SoftPAT3D: Physics-Augmented Text-to-3D Scene Generation with Deformables
SoftPAT3D: Physics-Augmented Text-to-3D Scene Generation with Deformables
This paper presents SoftPAT3D, a framework for generating physically valid 3D scenes with deformable objects like cloth and ropes. Unlike existing pipelines limited to rigid bodies, SoftPAT3D constructs rest shapes and material models from manufacturing information, optimizes placements via grid search and differentiable simulation adjoints, and verifies equilibrium using geometric checks and vision-language models. The system achieves zero interpenetration across sixty-five prompts and supports scene editing and robotic manipulation.
SoftPAT3D introduces a physics-augmented framework for text-to-3D scene generation that handles deformable objects by deriving rest shapes and material parameters from manufacturing cues rather than generated meshes, achieving intersection-free contact across sixty-five prompts without undesired interpenetration.
Source: arXiv Robotics — research abstracts · Read original article ↗
Loading article text…
What the source reports
Publisher-reported claims, with original evidence. These results have not been independently verified by RoboSignal.
- paper: not_reported
Source excerpts and review record
Automatically extracted; no manual editorial approval recorded.
ibrate in a single simulation, and the pipeline also runs from the text prompt alone. On all sixty-five prompts, SoftPAT3D produces no undesired interpenetration of any kind, whereas every other system does, and its scenes directly support scene editing and robotic manipulation.
Open source S6
Source：arXiv Robotics — research abstracts · arxiv.org