Neural-Networks-Based Inverse Kinematics for a Robotic Manipulator

Abstract

The solution of inverse kinematics problem of robotic manipulator is a fundamental problem in robot control. This paper involves the study of forward and inverse kinematics of a robotic manipulator platform with revolute joints. The (Denavit-Hartenberg) kinematic model of the prescribed manipulator is presented to robot links and joints. In addition inverse kinematics solution has been provided using geometric approach. This solution is used to develop a dataset used to train three artificial feed forward neural networks (ANN). Each network is used to calculate one joint variable using position and orientation of the end effector as an input. Experimental results have shown a good mapping over the working area of the robot with smaller RMS error than that reported in the cited literature.