Research on path planning for clear aligner flexible manufacturing

Kuntao Huang, Sheng Gao, Cheng Zou, Cong Zhang, Yili Peng, Xi LiuScientific 2024 · DOI

Abstract

In recent years, clear aligners have become increasingly popular for correcting dental defects and enhancing appearance. However, the manufacturing process still faces challenges, including low automation and high equipment costs. A significant hurdle is the coordinate system mismatch between the gingival curve point cloud and the dental CAD model. This paper proposes a PCA-ICP registration algorithm, combining Principal Component Analysis (PCA) for coarse matching and an improved Iterative Closest Point (ICP) algorithm for fine alignment. By utilizing z-level dynamic hierarchical searching, the system achieves a maximum registration error of 0.03 mm, which is within the tolerance of robotic machining. The effectiveness of this method is demonstrated through the generation of executable tool paths for ABB industrial robots, enabling flexible cutting and extraction of clear aligners.

What Problem Does This Paper Address?

While the design phase of clear aligners is highly digitized, the final manufacturing step—trimming the plastic aligner from its 3D-printed mold—remains a bottleneck. The primary technical challenge is a coordinate system mismatch. The "gingival curve," which defines the cutting path where the aligner meets the gums, is generated in dental CAD software. However, when the physical mold is mounted for machining, its real-world orientation rarely matches the digital model perfectly.

Point cloud representation of the gingival curve (left) and the dental model (right) highlighting the coordinate mismatch.

Standard alignment algorithms often fail because dental models are geometrically complex and "steep," causing software to converge on "local optima"—mathematically valid but physically incorrect alignments. Consequently, many manufacturers still rely on manual trimming with scissors, a process that is labor-intensive, inconsistent, and slow.

Methods: The PCA-ICP Registration Algorithm

To overcome these limitations, the researchers developed a two-stage registration process designed for high-precision robotic manufacturing.

The proposed PCA-ICP registration algorithm workflow.
  1. Coarse Registration (Improved PCA): The system first uses Principal Component Analysis (PCA) to identify the principal axes of both the dental model and the gingival curve point clouds. A critical improvement in this stage is a check to ensure a unified right-handed coordinate system, preventing the common "flipped axis" error where the software confuses the left and right sides of the dental arch.
  2. Fine Registration (Z-level Dynamic Hierarchical ICP): Standard ICP algorithms are highly sensitive to initial positioning. The researchers identified that the Z-axis (height) is the most critical factor for successful alignment in dental models. Their algorithm "traverses" the Z-axis in 0.25 mm increments, performing ICP registration at each step to identify the absolute global optimum.
  3. Path Planning and Machining: Once the models are perfectly registered, the system generates a 6-axis tool path for an ABB industrial robot. This path is calculated using a local coordinate system between the workpiece and the tool to avoid interference with the steep geometry of the mold.

Key Results and Insights

The study utilized real-world dental CAD data to validate the system's performance. The results demonstrate a significant leap in both precision and industrial efficiency:

  • Precision: The maximum registration error was recorded at 0.03 mm, which is smaller than the mechanical error of the robot itself, ensuring a perfect fit for the aligner.
  • Efficiency: The robotic system completes the trimming process in approximately 15 to 20 seconds per piece. In contrast, manual trimming and polishing typically require 120 to 135 seconds.
  • Computational Speed: The entire registration and path-generation sequence takes between 2.1 and 5.8 seconds, making it suitable for high-volume "flexible manufacturing" environments.

Conclusion

This research provides a robust solution for the automated production of clear aligners. By solving the coordinate mismatch problem through a hybrid PCA-ICP approach, the system eliminates the need for manual labor and reduces production time by over 80%. Future work will explore the use of Deep Learning to further optimize initial pose prediction and test the algorithm on a broader range of complex dental malformations.

Summary updated

Abstract

In recent years, clear aligner can enhance individual appearance with dental defects, so it used more and more widely. However, in manufacturing process, there are still some problems, such as low degree of automation and high equipment cost. The problem of coordinate system mismatch between gingival curve point cloud and dental CAD model is faced to. The PCA-ICP registration algorithm is proposed, which includes coarse match algorithm and improve-ICP registration algorithm. The principal component analysis (PCA) based method can roughly find the posture relationship between the two point clouds. Using z-level dynamic hierarchical, the ICP registration can accurately find the posture between these two clouds. The final registration maximum distance error is 0.03 mm, which is smaller than robot machining error. Secondly, the clear aligner machining process is conducted to verify the registration effectiveness. Before machining, the path is generated based on the well registered gingival curve. After full registration, the tool path is calculated by establishing a local coordinate system between the workpiece and the tool to avoid interference. This path is calculated and generated as an executable program for ABB industrial robots. Finally, the robot was used for flexible cutting of clear aligners and was able to extract products, ensuring the effectiveness of the proposed research. This method can effectively solve the limitations of traditional milling path planning under such complex conditions.
BibTeX
@inproceedings{huang2024research,
  author = {Kuntao Huang and Sheng Gao and Cheng Zou and Cong Zhang and Yili Peng and Xi Liu},
  title = {{Research on path planning for clear aligner flexible manufacturing}},
  booktitle = {Scientific},
  year = {2024},
  doi = {10.1038/s41598-024-64546-y},
}
Figure