Controle de posição utilizando algoritmo genético com minimização de entropia do erro

Abstract

This work proposes the synthesis of controllers applying a Genetic Algorithm, whose objective function is to minimize the entropy of the error. Recent studies demonstrate that methods used in systems that use the mean square error for error estimation do not present satisfactory performance when dealing with non-Gaussian and nonlinear signals, so it was necessary to search for new alternatives to solve more complex problems. The error entropy minimization method has been used in researches and presenting satisfactory performance in this area. The controllers used are data in the form of a transfer function and we searched for the tuning of the parameters of the genetic algorithm in search of better performance for the generated controller. For tests, simulations were performed using MATLAB software and the validation was performed in a torsion plant with MATLAB / Simulink. A comparison with the mean square error method is also presented. Satisfactory results were found for both methods, however, it was observed a longer execution time for the entropy minimization due to the greater complexity of its function, which uses Parzen’s windowing techniques to estimate the probability density function of the error.