Inverse Kinematics for Robotic Manipulators Using Iterative Preconditioned Optimization Algorithms

In this paper, we employ previously developed algorithms, the iteratively preconditioned gradient-descent (IPG) algorithm and its variant with momentum terms, to solve the inverse kinematics (IK) problem for robotic manipulators. Our algorithms belong to the class of quasi-Newton methods. By selecting appropriate hyperparameters, the algorithms can potentially achieve faster convergence than commonly used optimization algorithms. A two-link planar robot and a Universal Robots UR3 cobot are used as examples to present the performance of the algorithms.