[SIGGRAPH 2020] Incremental Potential Contact (IPC) 

I doubt it could compete with marvelous's solver in what its designed to do. (realtime interactive robust cloth simulation). im sure @Michen Li could speak better to that point though

@@RobertFletcherOBE There is no tetra in MD so it's not about replacing the original cloth solver, but add tetra support through IPC

@@workflowinmind I know. What I'm saying is marvellous has a specialised solver which is extremely optimised for its task. Dropping in IPC would not necessarily do a better job

Go ahead and implement it yourself.

I'd love to see this happen. I posted a suggestion here: blender.community/c/rightclickselect/mCgbbc/

Thank you! Compared to traditional MPM, IPC does not have dx gap and sticky artifacts, but if the number of contacts is massive, it can be less efficient than MPM. Likewise IPC currently does not support fracture. Houdini solver is a great idea, we are considering it:)

@@minchenli2292 can it fold paper more than 7 times?

@@bestboy007 :)

I think the difficulty depends, an implementation of an algorithm that runs more efficient will require more efforts, and methods often have a clear framework at a high level with certain details that would require more attention. For IPC it can be directly implemented into the current framework of elasticity simulation (line search based Newton solver), just with an extra energy term besides elasticity potential and a collision detection routine. However, to ensure robustness there are indeed some aspects to pay more attention when implementing. We will release our research prototype code used to generate these examples soon so all details can be easily referenced.

This is a good question! IPC doesn't handle this kind of exactly zero volume or zero distance simulation. It actually simulates these scenes with a tiny gap, e.g. 1mm or even 1um, so that it looks nothing different than zero volume/distance (in this situation we've observed that even ray intersection checks in rendering algorithms will be fooled and the rendering has artifacts). Our other argument for this is that in real-world, nothing like this exists, i.e. even atoms are separate to each other with around 1nm distance, matters that overlap exactly will trigger nuclear reactions:)

@@minchenli2292 what a fantastic answer! ty!

For simulation what we mainly develop is the algorithm to predict the geometry at different times. The rendering is usually done by using software like Houdini or Blender, etc.

Sounds cool! We have our open-source code at github.com/ipc-sim together with a C++ toolkit API, hope that can help.

IPC's source code is publicly available now on github.com/ipc-sim/IPC with a quick start guide on github.com/ipc-sim/IPC/wiki

working on it!

IPC is a contact method and currently we only tested it with elasticity. It would be interesting to see IPC applied on plastic deformation and since it's quite orthogonal it shouldn't be impossible!

working on it~

@@minchenli2292 Yay!

@@minchenli2292 omg

Yeah! We will release the source code soon and we are also working on a Houdini plugin, which will come later.

IPC's source code is publicly available now on github.com/ipc-sim/IPC with a quick start guide on github.com/ipc-sim/IPC/wiki