of nonlinear non-stationary objects with usage of multi-layer perceptron is proposed. The method consists of two stages. In the first stage, a problem of robust neurocontrol is solved by finding the turning parameters for adaptation algorithm on suggested set of object variants. Found turning parameters for adaptation algorithm are used in the second stage – model-free neurocontrol. In this stage it is a finding of optimal turning parameters for algorithm of model free neurocontrol. Regularization method is used to ensure stability of proposed method. Effectiveness and stability of suggested method was verified by the model example. In the presence of random disturbances, neurocontrol was stable and a degree of damping was more than 50 %.

Keyphrases: control synthesis, neural network, neurocontroller

@booklet{EasyChair:4435,
  author    = {Sergey V. Frolov and Sergey V. Sindeev and Artyom A. Korobov and Anton Y. Potlov},
  title     = {Combined Method of Neurocontrol for Nonlinear Non-Stationary Object},
  howpublished = {EasyChair Preprint 4435},
  year      = {EasyChair, 2020}}