Source-linked AI summary
Coverage Planning for Robotic Tooth Preparation in Densely Constrained Environments
Yunwen Li, Chen Chen, Xiangjie Yan, Chang Shu, Jianxia Hou, Shiji Song, Xiang Li
TL;DR
Robotic tooth preparation must remove material precisely in a confined oral workspace while avoiding adjacent structures. This paper presents an autonomous full-crown system combining anatomy-aware toolpath planning with residual-yaw handpiece orientation, achieving 0.117 mm RMSE after best-fit alignment in phantom-head experiments.
Problem
Autonomous tooth preparation requires precise three-dimensional surface shaping in a confined workspace with limited posterior access and adjacent teeth and soft tissues constraining tool motion.
Method
The system generates anatomy-aware toolpaths from technician-designed preparation models while accounting for bur geometry and adjacent-tooth safety, then uses residual yaw to orient the handpiece outward.
Results
0.117 mm RMSE after best-fit alignment, with 90% of surface points within 0.210 mm of the designed tooth in phantom-head experiments.
Takeaways & Limitations
The framework achieves sub-millimeter accuracy while maintaining safe distances from adjacent structures, indicating potential for clinical translation.
Abstract
from arXiv · showhide
Tooth preparation refers to the controlled removal of tooth structure to create an optimal substrate for fixed restorations and is a core procedure in restorative dentistry. Automating this task is particularly challenging for robots because the dental bur must operate within a densely constrained intraoral workspace, where even sub-millimeter deviations can compromise outcomes or damage adjacent structures. This paper presents a novel robotic system for autonomous full-crown tooth preparation. The proposed framework includes: 1) an anatomy-aware toolpath planning algorithm that conforms precisely to a technician-designed preparation model while protecting adjacent teeth, and 2) a clearance-oriented end-effector yaw assignment strategy that allows intraoral access while reducing the risk of soft-tissue interference. Together, these features enable the robot to accurately mill the irregular tooth surface with an average geometric deviation of 0.117 mm (RMSE), achieving both restoration quality and clinical safety. A series of simulations and phantom-head experiments validate the system's feasibility and effectiveness.
I. INTRODUCTION
Autonomous full-crown tooth preparation is difficult because robots must shape complex surfaces accurately while avoiding adjacent teeth and soft tissues in a confined oral workspace. The proposed framework addresses these constraints through anatomy-aware planning and feasible six-axis handpiece execution.
- Tooth preparation requires sub-millimeter control of taper, finishing line, and surface smoothness within a confined, irregular oral cavity.
- Robotic execution is challenging because full-crown preparation requires complex three-dimensional surface shaping alongside strict constraints from posterior access, adjacent teeth, and soft tissues.
- Existing systems remain limited by preparation type or by operating on isolated teeth without accounting for adjacent teeth and soft-tissue interference.
- The framework targets autonomous full-crown preparation that matches a technician-designed geometry, aligning robotic preparation with digitally planned restorative workflows.
- An anatomy-constrained coverage planner accounts for target geometry, dental-bur shape, and adjacent-tooth protection to support accurate and safe toolpaths.
- Residual yaw about the bur axis orients the handpiece body outward from the oral cavity, reducing soft-tissue interference during six-degree-of-freedom execution.
II. RELATED WORK
Prior robotic tooth-preparation systems use laser ablation or handpiece-based robotic arms, but they retain limitations in preparation scope, process effects, or environmental adaptation. This paper motivates anatomy-aware planning for freeform milling in the narrow oral cavity and describes a digitally integrated robotic setup.
- Laser-based systems demonstrated automatic crown preparation but were limited to shoulder-less full-crown designs and faced temperature-control and surface-discoloration challenges.
- Handpiece-based robotic-arm research has addressed veneer preparation, layered full-crown removal, grinding-process modeling, trajectory optimization, and extrusion-force optimization.
- Traditional freeform CNC machining assumes an open workspace, whereas tooth preparation requires toolpath adaptation to narrow conditions with limited freedom to retract or reorient.
- The limited prior adaptation of freeform surface machining to narrow oral conditions motivates anatomy-aware path planning tailored to the oral cavity.
- The proposed system combines intraoral scanning and technician-designed preparation models as inputs for a digitally planned robotic workflow.
- The prototype includes optical tracking, a six-degree-of-freedom UR3e manipulator with a high-speed dental handpiece, a dental chair, and computer control.
IV. METHODS
The method analyzes tooth geometry to construct anatomy-constrained toolpaths while satisfying precision, adjacent-teeth protection, and anatomical-safety requirements. It extracts preparation boundaries, separates tooth regions, and optimizes a fixed local tool frame for subsequent planning.
