Hi all,
I am looking for best source for learning FEMs for soft bodies. I have already looked into the 2 chapters 9-10 in Physics based Animation by Kenney Erleben but it is mostly dealing with Linear FEM (it covers the basics very well). In addition, I was also redirected to chapter 10 of game physics pearls.
I know how to do both linear FEM and corotational FEM (stiffness warping).
What I want to know is if there a recent thesis/survey paper/book/course/reference that lists all of the FEM based deformation method to the current state of the art for soft body modeling.
Best resource for learning FEM
-
Erwin Coumans
- Site Admin
- Posts: 4221
- Joined: Sun Jun 26, 2005 6:43 pm
- Location: California, USA
Re: Best resource for learning FEM
As far as I know, corotational fem (stiffness warping) is the state of the art of real-time FEM, so the two resources seem suitable although they don't cover collision detection properly. The same method is also described in the paper about Pixelux DMM here: http://graphics.cs.berkeley.edu/papers/ ... TD-2009-08
In case someone missed it, I recently posted a link to source code of working FEM under the permissive Zlib license, extracted from the OpenTissue framework here.
The SOFA framework is showing real-time corotational FEM running on the GPU. They can simulate 45k tetrahedra in real-time (90FPS) on the most recent NVIDIA GPUs.
You can checkout svn://scm.gforge.inria.fr/svn/sofa/branches/Sofa-1.0 and see the Sofa-1.0\modules\sofa\gpu\CUDA folder.
See also http://codrt.free.fr/allardj/pub/2011/ACF11a and http://codrt.free.fr/allardj/
The upcoming GPU Computing Gems Jade edition has a chapter on this.
Thanks,
Erwin
In case someone missed it, I recently posted a link to source code of working FEM under the permissive Zlib license, extracted from the OpenTissue framework here.
The SOFA framework is showing real-time corotational FEM running on the GPU. They can simulate 45k tetrahedra in real-time (90FPS) on the most recent NVIDIA GPUs.
You can checkout svn://scm.gforge.inria.fr/svn/sofa/branches/Sofa-1.0 and see the Sofa-1.0\modules\sofa\gpu\CUDA folder.
See also http://codrt.free.fr/allardj/pub/2011/ACF11a and http://codrt.free.fr/allardj/
The upcoming GPU Computing Gems Jade edition has a chapter on this.
Thanks,
Erwin
-
mobeen
- Posts: 122
- Joined: Thu May 05, 2011 11:47 am
Re: Best resource for learning FEM
HI Erwin,
Thanks for the links and info.
Yeah i had a look through SOFA (it gives u a collection of objects/methods that you can play with to understand how the methods perform wthout going into their implementation details) but I dont think it could be used for realtime soft body deformation research because even if u come up with a new method of doing deformation (based on existing modules) however, it is pretty difficult to highlight your novelity unless u extend existing modules to make ur own custom modules but doing that would require a considerable understanding of the whole framework. Ofcourse it is ideal for doing prototyping and comparisons.
In addition, I have compiled the runSofa and other applications fine. However, my own codes for e.g. by following the tutorials donot work I get a lot of errors and assertions. Another thing that is problematic is the unnatural compile time (template instantiation is taking quite some time) for a simple one particle demo and the documentation is also not good just a basic tutorials. These are some things that prevent me from going ahead with SOFA.
Yeah I have read this short talk but I was hoping to get a bit more details on their method which are not covered in the linked paper. I would need to go through the gems book.
Thanks for the links and info.
Yeah i had a look through SOFA (it gives u a collection of objects/methods that you can play with to understand how the methods perform wthout going into their implementation details) but I dont think it could be used for realtime soft body deformation research because even if u come up with a new method of doing deformation (based on existing modules) however, it is pretty difficult to highlight your novelity unless u extend existing modules to make ur own custom modules but doing that would require a considerable understanding of the whole framework. Ofcourse it is ideal for doing prototyping and comparisons.
In addition, I have compiled the runSofa and other applications fine. However, my own codes for e.g. by following the tutorials donot work I get a lot of errors and assertions. Another thing that is problematic is the unnatural compile time (template instantiation is taking quite some time) for a simple one particle demo and the documentation is also not good just a basic tutorials. These are some things that prevent me from going ahead with SOFA.
Yeah I have read this short talk but I was hoping to get a bit more details on their method which are not covered in the linked paper. I would need to go through the gems book.