RBD packed primitive convex hull jitter

[pgrochola 1]

hey,

 

Anyone else notice when using RBD packed objects,  the convex hull generated changes typology thorough out the simulation?

Causing jitter and general strange behaviour.  

This is particularly evident on small objects because it happens at the edges (or when collision padding is high)

 

Check out below file, inside the DOP i have the same object as packed or regular rbd fractured object. 

Apart from the typology changing  on the packed objects, the behaviour of the RBD fractured object looks much more natural to me.

Not sure if this is related to typology change, but there is something odd about the pivot location of the packed object.

 

 

Thanks
Pav

 

packed_vs_standard.hipnc

Edited May 3, 2014 by pgrochola

[goldleaf 107]

Interesting, I see some flickering on the convex geometry too. If no one else responds in the next day or two, might be a good bug to submit to SESI?

 

I did a separate file, and I don't see the drastic topology changes as your hipfile has, but they are definitely quivering.

[pgrochola 1]

thanks for looking.

I noticed the geo flickering also but put it down to GL bug.

If you look at it in wireframe its fine. I also scattered some points on the primitive and they don't seem to flicker around so put that one down to display.

The convex hull flicker gets worse at small scale objects or large collision padding.

[acc3d 2]

I don't think this is a display/OpenGL bug, because the re-topologizing appears to affect the simulation. When the topology changes, the objects jitter because they're trying to resovle new collisions with the ground plane. Also, the re-topologizing is visible in wireframe mode too.

 

 

I posted a file last week that demonstrates the same problem:

http://forums.odforce.net/topic/16254-testing-bullet-why-is-everything-poppingblowing-up/page-2

 

 

-AC

[goldleaf 107]

AC or pawel, you guys should send the hip file as a bug to SESI. Definitely something fishy going on here.

[acc3d 2]

Pawel, I just submitted a bug report to SESI with your example file. We will see what they have to say.

[pgrochola 1]

On my system I see the "display geometry" (original geo from sops) flickering. That's what I think gl problem.

the convex hull collision geo changes typology, which is not a gl issue.

Anyway.. Thanks for submitting bug. If we don't hear from them soon ill submit one also.

[pgrochola 1]

hey,

any info from SESI on this?

[acc3d 2]

About 2 weeks ago, I got an email from SESI that said they're looking into the issue. I've been watching the Houdini Journal and so far there is no fix:

[rayman 456]

Thursday, May 22, 2014

Houdini 13.0.422: Fixed a bug that could cause the Bullet solver to create incorrectly transformed collision shapes for packed primitives with varying scales.

 

Thursday, May 22, 2014

Houdini 13.0.422: Fixed a bug in the Bullet Solver that caused it to re-evaluate collision geometry of RBD Objects every frame if "Inherit Velocity From Point Velocity" was enabled.

 

 

Can you confirm that this bug is resolved?

[acc3d 2]

I just tested, and the bug does NOT appear to be fixed. I will try to follow up with SESI and see what's going on. It may actually be a bug in Bullet, not Houdini...

[kgmcnamara 38]

Yea this bug is definitely not fixed...bummer since I really want to use bullet.

 

Has anyone found a fix?

[kgmcnamara 38]

I poked SideFX again - they have submitted this as bug #62235.

[meeotch 3]

Any progress on / workarounds for this?  I'm still seeing it in 13.0.447.

[Why Gee 0]

I've had a similar problem when "emitting" pack prim rbds where at some point during the emission, the emitter had zero points/prims, meaning at a certain frame it created a pack prim rbd with no geometry which messed up the solver causing this jittering behavior. I created a simple expression to make sure no rbds were generated during those null frames and it fixed it completely. Should be caught by dops but for now this is a good workaround.

So watch out for empty pack prim rbds.

[Skybar 360]

I'm getting jitter as well, using H14.0.258. Here's a gif visualizing the bullet convex hull (had to reduce the colors a bit, hopefully you can still see):

 

Notice that even the red (deactivated) collision meshes are jittering.. 

It's a really simple scene, just a hollow tube cut with Voronoi into 10 pieces and dropped on the ground. Has anyone heard anything more about this? Do you have any workarounds or is it working for you guys? 

 

I guess one could set it to Concave, that works alright, but is just painfully slow.

[Skybar 360]

Interestingly enough, skipping the packed part and just doing a normal bullet sim (with RBD Fractured Object) - the Convex Hulls works just fine.

[acc3d 2]

True, but it's MUCH slower to simulate non-packed. In any case, kgmcnamara submitted this bug to SESI a while ago. Is there any way for us to inquire about it?

[DocWhite 0]

Same problem! But if I change the Geometry representation to Concave it is much more stable. But slower :-(

[Skybar 360]

I emailed with support today (Jenny) and she kindly shared this hipfile with me:

tileDestruction_restingIssue_v003.hip

 

I quote what she said:

"The main problem seems to be that the Bullet engine is producing somewhat jittery results when the collision padding is very small or the objects themselves are small.

If you can use the default collision padding of 0.02 and looks much better.

Attached is a simplified example that shows one approach to scaling the scene - I scaled the objects up before the sim, and then scaled them back down after the DOP Import SOP. I also adjusted the gravity by the same factor."

 

She also said that they have a bug for the drawing of the guide geometry for packed objects (ID# 61536), which probably is why they seem jittery compared to non-packed.

I draw the conclusion from this that the jitter is purely circumstantial and is not related to the visual bug of the packed guide geometry. In the scene above you can see that the convex hulls are still changing, but the simulation itself is stable.

 

Good to have this clarified! Case closed hopefully