Point-based Diffusion Model for Predicting Spatio-Temporal Dynamics in Physical Systems

Conventional diffusion models applied to physics prediction rely on grid-based and snapshot-level representations, limiting their adaptability to irregular domains and geometric variability. This study introduces a novel point-wise conditional diffusion framework that enables efficient and generalizable prediction of complex physical systems with diverse and irregular geometries, without relying on fixed grids. We validate this approach across 2D spatio-temporal systems (Eulerian and Lagrangian) and 3D large-scale aerodynamics, demonstrating robust generalization to diverse geometric complexities.

The 9th Asian Pacific Congress on Computational Mechanics/The 7th Australasian Conference on Computational Mechanics (APCOM-ACCM 2025)