Neural Network Based Inverse Kinematic Solution of a 5 DOF Manipulator for Industrial Application

Abstract

This paper showcases the application of Neural Networks for manipulation. Algorithm based approaches work well although are limited in their ability to find solutions sometimes. The tedious task of programming a manipulator can be replaced with Neural Networks which can learn how to solve Inverse Kinematics. We present a pick and place scenario using Neural Network to solve Inverse Kinematics. Their ability to learn from examples make them a good candidate to solve the inverse kinematics problem. For this purpose, a unique configuration of 5 DOF arm is designed to suit industrial needs. To train the network, a dataset of random joint positions is created and forward kinematics is derived for the corresponding joint angles. The joint variables are then fed to a path planner in Simulink and then the process is simulated.