Real-time voxelized mesh fracture with Gram–Schmidt constraints

Much previous research about fracture of deformable bodies has focused on physical principles (e.g. energy and mass conservation), leading to simulation methods that are very realistic, but not yet applicable in real-time. We present a stylized animation method for destruction of soft bodies that is...

Full description

Saved in:
Bibliographic Details
Published in:Computers & graphics Vol. 132; p. 104382
Main Authors: McGraw, Tim, Zhou, Xinyi
Format: Journal Article
Language:English
Published: Elsevier Ltd 01.11.2025
Subjects:
Animation
Fracture
Parallel programming
Physics
Simulation
Animation
Fracture
Parallel programming
Simulation
Physics
ISSN:0097-8493
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:Much previous research about fracture of deformable bodies has focused on physical principles (e.g. energy and mass conservation), leading to simulation methods that are very realistic, but not yet applicable in real-time. We present a stylized animation method for destruction of soft bodies that is visually plausible and capable of running at hundreds of frames per second by sacrificing visual realism and physical accuracy. Our method uses a new volume-preserving voxel constraint based on Gram–Schmidt orthonormalization which, when used in tandem with a breakable face-to-face voxel constraint, allows us to animate destructible models. We also describe optional LOD constraints which speed convergence and increase apparent stiffness of the models. The creation pipeline and constraints presented here are designed to minimize the number of partitions needed for parallel Gauss–Seidel iterations. We compare the proposed techniques with shape constraints and the state-of-the-art material point method on the basis of memory usage, computation time and visual results. [Display omitted] •Destructible models in a voxelized visual style are created from watertight meshes.•Position-based dynamics constraints are based on Gram–Schmidt orthonormalization.•Model structure permits an efficient parallelization scheme.•LOD constraints speed up convergence efficiently.
ISSN:0097-8493
DOI:10.1016/j.cag.2025.104382