Um Estudo Comparativo Entre Diferentes Técnicas de Otimização do Treinamento de Neurocontroladores

This paper presents a method of on-line neurocontrol and compares three different optimization methods applied to insuring convergence velocity. The question about the convergence time is critical in neurocontrollers with real time training, which the controller training must be during the sampling period, usually small. We present the results presented by three differents optimization methods in a linear and a nonlinear plant. We will see that different methods conduce to different results in the controller performance.