Self-Organizing Fuzzy Clustering Neural Networks Controller for Robotic Manipulators

Abstract

This paper presents a self-organizing fuzzy clustering neural network (SOFCNN) controller suitable for motion control of multilink robotic manipulators. It overcomes the defect of traditional PID control which is difficult to control nonlinear and uncertainties event, the defect of simply fuzzy control which can not remove steady error thoroughly, the defect of neural network need tedious computing time which is not adapt to real-time control. The SOFCNN is based on the fuzzy clustering method optimaling training data before learning fuzzy rules, in order to remove redundant data and resolve conflicts in data. The approach not only reduce computational burden of neural network, but also make the control rules reasonable and suitable for the robotic manipulators. The feature of the SOFCNN controller has dynamic self-organizing structure, fast learning speed and flexibility in learning. The simulation results show that is very fine.