SoMA: A Real-to-Sim Neural Simulator for Robotic Soft-body Manipulation
Mu Huang, Hui Wang, Kerui Ren, Linning Xu, Yunsong Zhou, Mulin Yu, Bo Dai, Jiangmiao Pang · 2026
Read SoMA: A Real-to-Sim Neural Simulator for Robotic Soft-body Manipulation by Mu Huang, Hui Wang, Kerui Ren, Linning Xu, Yunsong Zhou, Mulin Yu, Bo Dai, Jiangmiao Pang (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2602-02402.txt.
Description
Mu Huang and 7 co-authors present a neural simulator that copies soft-body robot tasks from the real world into simulation, in a 2026 arXiv paper. Cloth and other deformable objects are hard to simulate because their motion depends on both robot actions and environment forces. Older simulators use fixed physics rules or learned dynamics without robot control.
Their system, SoMA, represents scenes as 3D Gaussian splats, small fuzzy blobs that render realistic images. Deformation, environmental forces, and robot joint actions are modeled together in one learned latent space. No predefined physical model is required. SoMA improves resimulation accuracy and generalization on real robot tasks by about twenty percent. It stays stable over long horizons, including complex jobs like folding cloth step by step.
Catalog record
| CREATOR | Mu Huang, Hui Wang, Kerui Ren, Linning Xu, Yunsong Zhou, Mulin Yu, Bo Dai, Jiangmiao Pang |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 46 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2602-02402