An experimental study of robotic modeling and control using a fuzzy neural network

Abstract In this paper, a fuzzy neural network (FNN) is applied to modeling and control of a robot. Comparisons are made between the FNN and standard back propagation neural networks as well as a commercially available neural network software package for modeling the robot, and observations on the robustness of these networks are made. A number of experiments demonstrate that the FNN can learn faster and more accurately than the standard backpropagation and certain commercial neural networks for: 1) modeling and control of a real robot, 2) solving a benchmark encoder problem. The FNN will be used as a prototype dynamic model and a control component in the control system of a 7 degree of freedom robot.