TL;DR
This paper introduces a novel, rapid method for coordinate transformation in industrial robot calibration. By aligning "characteristic lines" derived from point clouds, it achieves high accuracy comparable to existing methods but with significantly simpler computation and greater flexibility, making it ideal for dynamic measurement systems.
What Problem Does This Paper Address?
Industrial robots, while precise in repeatability, often suffer from poor absolute positioning accuracy. This is exacerbated by errors accumulating during coordinate transformations, which are crucial for integrating data from multiple sensors (like cameras and laser trackers) into a common frame for calibration. Existing methods for coordinate transformation, often based on solving complex equations from point clouds, can be computationally intensive, prone to ill-conditioned matrices leading to no solutions, and sometimes require specialized sensors with large fields of view. This paper aims to provide a more robust, efficient, and flexible coordinate transformation method that is essential for accurate robot calibration.
Methods
The core of the proposed method is a geometric approach to coordinate transformation, moving away from solving equations. It involves:
- Characteristic Line Construction: Instead of using entire point clouds, the method extracts "characteristic lines" from these clouds to represent coordinate systems.
- Geometric Coincidence: The key idea is to make these characteristic lines coincide through a series of rotations and translations. This geometric alignment directly implies the coincidence of the coordinate systems.
- Decomposition of Rotation: The complex rotation around an arbitrary axis is decomposed into simpler rotations around the X, Y, and Z axes, making the calculations manageable.
- Matrix Transformation Theory: Standard matrix transformation theory is then used to derive the overall transformation matrix from these sequential rotations and translations.
- Iterative Refinement: The process is applied iteratively, first aligning one characteristic line, then another, to achieve full coincidence of the coordinate systems.
This method avoids solving complex systems of equations and is less susceptible to issues arising from ill-conditioned matrices.
Datasets
The paper describes experiments using a KUKA KR 5 arc robot, a TENYOUN 3DMoCap-GC130 photogrammetric system, and a FARO Xi laser tracker. Specific experimental setups involved:
- Calibration Cube: An 80-point cube (200mm x 200mm x 200mm) was used to facilitate the selection of characteristic lines.
- Public Points: Five common points were used for evaluating the accuracy of the coordinate transformation method itself.
- Robot Calibration Points: Sixty points in space were used for robot kinematic parameter calibration.
- Demonstration Experiment: Data was collected across different regions of the robot's workspace to assess the calibration accuracy distribution.
No new benchmark datasets were introduced; rather, existing measurement systems were utilized for evaluation.
Key Insights and Ideas
- Geometric Simplicity: Shifting from equation-solving to geometric alignment of characteristic lines offers a more intuitive and computationally efficient approach to coordinate transformation.
- Robustness to Sensor Errors: The method's reliance on geometric relationships, rather than specific sensor readings in a complex equation, makes it less susceptible to individual sensor inaccuracies or environmental factors.
- Flexibility for Dynamic Systems: The "rapid" nature and simplicity of the method make it highly suitable for on-line calibration systems where coordinate systems might change frequently.
- Error Suppression: The characteristic line method appears to suppress large error terms and avoid error accumulation, which is a significant advantage over some other methods.
Key Results
- High Accuracy: The proposed method achieved a coordinate transformation accuracy comparable to high-end sensors, with average errors of approximately 0.106 mm (photogrammetric) and -0.015 mm (laser tracker) in the X-direction.
- Robot Calibration Improvement: After applying the proposed coordinate transformation and subsequent robot calibration, the position accuracy of the robot manipulator was improved by 45.8%.
- Efficiency: The method is described as "simple and fast," with a short execution time compared to other algorithms.
- Comparison: The method demonstrated accuracy on par with Singular Value Decomposition (SVD) and Quaternion methods, while being simpler and more stable. It outperformed the Three-Point method.
Limitations and Future Work
- Working Range Dependency: The paper notes that calibration accuracy is not consistent across the entire working range of the robot, suggesting the method is more applicable to smaller working ranges.
- Characteristic Line Selection: While the method is robust, the selection of appropriate characteristic lines is crucial for accuracy. The paper mentions that at least two lines are required.
- Future Work: The paper implies that the method's simplicity, speed, and accuracy make it suitable for a wide range of industrial applications where coordinate systems frequently change. Further exploration of its application in more complex multi-sensor fusion scenarios could be a natural next step.
Summary updated