The article is devoted to development of a method of nonlinear objects control with the evolving multilayer perceptron. The evolving multilayer perceptron is used to build a nonlinear model of the object, which is then used to predict the behavior of a recursive object in the model predictive control. For the neural network training a genetic algorithm is used, what significantly speed up the training process. The simulation results confirm the effectiveness of the proposed control method.
forecasting model, genetic algorithm, multi-layered perceptron
"Reshenye zadachy upravlenyia s prohnozyruiushchei modeliu na osnove эvoliutsyonyruiushcheho mnohosloinoho perseptrona" [The model predictive control problem solving on the basis of the evolving multilayer perceptron],
Information Processing Systems,