Metamaterial geometry generation improves novelty and plausibility trade off
ReGDiff: Guided Diffusion in Regulated Latent Space for Exploring Metamaterial Voxel Geometry
Computer Vision and Pattern RecognitionArtificial Intelligence
Summary
Metamaterials are unusual materials whose special properties come from their shape rather than what they are made of. Creating new shapes using cubes (voxels) can be tricky because sticking too close to known shapes keeps them realistic, but exploring new designs risks nonsense shapes. The authors propose REGDIFF, a method that helps generate new, plausible shapes by carefully guiding the design process to avoid being too similar or too strange. Their tests show REGDIFF creates more believable, more diverse, and more innovative metamaterial designs than previous methods.
What this means in practice
- •For metamaterial designers: Generate new metamaterial voxel shapes that balance innovation with realistic structural features for advanced material design.
- •For 3d printing engineers: Create diverse and plausible geometric templates for voxel-based metamaterial components to improve manufacturing trials.
Authors
Wangzhi Zhan, Jianpeng Chen, Dongqi Fu, Dawei Zhou
Abstract
Metamaterials are artificially engineered structures whose mechanical and physical behaviors are strongly shaped by geometry rather than composition. Voxel representation provides a unified format for metamaterial geometry generation, as it can express diverse classes such as truss, shell, and porous structures within a single cubic discretization. However, voxel-based generation faces a plausibility-novelty trade-off: staying close to known geometries helps preserve geometric regularities, while moving away from them is necessary for novelty but may produce degenerate geometries. To address this challenge, we propose REGDIFF, a generative framework that couples voxel representation with latent space regulation and guided diffusion. REGDIFF introduces a repel-and-sink (RAS) mechanism to smooth the latent distribution of plausible geometries, and short-range repulsion (SRR) guidance to discourage generation overly close to known samples while maintaining geometric plausibility. We further contribute a voxel-based benchmark covering truss- and shell-type metamaterial geometries, together with an evaluation module for geometric plausibility, novelty, and diversity. Experiments show that REGDIFF outperforms voxel-based generative baselines, achieving +8.9% in geometric plausibility, +46.4% in novelty, and +128.6% in diversity on average across two datasets. These results suggest that REGDIFF is a strong geometry candidate generator for downstream evaluation. Our code is provided at https://github.com/wzhan24/ReGDiff.