Neuro-fuzzy-genetic controller design for robot manipulators

Abstract

In this paper, a new form of neuro-fuzzy-genetic controller design for rigid-link flexible-joints robot manipulator applications has been presented. The control algorithm uses fuzzy logic with neural membership functions and a rule base without needing the knowledge of the mathematical model or the parameter values of the robot. The genetic algorithms are applied for fuzzy rules set optimization. The proposed controller is capable of compensating the elastic oscillations at the robot joints. The obtained membership functions and fuzzy rules are implemented with backpropagation feedforward neural networks. The membership functions are modified through a learning process as a fine tuning. Results of computer simulations, applied to four degree-of-freedom rigid-link flexible jointed SCARA robot manipulators, show the validity of the proposed method.