Problems in the control of redundantly actuated parallel manipulators caused by geometric imperfections

Abstract

Redundant actuation is considered as a way to improve the properties of parallel manipulators as it increases and homogenizes the kinematic dexterity and the stiffness, and eventually eliminates singularities. It further allows for a purposeful distribution of control forces, taking into account secondary tasks, such as optimal force distribution, stiffness control, and backlash avoiding control. The actual model-based control builds upon an exact model of the manipulator. In this paper the effect of kinematic uncertainties on the applicability of established model-based control schemes is analyzed. It is shown that kinematic model uncertainties lead to parasitic perturbation forces that cannot be compensated by the controls. To tackle this problem amended versions of the augmented PD and computed torque control schemes are proposed.