H10.0.260

I am not able to get rbd/cloth mutual collisions to work, only cloth is being deformed by RBD. I tried to lower RBD's density/mass to extremely low numbers but with no difference. Am I missing something obvious?
see attached file

I am getting error in cloth solver:
“Cannot retrieve data from previous solve”
it happens a lot
even in ParticleAndCloth Example file from help examples
EDIT: the attached file is only for the collisions issue, i am not getting this error in this particular file. But see above mentioned example file for this error

RFE: I have noticed that cloth collides with its faces (maybe edges/points too) this is very cool, but is it possible to give user control over which collisions will be used?
Like points/edges/face checkbox near the Cloth-External Collisions
and the same near the Cloth-Cloth Collisions
This is only if we can benefit in speed increase when face/edge collisions are turned off for initial tests for example.


thanks for any suggestions

tamte
I am getting error in cloth solver:
“Cannot retrieve data from previous solve” it happens a lot

I get this a lot too. :x

Cloth and RBDs do not pass forces back and forth between each other unfortunately. So what's happening is that the RBD sim is just ignoring the cloth that's there and having the ball keep falling, so the cloth eventually gets stretched so much that the solver freaks out. (I'm doing this on build 245 btw, it has changed a bit since then but that basic problem is the same).

One way to do this would be to make the sphere a very stiff cloth object, and drop that on there. If you didn't want to see the deformations that would technically result on the sphere you can make another sphere geo follow around that centroid.

It's good that you have the sphere as a polygonal one, as primitives or poly meshes won't work with cloth.

I'm not 100% clued up on the collision model, but I really doubt you can be that selective about how to calc collisions, mainly because the model needs to take into account some fairly crazy eventualities. IE a giant poly with a sphere of tiny ones hitting it, or an object that cuts a piece of cloth very fast with strange topology, etc. The cloth solver actually triangulates everything (I think) and then bases collision off that, with non planar polys that have N sides things get interesting.

Turning off bits and pieces of that algorithm would cause things to get very unstable I think. The Symmetrize max sides parm on the cloth solver is actually the input for that triangulation method, I'd need to refresh my memory on it before I offer any more advice there.

edit: as for the “cannot retrieve data from previous solve thing”, it will actually give you the error as a node message on the frame it dies, but since the next frame will look at THAT frame and go “oh no I can't retrieve stuff” it wipes the message. It can be hard to know what's going wrong unfortunately, it would definitely be good to get more feedback with it.

thank you for your long answer msawtell

i am confused about cloth/RBD interaction, i thought that it is supposed to work together and it worked (more or less) in H9.5 so this have to be a bug then.

To the collision toggles, I was used to those type of choices (point/edge/face) in some other cloth systems and it is good to be able to turn off face and edge collisions if i do not need them to speed up simulation (I know that it will lower the accuracy for the reasons you described). But if it is not possible with houdini's cloth model then it is still ok with me. I have kind of used to accuracy-over-speed in houdini, however i am still hoping for both in the future.

the error message now makes more sense
but raises the question, why did my simulation stop?
i hope it will be resolved soon

Nevertheless I have to say that there is very notable improvement on H10 cloth, great work deveopers

In Houdini 10.0, collisions between cloth and non-cloth are no longer using SDFs. Instead the cloth collides with polygons in the non-cloth geometry. In particular, cloth collides with the points, edges, and facets of the polygons in the non-cloth geometry. This has several advantages. For example, collisions with animated geometries (deformable Static Objects) now work better. Also, it is no longer possible for an small Static/RBD object to move completely through the cloth in one frame.

Support for feedback from RBD/Static to cloth will need to be re-added for this new collision code path. So currently, there is only a one-way interaction: RBD and Static affect cloth, but not the other way around.

As noted by Martin, you can work around this issue in many cases by constructing “quasi-rigid objects” from cloth. Instead of creating an RBD/Static object, you would create a cloth object with high stiffness coefficients for “Stretch, Shear and Bend” (see the cloth material tab). It is recommended to use the Strong bend model in this case.

Michiel

michiel
Support for feedback from RBD/Static to cloth will need to be re-added for this new collision code path. So currently, there is only a one-way interaction: RBD and Static affect cloth, but not the other way around.

Is there a timeframe for that? The fact that this type of interaction is missing (while it was available before) is kinda worrying. I'm impressed by the advance of cloth in H10, but I also hope that advertised features were taken out only for a short time.

Dragos