Intelligent Control Using Cubic Neural Network for Swinging up and Stabilizing Pendulum

Abstract

In this paper, we discusse an intelligent control method of integrated the cubic neural networks (CNN) proposed by one of the authors proposed. The CNN consists of multilevel parallel processing networks at different degrees of abstraction. By an appropriate qualitative level control, the CNN enables us to realize robust control. In this study we deal with nonlinear or multipurpose control problems via intelligent control of integrated CNNs. As a practical example, the integrated CNN is applied to the swing up and inverted control of pendulum. A neural network integrator is used for integrating swinging up and inverting CNN controls. The genetic algorithm is used for integrating the CNNs for stand up control of the pendulum. As a result of simulation and experiment, it is demonstrated that the integrated CNN controller can perform the stand up control of the pendulum without touching the cart position limit. It is also shown that the integrated CNN controller can perform the stand up control of a pendulum even in the case of sensing failure of the pendulum angle.