Genetic design of fuzzy-logic controllers for robotic manipulators

In this paper, genetic algorithms are used to design multivariable fuzzy-logic controllers for robotic manipulators using no information regarding the dynamical parameters of such manipulators. In particular, it is shown that genetic algorithms provide a very effective means of determining both the optimal set of fuzzy rules and the domains of the associated fuzzy sets for such fuzzy-logic controllers. It is demonstrated, in the case of a particular direct-drive two-link robotic manipulator, that such fuzzy-logic controllers can be readily robustified