- Planning constraints: Toolpath planning must satisfy precision, adjacent-teeth protection, and anatomical safety within the constrained oral cavity.The finishing line provides a geometric safety margin for neighboring teeth, while handpiece orientations must avoid soft-tissue collisions.
- Mesh analysis: The mesh-analysis pipeline represents original and prepared teeth as triangular meshes whose facet normals encode local surface orientation.These geometric representations support boundary extraction, region separation, and tool-axis selection.
- Mesh analysis: The finishing line marks the boundary between prepared and unprepared tooth structure and serves as both the first toolpath layer and an adjacent-teeth safety boundary.It is extracted from boundary edges of the designed prepared-tooth mesh.
- Mesh analysis: For premolars and molars, dynamic programming connects radial candidate points into a smooth closed loop separating occlusal and axial surfaces.The selected loop follows upward-facing facets whose normals are approximately 45° upward.
- Mesh analysis: The bur’s fixed local z-axis maximizes the number of target facets whose normals form non-obtuse angles with that direction.The optimization searches the unit sphere, using an indicator objective that is later smoothed with a Sigmoid for differentiable optimization.
- Mesh analysis: For premolars and molars, the local-frame origin is based on the finishing line, whereas incisors and canines use the highest mesh point projected to the finishing-line height.The lateral direction is manually selected toward the outside of the mouth for later handpiece-pose optimization.
B. Precise Path Generation
The path-generation method uses a geometric milling model with the bur held parallel to an optimized axis. It generates cutter locations across layers and pushing directions while using an inward-offset finishing line to protect adjacent teeth.
- B. Precise Path Generation: The path-generation model is purely geometric and keeps the bur parallel to the previously optimized z-axis throughout preparation.This fixed orientation supports consistent geometric conformity during material removal.
- B. Precise Path Generation: The finishing line is inwardly offset by the cutter radius and added as the first layer, establishing the initial shoulder shape and a safety boundary for adjacent teeth.Subsequent cutter locations are generated for all layers and pushing directions.
1) Incisors, Canines, and Axial Surfaces:
For incisors, canines, and axial premolar and molar surfaces, the system generates geometry-conforming toolpaths through layer-by-layer radial cutter placement and contact calculations.
- Incisors, Canines, and Axial Surfaces: The strategy generates closed loops layer by layer from the finishing line until the complete target geometry is reached.
- Incisors, Canines, and Axial Surfaces: Radial sampling varies each path’s vertical position with angular morphology rather than relying on planar iso-height slices.
- Incisors, Canines, and Axial Surfaces: A cylindrical flat-end cutter with a 1.0 mm diameter is pushed into the model along discrete XY-plane angular directions.
- Incisors, Canines, and Axial Surfaces: For each direction, candidate facets, edges, and vertices are examined to calculate the precise cutter-location point from cutter-surface contact.
- Incisors, Canines, and Axial Surfaces: The cutter contacts an edge by projecting and lifting it to the path height, then using the cutter radius to determine contact.
- Incisors, Canines, and Axial Surfaces: The earliest valid contact along the pushing direction is selected as the cutter location, and paths are generated across radial directions and layers.
2) Occlusal Surfaces:
Occlusal premolar and molar surfaces use a downward cutter approach guided by an outside-in spiral, with interpolated layers smoothing the axial-to-occlusal transition.
- Occlusal Surfaces: For premolar and molar occlusal geometry, the cutter drops downward along the z-axis until it touches the designed prepared tooth.
- Occlusal Surfaces: An outside-in spiral guide path determines the locations from which the cutter drops.
- Occlusal Surfaces: Several interpolated layers are added between the uppermost axial and outermost occlusal toolpaths for a smoother transition.
- Occlusal Surfaces: The geometric contact calculations are implemented using a modified version of OpenCAMLib.
C. Adjacent-Teeth Protection
Adjacent-tooth protection constrains toolpaths within a finishing-line boundary, while handpiece orientation addresses soft-tissue interference under incomplete intraoral environment modeling.
- Adjacent-Teeth Protection: Even small toolpath overshoots beyond the target margin can cause unintended contact with neighboring teeth.
- Adjacent-Teeth Protection: The finishing line is offset inward by the cutter radius, and each higher layer must remain within the footprint of the layer below.
- Adjacent-Teeth Protection: Where interproximal clearance is insufficient, the cutter may penetrate the target preparation to keep its surface inside the finishing-line boundary.
- Adjacent-Teeth Protection: Aligning the handpiece y-axis with the tooth frame’s outward y-axis extends the path from five to six degrees of freedom while maintaining a safe buccal-mucosa orientation.
