Performance Evaluation of Multigrid Brute-Force Solutions of Inverse Kinematics Problem for the Robotis OP2 Humanoid Hand

Abstract

Humanoid robots target to remove human labor from multiple working environments including the ones that were initially constructed for a human. Robot limbs operation requires solving an inverse kinematics problem, and a standard solution could involve algebraic, geometric, or numerical approaches. This paper presents two brute-force off-line approaches for a Robotis OP2 humanoid upper limb positioning via forward kinematics. Both approaches calculate and structure all possible solutions for an end-effector pose within a robot workspace in advance using a powerful PC, in the off-line mode. Several levels of workspace and joint space discretization allow a user to select a required for his/her task level of the solution precision considering available onboard resources of the robot. Different discretization levels were evaluated at an offboard PC and at an onboard computer of the Robotis OP2 humanoid. The solutions with different discretization levels were compared in terms of memory consumption and precision. The solutions were initially obtained in the Gazebo simulation and then successfully validated with a real Robotics OP2 humanoid. The presented analysis might be useful for a discretization level selection under onboard memory limitations while dealing with manipulator kinematics.