Solving Inverse Kinematics for 3R Manipulator Using Artificial Neural Networks

Abstract

Manipulator robots are one of the widely famous robotics types. One of the major problems that appear when preparing these robots is the Inverse Kinematic Problem (IKP). Overcoming the IKP in some cases requires solving complex equations, time-consuming, and can lead to singularity solutions. In this paper, a modeling of the forward kinematics for a 3-R manipulator robot was carried out to generate a dataset for supervised learning. Furthermore, a training model for solving the inverse kinematics of this robot with avoiding the singularity points based on Artificial Neural Networks (ANN) is proposed. The reliability of the method was validated by trajectory tracking. The results demonstrate the effectiveness of the robot to track the referenced path successfully with maximum mean absolute error in the X, Y, and Z axes around 1.375 mm.