A control strategy for designing an intelligent controller for highly dynamic/perturbed systems

Abstract

In this paper a control strategy is proposed to deal with highly dynamic systems. It is based on designing an ANFIS based controller, with minimum number of inputs, by training it from the response of a conventional controller working on the same system. The performance of the resultant fuzzy controller is enhanced by optimizing its PID-like gains using Genetic Algorithms. The proposed control strategy is applied to a double inverted pendulum system, when a payload of significant mass is attached to the second link. This issue is important because of its relevance to some applications like humanoid robots. The added mass increases the system perturbation due to its inertial effect. The purpose of the controller is to balance the links around the equilibrium position, defined as the vertical upward one.