Adaptive Neural Fuzzy Observer Design for Flexible Robot Joint Control

Industrial robotic arms are one of the most important and practical areas of robotics that have the ability to operate in hazardous, unpredictable situations and in cases where humans themselves are unable to perform. However, the high nonlinearity of the robot arm descriptive equations has complicated the design of suitable controllers for different operating conditions. Therefore, their proper performance depends on the use of robust and flexible control methods. In this research, the modeling and control of a robotic arm is investigated. First, the dynamic modeling of the nonlinear robot system in the form of state space extracted using the Lagrange-Euler method and then a new neural-fuzzy control strategy is proposed for tracking the reference points and improving the robustness of the robot. The proposed controller is adaptive and has two separate fuzzy Neural Identifier and Identifier networks that are trained online and offline. Therefore, changing the working position and parameters of robot modeling are not affect the robust, optimal and adaptive performance of the controller. For comparative purposes, the results of applying the proposed control for the robotic arm are compared with the results of applying the PID control for the robot. The simulation results performed with the MATLAB software show well the effectiveness of the control method in damping the oscillations at angles of the robot joints as well as the movement of the robot tool in the Cartesian coordinates .