Título: Controle de Posição de um Robô Manipulador SCARA utilizando Redes Neurais Artificiais com Treinamento On-Line
Autores: Battistella, Sandro; Pieri, Edson R. De; Stemmer, Marcelo R.
Resumo: The objetive of this work is the evaluation of the employment of artificial neural networks in position control of robot manipulators. Employing a on-line training method with MLP networks, it is possible to obtain more compact structures, avoiding the need for a training set to train the neural network. Acting in a parallel structure together with a classical PD controler, the neural network is capable to compensate the non-linearities, disturbances and non-modelled dynamics existing on the robot structure. The neural controller performance is compared with the classical PD control in the joint space. Numerical simulations of the whole structure of a SCARA robot are shown.
Artigo em pdf: 4cbrn_080.pdf
Arquivo BibTex: 4cbrn_080.bib