Intelligent stabilization control to arbitrary equilibrium point of double inverted pendulum

This study aims at establishing a robust and effective intelligent control method for nonlinear and complicated systems by extending the integrated Cubic Neural Network (CNN). In the method, an integrator neural network acquires optimum switching and integration of several controllers for a different local purpose by calculating the fitness based on the system object using the genetic algorithm. The proposed method is applied to an equilibrium point transfer and stabilization control of a double pendulum that possesses four equilibrium points, namely Down-Down, Down-Up, Up-Down and Up-Up points. In order to verify the effectiveness of the proposed method, simulations and experiments were carried out. As a result, it was demonstrated that the integrated intelligent controllers can transfer and stabilize the double pendulum from the arbitrary equilibrium points to arbitrary one without touching the cart position limit. In addition, by taking account of energy variation, the double pendulum was transferred from Up-Down to Up-Up and was stabilized without falling down.