Solving the inverse kinematics of robotic arm using autoencoders

Abstract

In the modern era, robotics is an attractive field for many researchers since robots are involved in many aspects of everyday life due to the conveniences and solutions that they provide in various daily difficulties. For this reason, the inverse kinematics of robotic arms is a challenging problem that seems more appealing to researchers as years pass by. In this paper, a novel approach to solve this problem is assessed, which is based on autoencoders. In our implementation the goal is not only to find one random (of the infinite solutions) of this problem, but to determine the one that minimizes both the position error between the actual and desired position of the end-effector of the robotic arm and the joint movement. For the training of the Neural Network of the autoencoder, four different types of the loss function and their corresponding results are examined. A robotic arm with three Degrees of Freedom is used for the evaluation of our implementation and the accurate results demonstrate the efficiency and effectiveness of our proposed method.