Q Learning Based PSO Algorithm Application for Inverse Kinematics of 7-DOF Robot Manipulator

Abstract

Solving inverse kinematics problems is one of the fundamental challenges in serial robot manipulators. In this study, a learning-based algorithm was developed to minimize the complexity of solving the inverse kinematics problem for a 7-degree-of-freedom serial manipulator. The parameters of the Particle Swarm Optimization algorithm, modified with Q-learning, a reinforcement learning technique, are updated depending on the states. This approach aimed to increase the efficiency of the algorithm in finding solutions. In the simulation studies, two different end positions of the robot, measured in meters, were used to compare the performance of the proposed algorithm. The location error of the proposed algorithm was statistically compared, and meaningful results were obtained regarding the reliability of the outcomes through Wilcoxon analysis. The simulation results demonstrated that the reinforcement learning-based particle swarm optimization algorithm can be effectively used for inverse kinematics solutions in serial robot manipulators.