Enable alternative material models for triangle/tetrahedra mesh

[ShaoxiongYao]

Hi,

Thank you for providing this library!

I want to ask if there is a plan to enable alternative material models in integrator_euler.py. The current tetrahedra mesh assumes the Neo-Hookean model and alternative models have been implemented but commented out. For example, in eval_triangles there is alternative St. Venant-Kirchoff model implemented. I wonder if there will be flags to enable different material models in later iterations.

[mmacklin]

Hi @ShaoxiongYao ,

I think having the material models on a flag is a good idea - it hasn't been a high priority, but it should be fairly straightforward to implement.

Generally I think Neo-Hookean should be preferred over St. Venant Kirchoff since it is has better volume preservation. Are there any other models you would like to see?

Cheers, Miles

[ShaoxiongYao]

Hi @mmacklin,

Thank you for the reply! I do agree St. Venant Kirchoff's model is generally less preferred than Neo-Hookean (element inversion problem). But I think it would be a nice feature for users playing around with different options. If possible, I would like to see linear/co-rotated linear models that are more efficient, I am also happy to contribute if you are busy :)

Yours, Shaoxiong

[touste]

Hi, as a user interested in using warp for biomechanical applications, I strongly support the idea of separating the constitutive model from the integrator to ensure easier development of other hyperelastic materials (e.g. Mooney-Rivlin, Ogden, ...). Typically, you could have the constitutive model in a separate function that takes F and dFdt as input and expect PK1 as output. I'd be happy to contribute then. Thanks!

[shi-eric]

Sorry we weren't able to address this issue before removing warp.sim from the codebase. As of Warp v1.10, the warp.sim module has been removed. It was deprecated back in v1.8 and has been superseded by the Newton physics engine, which is now maintained as an independent Linux Foundation project with a redesigned API focused on robotics and robot learning.

If this issue is still relevant to your work, we'd encourage you to check out Newton. There's a migration guide available that should help with the transition. You can open issues in the Newton repository or ask questions in the Newton Discussions section. Many of us from the Warp team are also involved with Newton and happy to help there.