Adaptive Volumetric Mechanical Property Fields Invariant to Resolution

2026-06-16

AI Summary

Existing 3D assets commonly lack high-fidelity, spatially varying mechanical properties—such as Young’s modulus, Poisson’s ratio, and density—hindering realistic physics-based simulation. To address this limitation, this work proposes AdaVoMP, the first method to leverage Sparse Adaptive Voxels (SAV) representation combined with a sparse Transformer-based autoregressive generative architecture for efficiently modeling high-resolution volumetric mechanical property fields of arbitrary 3D objects. The approach achieves up to a 4096× increase in prediction resolution while maintaining memory efficiency and substantially reducing computational overhead at test time, thereby enabling complex 3D models to be directly employed in high-fidelity deformation simulations.

Abstract

Accurate mechanical properties (or materials) Young's modulus ($E$), Poisson's ratio ($ν$) and density ($ρ$) are essential for reliable physics simulation of digital worlds, but most 3D assets lack this information. We propose AdaVoMP, a method for predicting accurate dense spatially-varying ($E$, $ν$, $ρ$) for input 3D objects across representations, improving the resolution, accuracy, and memory efficiency over the state-of-the-art. The foundation of our technique is a sparse and adaptive voxel structure SAV that efficiently represents both the input 3D shape and the material field output. We replace the fixed-voxel model of the most accurate prior method, VoMP, with a novel sparse transformer encoder-decoder model that learns to generate a unique SAV autoregressively for every input shape to represent its materials, achieving a resolution $16^3\times$ higher than prior art. Experiments show that AdaVoMP estimates more accurate volumetric properties, even with lesser test-time compute than all prior art. This allows us to convert high-resolution complex 3D objects into simulation-ready assets, resulting in realistic deformable simulations.

Problem

• mechanical properties
• 3D assets
• physics simulation
• volumetric material fields
• Young's modulus

Innovation

• AdaVoMP
• sparse adaptive voxel
• volumetric mechanical properties
• spatially-varying material
• transformer-based 3D representation