Solving Inverse Kinematics Using Artificial Neural Networks

Abstract

Finding the inverse kinematic solution is among one of the most important task in the field of robotics i.e. provided the position and orientation of the end factor compute the set of joint angles which would represent the desired position of end effector. For this purpose there exist many solutions such as geometric, algebraic and iterative as the complexity of the joint structure of the manipulator increases the computation becomes more difficult and time consuming. In order to elude cumbersome task of calculating manually a Neural Network based approach is proposed to solve Inverse kinematics. In this study inverse kinematics of six-degrees-of-freedom PUMA 560 was solved using ANN. Comprehensive study was conducted to find the finest configuration for this problem.