Research map / Probabilistic and scientific modeling / Dynamical systems
Learning Particle Dynamics for Manipulating Rigid Bodies, Deformable Objects, and Fluids
Yunzhu Li, Jiajun Wu, Russ Tedrake, Joshua B. Tenenbaum, Antonio Torralba
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add hierarchical modeling (ablation) | DPI-Net · BoxBath | rollout MSE (×10^-2): 2.64 $\pm$ 0.69 -> 2.03 $\pm$ 0.41 | Improved |
| add hierarchical modeling (ablation) | DPI-Net · RiceGrip | rollout MSE (×10^-2): 0.29 $\pm$ 0.13 -> 0.13 $\pm$ 0.07 | Improved |
| add hierarchical modeling (ablation) | DPI-Net · FluidFall | rollout MSE (×10^-2): 0.15 $\pm$ 0.03 -> 0.15 $\pm$ 0.03 | No clear effect |
| add hierarchical modeling (ablation) | DPI-Net · FluidShake | rollout MSE (×10^-2): 1.89 $\pm$ 0.36 -> 1.89 $\pm$ 0.36 | No clear effect |
| use dynamic interaction graphs instead of HRN's graph approach | particle dynamics modeling · FluidShake | rollout MSE (×10^-2): 3.58 $\pm$ 0.77 -> 1.89 $\pm$ 0.36 | Improved |
| use state-specific motion predictors instead of a unified predictor (ablation) | DPI-Net on BoxBath · BoxBath | rollout MSE | Improved |
| use fewer hierarchy roots (ablation) | DPI-Net on RiceGrip · RiceGrip | rollout MSE | Improved |
| increase propagation steps (ablation) | DPI-Net on RiceGrip · RiceGrip | rollout MSE | Mixed |
| increase neighborhood size (ablation) | DPI-Net on RiceGrip · RiceGrip | rollout MSE | Improved |
| increase neighborhood size (ablation) | DPI-Net on RiceGrip · RiceGrip | forward-step runtime: 30ms for d=0.04, 33ms for d=0.08, and 40ms for d=0.12 | Hurt |
| add online system identification | RiceGrip control · RiceGrip | Chamfer distance | Improved |
| use model-based search and trajectory optimization instead of model-free RL | FluidShake control · FluidShake | training time steps: 600K time steps vs more than 10 million time steps | Improved |
Closest papers by meaning
- Learning 3D Particle-based Simulators from RGB-D Videos (ICLR 2024 · similarity 0.90)
- DEL: Discrete Element Learner for Learning 3D Particle Dynamics with Neural Rendering (NeurIPS 2024 · similarity 0.87)
- NeuMA: Neural Material Adaptor for Visual Grounding of Intrinsic Dynamics (NeurIPS 2024 · similarity 0.86)
- Thin-Shell Object Manipulations With Differentiable Physics Simulations (ICLR 2024 · similarity 0.85)
- DeformGen: Dynamics-Based Topology Augmentation for Deformable Manipulation Policy Learning (NeurIPS 2026 · similarity 0.84)
- Latent Intuitive Physics: Learning to Transfer Hidden Physics from A 3D Video (ICLR 2024 · similarity 0.84)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
