Abstract
This study aims at establishing an advanced visual feedback control method by extending the Cubic Neural Network (CNN) intelligent method. In this study, a qualitative controller was designed in consideration of the system characteristics, especially observational noise by using the genetic algorithm. This study deals with a problem of stabilizing a two-dimensional inverted pendulum by feedback control using stereo camera. To verify the noise tolerance of the proposed method, the computational simulations and experiments using a real apparatus were carried out. It was confirmed that the designed CNN controller had noise tolerance.
