FluidFormer: Transformer with Continuous Convolution for Particle-based Fluid Simulation
This work addresses the need for more stable and accurate neural fluid simulators for applications in computer graphics and physics-based modeling, representing a new paradigm rather than an incremental improvement.
The authors tackled the problem of error accumulation in learning-based fluid simulation by proposing FluidFormer, a Transformer architecture that integrates local features via continuous convolutions and global dependencies via self-attention. This approach achieved state-of-the-art performance with improved stability in complex fluid scenarios.
Learning-based fluid simulation networks have been proven as viable alternatives to traditional numerical solvers for the Navier-Stokes equations. Existing neural methods follow Smoothed Particle Hydrodynamics (SPH) frameworks, which inherently rely only on local inter-particle interactions. However, we emphasize that global context integration is also essential for learning-based methods to stabilize complex fluid simulations. We propose the first Fluid Attention Block (FAB) with a local-global hierarchy, where continuous convolutions extract local features while self-attention captures global dependencies. This fusion suppresses the error accumulation and models long-range physical phenomena. Furthermore, we pioneer the first Transformer architecture specifically designed for continuous fluid simulation, seamlessly integrated within a dual-pipeline architecture. Our method establishes a new paradigm for neural fluid simulation by unifying convolution-based local features with attention-based global context modeling. FluidFormer demonstrates state-of-the-art performance, with stronger stability in complex fluid scenarios.