Genetic rule induction in the design of computed-torque/fuzzy-logic controllers for robotic manipulators

Abstract

In this paper, genetic algorithms are used to design computed-torque/fuzzy-logic controllers for robotic manipulators. It is shown that this use of genetic algorithms provides a very effective means of determining the optimal set of fuzzy rules as well as the optimal domains of the associated fuzzy sets of the fuzzy-logic components of such controllers. It is demonstrated that these computed-torque/fuzzy-logic controllers are more robust than computed-torque/fuzzy-logic controllers in which only the domains of the fuzzy sets are genetically determined.