Learning-Based Inverse Kinematics for Industrial Robots: A DDPG Approach

This paper explores the use of deep reinforcement learning (DRL) for solving the inverse kinematics (IK) problem of a 6-DOF industrial robot. A model-free method based on the DDPG algorithm is developed and compared against a classical analytical solution. While the analytical solver achieves high precision, it lacks flexibility and robustness. In contrast, the Deep Deterministic Policy Gradient (DDPG) agent learns effective joint configurations from interaction, reaching sub-centimeter accuracy in less than one second. The results highlight the potential of DRL as a scalable and adaptable alternative to traditional IK methods in robotics.