SURFSUP: Learning Fluid Simulation for Novel Surfaces
This addresses a bottleneck in fluid simulation for applications like design and graphics, though it is incremental as it builds on prior learning-based methods.
The paper tackles the problem of accurately modeling fluid interactions with novel surfaces not seen during training by introducing SURFSUP, a framework that uses signed distance functions for implicit object representation, resulting in more accurate and efficient simulations that generalize to complex real-world scenes.
Modeling the mechanics of fluid in complex scenes is vital to applications in design, graphics, and robotics. Learning-based methods provide fast and differentiable fluid simulators, however most prior work is unable to accurately model how fluids interact with genuinely novel surfaces not seen during training. We introduce SURFSUP, a framework that represents objects implicitly using signed distance functions (SDFs), rather than an explicit representation of meshes or particles. This continuous representation of geometry enables more accurate simulation of fluid-object interactions over long time periods while simultaneously making computation more efficient. Moreover, SURFSUP trained on simple shape primitives generalizes considerably out-of-distribution, even to complex real-world scenes and objects. Finally, we show we can invert our model to design simple objects to manipulate fluid flow.