Abstract
This paper reviews the dynamics modeling and redundancy resolution of redundantly actuated Parallel Kinematics Machines (PKM). While redundant actuation offers significant benefits like singularity avoidance and backlash elimination, it introduces challenges related to kinematic parameter uncertainties. We analyze how geometric errors lead to uncontrollable internal forces and propose amended control schemes—specifically augmented PD and computed torque control—to ensure system stability despite these uncertainties.
Introduction: The Promise and Peril of Redundancy
Parallel Kinematics Machines (PKMs) are prized in industrial applications for their high stiffness and precision. However, their operational workspace is often limited by singularities—mathematical "dead zones" where the robot loses degrees of freedom or becomes uncontrollable.
To overcome this, engineers employ redundant actuation, adding more motors than the robot's degrees of freedom. This strategy not only expands the usable workspace but also allows for "prestressing" joints to eliminate mechanical backlash. However, this redundancy creates a complex control environment. If the physical robot deviates even slightly from its mathematical model due to manufacturing tolerances, the redundant actuators may work antagonistically, generating destructive internal forces that standard controllers cannot detect.
Modeling and Control Methods
The research utilizes multibody system (MBS) dynamics to derive rigorous equations of motion. A central technique involves the Voronets Equations, which transform complex closed-loop kinematics into a simpler "tree" topology by identifying specific "cut-joints."
To resolve the redundancy (i.e., deciding how to distribute force among the extra motors), a weighted right pseudoinverse is used. This allows for goal-oriented control, such as minimizing power consumption. Furthermore, a one-dimensional optimization problem calculates "prestress" parameters ($\sigma$) to ensure actuator forces never cross zero, effectively pinning gears in one direction to prevent the "clunking" associated with mechanical backlash.
The Challenge of Geometric Uncertainty
In standard non-redundant robots, geometric errors typically result in minor positioning inaccuracies. In redundant systems, these same errors manifest as "parasitic" internal forces. Because these forces exist in the "null-space"—meaning they do not result in end-effector motion—a standard controller may be unaware that it is straining the motors toward failure.
The paper introduces a perturbation matrix ($S$) to characterize this interference. When $S$ is non-zero, the robot's actuators are fighting a "ghost" created by the model mismatch. To mitigate this, the author proposes Amended Control Laws. By restricting linear feedback to the subspace of independent coordinates, these modified controllers prevent the accidental amplification of internal stress caused by model uncertainties.
Key Results: Singularity Elimination and Efficiency
The primary advantage of redundancy is the homogenization of the workspace. By adding a third limb to a 2-DOF robot (the RP/2RPR configuration), the singularities typically found at the workspace boundaries are eliminated.
Beyond workspace improvements, the paper provides closed-form solutions for "simply-redundant" systems (those with exactly one extra actuator). This allows for efficient inverse dynamics calculations without the need for computationally expensive real-time matrix inversions. Validation using a 4RRR manipulator model showed that even a 5% joint displacement error creates significant uncontrollable counter-actions, reinforcing the need for the proposed amended control schemes.
Conclusion and Future Outlook
Redundant actuation is a powerful tool for enhancing PKM performance, but it is not a "free lunch." The proposed amended control schemes represent a critical first step in ensuring these systems remain applicable in the face of real-world manufacturing tolerances. Future work must focus on developing more sophisticated robust control concepts that can fully compensate for geometric errors while managing the unique null-space dynamics of redundant parallel robots.
Summary updated