Fast Fluid Simulation via Dynamic Multi-Scale Gridding

Recent works on learning-based frameworks for Lagrangian (i.e., particle-based) fluid simulation, though bypassing iterative pressure projection via efficient convolution operators, are still time-consuming due to excessive amount of particles. To address this challenge, we propose a dynamic multi-scale gridding method reduce the magnitude elements that have be processed, by observing repeated particle motion patterns within certain consistent regions. Specifically, hierarchically generate micelles in Euclidean space grouping particles share similar patterns/characteristics based super-light and scale estimation modules. With little internal variation, each micelle is modeled as single rigid body with only applied representative particle. In addition, distance-based interpolation conducted propagate relative message among micelles. our design, network produces high visual fidelity simulations inference time 4.24 ms/frame (with 6K particles), hence enables real-time human-computer interaction animation. Experimental results multiple datasets show work achieves great simulation acceleration negligible prediction error increase.