Intelligence in Fuzzy Control of Rigid-Flexible Manipulator using Dynamic Recurrent Neural Network

This article deals with intelligence in fuzzy control of manipulators with fixed base and the last flexible arm. The tip is moving on a path which is pre-specified. It is assumed that the dynamics of the system is unknown and variable. In control based on fuzzy logic, membership functions remain unchanged in the beginning of selected path to the end. This research deals with modifying membership functions during movement in order to improve system response by using neural network. Dynamic recurrent neural network is used to accelerate updating membership functions. Parametric equations of updating network weights are obtained based on reduction of joint angle errors, rate of joint angle errors and bending deformation. To verify the effectiveness of the proposed method, simulation for manipulator is made with three arms. System responses are compared in both the fuzzy controller and the proposed controller with each other. In order to prove the robustness of the control system, an example will be presented. The obtained results and surveys show the effectiveness of the proposed method.