A Comparison of Damped Least Squares Algorithms for Inverse Kinematics of Robot Manipulators

Abstract

A critical review of inverse kinematics algorithms for robots in presence of kinematic singularities is addressed in this paper. In particular, with the aim to assess the efficiency in handling joint velocity limits and the possibility that the target value is physically not reachable. Real-time control of redundant robot driven by operator can not guarantee, in fact, that the selected target corresponds to a non-singular configuration. On the other hand, activating the regularization, or damping, factor too far from the singularity corresponds to large tracking errors and severe velocity constraints. In addition, as it will be shown in the paper, reaching a singular configuration with an error different from zero usually affects the efficiency of several algorithms. Various approaches presented in the literature have been tested in different scenarios on the Kinova Jaco2, 7 degrees-of-freedom manipulator. Conclusions show that none of the tested algorithms performs in a satisfactory way.