In this paper, a comparison in the performance of metaheuristic algorithms was performed for solving the inverse kinematics problem of redundant robot manipulators.
First, a review of several popular metaheuristic techniques and their variants was made. Then, the forward kinematics for the 6-DOF UR5 universal robot and the 7-DOF Motoman SIA20D were obtained using the DH convention. Two optimization problems were defined to evaluate the performance of the different heuristic approaches in order to identify their weaknesses and strengths.
The first one consists in solving the inverse kinematics only for the position of the defined Tool Center Point (TCP) of the manipulators; the second one, which is the major contribution of this work, extends the analysis of the studied algorithms to solve the inverse kinematics for a desired pose of the end-effector of the robot. To validate the robustness of each metaheuristic, 30 trials were performed for each optimization problem. Each trial consisted of finding the joint values for 50 random positions within the robot’s workspace and validating how many optimal solutions were obtained during that trial. The results present the best and worst performing metaheuristics based on the optimal solutions obtained, standard deviation of the optimal solution and the execution time.
Full article (pdf):