14k particles
800 steps
ground truth simulator: SPH
19k particles
400 steps
ground truth simulator: MPM
15k particles
300 steps
ground truth simulator: MPM
6k particles
800 steps
ground truth simulator: SPH
2k particles
1000 steps
ground truth simulator: MPM
2k particles
320 steps
ground truth simulator: MPM
2k particles
400 steps
ground truth simulator: MPM
2k particles
1000 steps
ground truth simulator: MPM
8k particles
1000 steps
ground truth simulator: MPM
3.5k particles
600 steps
ground truth simulator: MPM
3.5k particles
400 steps
ground truth simulator: MPM
2.5k particles
600 steps
ground truth simulator: MPM
1.4k particles
2000 steps
ground truth simulator: MPM
1k particles
150 steps
ground truth simulator: PBD
2k particles
1000 steps
ground truth simulator: MPM
5k particles
400 steps
ground truth simulator: MPM
trained on: friction angle range [0, 30], [55-80]
inference: friction angle range [0, 90]
trained on WaterRamps (2.5k particles, 600 steps)
inference: 2x2 domain, 28k particles, 2500 steps
trained on WaterRamps (2.5k particles, 600 steps)
inference: 2x2 domain, 5k particles, 2000 steps
trained on WaterRamps (2.5k particles, 600 steps)
inference: 8x4 domain, 85k particles, 5000 steps
trained on SandRamps (3.5k particles, 400 steps)
inference: 1x2 domain, 3.5k particles, 2000 steps
trained on MultiMaterial (2k particles, 1000 steps)
inference: 1x1 domain, 4.5k particles, 2000 steps
BoxBath domain
While DPI uses hard-coded constraints to keep the box shape consistent, our model achieves this without any special treatment of the solid particles.
Comparison in the following domains:
Water-3D-S (SPH)
BoxBath (PBD)
Sand, Water, Goop, MultiMaterial (MPM)
Above videos are indicative for our model's average performance. However, in our comprehensive experiments we have also found some interesting examples of failure cases:
Over very long rollouts, solids may become deformed
Some model seeds learn to predict large pieces of goop sticking to the wall instead of sliding down
See supplementary material for additional discussion.