D. Pre-finishing Toolpath
A pre-finishing toolpath gradually removes substantial excess material before accurate finishing, reducing bur-binding risk while preserving adjacent-tooth protection.
- Pre-finishing Toolpath: Direct generation from the target geometry can require removing substantial material, increasing the risk of bur binding.
- Pre-finishing Toolpath: The pre-finishing stage lifts an offset finishing line to multiple heights, creating successive layers that remove material gradually through spiral interpolation.
- Pre-finishing Toolpath: Each pre-finishing layer is lifted from a finishing line shrunk by the cutter radius to protect adjacent teeth.
- Pre-finishing Toolpath: For incisors and canines, pre-finishing uses proximal top-down passes plus limited outside-in labial and lingual passes to reduce preparation time.
E. Handpiece Pose Optimization
The framework extends a five-DOF preparation path to six-DOF motion by selecting residual yaw to orient the handpiece outward, supporting safer intraoral access. Simulations cover three tooth types, while conservative feedrates prioritize safety over preparation speed.
- Handpiece pose optimization: Yaw is selected to align the handpiece body outward from the oral cavity, reducing soft-tissue interference despite incomplete intraoral geometry.The strategy provides implicit avoidance rather than explicit obstacle modeling of the full environment.
- Handpiece pose optimization: The remaining yaw rotates the handpiece body around the bur axis without changing bur-tip position or tool-axis direction.This redundancy determines the handpiece body's orientation around the local bur z-axis.
- Validation scope: The planning algorithm was validated in simulation on an incisor, canine, and molar, followed by a molar phantom-head experiment using a 1 mm-diameter flat-end bur.The experiment used the SF-10 bur.
- Preparation time: 28.4 min was required for the molar preparation under a staged conservative feedrate profile, while incisor and canine preparations took approximately 41 min.The molar profile used 1.5 mm/s for entry and exit, 0.6 mm/s for pre-finishing, 1.4 mm/s for axial shaping, and 0.4 mm/s for occlusal reduction.
- Preparation time: Preparation time depends primarily on layer count and feedrate, with future optimization constrained by material removal rate and cutting force.The current feedrate selection is explicitly safety-oriented.
B. Simulation
Simulation and phantom-head evaluation assessed toolpath execution, registration, surface deviation, and safety. Best-fit alignment yielded sub-millimeter geometric agreement, while the safety boundary and yaw strategy protected surrounding structures.
- Simulation: The simulated toolpath produced the intended tooth shape, a clean finishing line, and protection of adjacent teeth through the proposed safety boundary.Residual material above and outside the finishing line was expected to detach under clinical conditions.
- Experimental validation: The phantom-head evaluation registered the target using optical tracking and compared post-preparation scans with both unprepared-teeth and best-fit alignment references.The two references separate accumulated registration effects from geometric discrepancy.
- Surface accuracy: 0.321 mm RMSE in the in-situ frame included toolpath, registration, and execution errors, whereas 0.117 mm RMSE followed best-fit alignment.The in-situ and best-fit frames measure different combinations of error sources.
- Surface accuracy: 90% of surface points lay within 0.210 mm of the designed tooth after best-fit alignment.The reported signed-distance analysis distinguishes unremoved material from over-reduction.
- Limitations: Layer-wise planning created staircase-like texture that mesh deviation metrics may not fully capture because scanning and surface reconstruction smooth it.More layers, a ball-end bur, or final finishing can reduce the effect.
- Execution safety: The safety boundary protected adjacent teeth, and the yaw strategy produced no observed interference with anatomical tissues or optical markers.Figure 13 contrasts the proposed yaw assignment with possible interference without it.
VI. CONCLUSIONS
The paper presents an autonomous full-crown tooth-preparation system combining anatomy-aware planning with clinically feasible handpiece execution. Phantom-head experiments and evaluations achieved sub-millimeter accuracy while maintaining safe distances from adjacent structures, with future work targeting robustness and efficiency.
- VI. CONCLUSIONS: The system generates toolpaths conforming to technician-designed preparation models and defines feasible handpiece orientations for confined intraoral shaping.The framework integrates anatomy-aware planning with clinically feasible execution.
- VI. CONCLUSIONS: Phantom-head experiments and evaluations achieved sub-millimeter accuracy while maintaining safe distances from adjacent structures.The conclusion identifies potential for clinical translation within the demonstrated scope.
- VI. CONCLUSIONS: Future work will improve registration accuracy and optimize feedrate scheduling offline and online for faster, more reliable operation.The stated targets are robustness and efficiency.