Optimization of fuzzy controllers for industrial manipulators via genetic algorithms

This paper describes a design procedure for the decentralized fuzzy control of a 5-dof robotic manipulator based on Genetic Algorithms (GAs). Compared to traditional PID, fuzzy controllers better lend themselves to the nonlinear, coupled dynamics of industrial manipulators, thanks to their universal approximation capabilities. In addition, GAs allow a full exploitation of the potentialities of fuzzy control, being able to optimize the set of controllers even with relatively scarce information on the plant, exploring large search spaces and using multi-objective merit figures. The preliminary results, obtained on a detailed model of an industrial manipulator developed within the SimMechanics Matlab environment, show the effectiveness of this fully automated procedure: GA-tuned fuzzy controllers guarantee better performances than PID in a wide range of operating conditions.