Deformable bodies

[Ishfaz]

Hello
Does anyone here have experience working with deformable bodies in Isaac Sim?
I’m trying to figure out if there is any way to make a deformable object anisotropic (different stiffness in different directions).
If you’ve tried this or know any useful resources, please let me know.

[PeterL-NV]

Isaac Sim’s volumetric deformable bodies are currently limited to isotropic materials (one stiffness value, via Young’s modulus), so there is no direct way to make a FEM soft body stiffer in one spatial direction than another.

The Omniverse Deformable Schema introduces a separate shell material type that adds distinct stiffness values for stretch, shear, and bend, intended for thin-shell objects (cloth-like). These parameters give a kind of anisotropic control over different deformation modes, but they are not general 3D directional stiffnesses and apply only when using the shell material path, not the volumetric FEM soft bodies used for typical deformables in Isaac Sim.

In practice, to approximate anisotropy in Isaac Sim today, typical options are:

• Shape-based tricks: Use geometry and constraints (e.g., elongated shapes plus attachments or partial kinematic control) so that deformations are mechanically restricted in some directions even with isotropic material.

• Different asset types: For cables, garments, or sheets, consider shell- or cloth-type deformables where stretch/shear/bend parameters can be tuned separately, acknowledging this is not full tensor anisotropy.

• Hybrid modeling: Combine rigid bodies, joints, and tendons with deformables, using the rigid structure to enforce directional stiffness while the deformable part provides local compliance.

[Ishfaz]

Thank you for the detailed explanation.
I am currently working with deformable fish models, so unfortunately the available workarounds such as shell materials, cloth parameters, or shape-based mechanical constraints are not suitable for my use case. I need true directional stiffness inside a volumetric body.

I would like to confirm whether it is technically possible to introduce anisotropy by extending the existing implementation—either through a custom Warp kernel or a dedicated Omniverse extension that overrides the material model inside the deformable solver.

If this path is feasible, could you please point me to any example that demonstrates how to inject a custom constitutive model or modify the FEM material behavior? Even guidance on which files, classes, or schemas need to be touched would be extremely helpful. I am mainly trying to understand the correct procedure and integration points.

Thank you very much

[PeterL-NV]

Closing as there's no new response for a long time. Feel free to re-open if you have new thoughts.