GPU-accelerated minimum distance and clearance queries

Area of Science:

• Computer-Aided Design (CAD)
• Geometric Modeling
• High-Performance Computing

Background:

• Distance queries are crucial in CAD for tasks like collision detection and design analysis.
• Existing methods for complex NURBS surfaces can be computationally intensive and lack precision.

Purpose of the Study:

• To present practical algorithms for accelerating distance queries on trimmed NURBS surfaces.
• To provide a generalized framework for using GPUs as coprocessors in CAD operations.
• To improve the accuracy and speed of geometric computations.

Main Methods:

• Developed algorithms utilizing programmable Graphics Processing Units (GPUs) for distance queries.
• Implemented a surface bounding-box hierarchy on the GPU to supplement surface data.
• Focused on model space precision for accurate results, unlike image space precision methods.

Main Results:

• Achieved significant acceleration for distance queries on NURBS models.
• Demonstrated at least an order of magnitude speed increase compared to commercial kernels.
• Obtained approximately two orders of magnitude improvement in accuracy over ACIS.

Conclusions:

• The proposed GPU-accelerated algorithms offer a practical and efficient solution for distance queries on NURBS surfaces.
• The framework generalizes GPU usage for various CAD operations.
• The methods provide accurate and fast solutions, outperforming existing commercial software.