Abstract
Parallel Kinematics Machines (PKMs) offer high precision but suffer from limited workspaces and singularities. Actuation redundancy (RA-PKM) can overcome these limitations but introduces internal "fighting" between motors and modeling complexities. This paper proposes an antagonism filter to eliminate parasitic energy waste and a redundant coordinate formulation for robust control across the entire workspace.
Introduction: The Promise and Peril of Redundancy
Parallel manipulators are constrained mechanical systems where an end-effector (EE) is connected to a fixed platform by several linkages. Traditionally, the number of actuators equals the degrees of freedom (DOF). However, adding extra actuators—actuation redundancy—can significantly improve stiffness, acceleration, and workspace usability by eliminating singularities.
As shown in Figure 1, a planar PKM with 2-DOF can be equipped with three actuators. While beneficial, this redundancy creates a situation where actuator forces are not independent, potentially leading to mechanical prestress and internal "fighting" between motors.
The Problem: Antagonistic Forces and Mathematical Fragility
Two major hurdles prevent the widespread adoption of RA-PKMs:
- Internal Stress: In decentralized control systems, small measurement or calibration errors cause actuators to push against each other. This "antagonism" creates internal stress that does not contribute to motion but wastes significant electrical energy.
- Singularity Issues: Most model-based controllers use "minimal coordinates," which become mathematically invalid at certain points in the workspace, leading to software crashes or the need for complex "model switching."
Figure 3 demonstrates how different pairs of actuators might attempt to drive the end-effector along divergent paths due to synchronization errors, resulting in the parasitic forces the authors seek to eliminate.
Proposed Methodology
The research introduces two distinct strategies to address these challenges:
1. Antagonism Filter (AF) for Decentralized Control
For robots using simple PD control, the authors propose an Antagonism Filter. This mathematical projector is applied to control commands to strip away any signal components that would cause internal fighting (the null-space of actuator forces). This ensures only "effective" motion-producing signals reach the motors.
2. Redundant Coordinate Formulation for Advanced Control
To avoid the fragility of minimal coordinates, the authors derive robot dynamics using redundant coordinates. By employing a null-space projector to eliminate Lagrange multipliers, they create a globally valid dynamic model. This allows for Computed Torque Control (CTC) that remains stable even when crossing points that would be singular in traditional models.
Experimental Results and Discussion
The methods were validated on a planar 2-DOF RA-PKM prototype. Key findings include:
- Energy Efficiency: The Antagonism Filter reduced electrical energy consumption by 65%, 54%, and 36% across the three motors by eliminating internal fighting.
- Robustness: The redundant coordinate CTC successfully guided the robot through trajectories that crossed multiple parameterization singularities, maintaining smooth motor torques where standard controllers would fail.
- Steady-State Performance: The filter effectively removed antagonistic torques at rest, typically caused by encoder quantization.
Conclusion
While these techniques significantly improve the efficiency and reliability of redundant parallel manipulators, they are not without trade-offs. The Antagonism Filter remains dependent on an accurate kinematic model (the Jacobian), and the redundant coordinate approach increases computational overhead. Future work will focus on extending these models to 3D spatial manipulators and exploring active stiffness control.
Summary updated