Feedforward control using fuzzy logic learning controller

Abstract

A well known method to control dynamic systems is the dynamic model inversion technique. Compensators obtained by this method may be used as feedforward controllers in the open-loop or closed-loop schemes. This technique has been widely used in the control of robotics arms. The method is conceptually simple but often requires a large amount of computation. In order to overcome this problem, a new method to design a fuzzy logic feedforward controller is presented in this paper. When designing a fuzzy logic controller, most of the time is spent in developing the fuzzy rules base to describe how the fuzzy controller should respond to the various inputs. Obtaining the rule base is usually a trial and error process. A starting set of rules are obtained and tested. The results are studied and the rules are then adjusted. The process is repeated until the desired results are achieved. An automated process would greatly simplify the design process of a fuzzy logic controller. In this paper, a two-level learning algorithm is used to obtain the inverse dynamics of a system. Changes to the learning algorithm that improves the performance of the existing algorithm are also presented. The learning algorithm allows the rules to be obtained by processing data collected from the system. The inverse dynamics are then implemented as a feedforward controller. To demonstrate the effectiveness of the learning algorithm, simulation results are presented.