Replacing low res geo RBD Bullet sim with high res geo. Wrong orientation.

[hellbydante]

Does anyone know how to replace a high res geo in a low res RBDBullet sim?

Was following the "Unlocking the Power of RBDs" tutorial by PaulEsteves Unlocking the power of RBDs and I was trying to replace the low res objects in the RBD bullet sim with their high res objects. I've tried to copy the high res objects onto the RBD bullet sim's simulation points via copy to points but it resulted in high res objects appearing with a totally messed up orientation. I've looked into the @orient attribute and quarternion stuff but I can't seem to correct the orientation of the high res objects to match that of their low res copies in the simulation. I've also tried to establish the normal direction of the packed high res geo using an attribute vop. I read that houdini would look to match the @orient attribute first than looking at the @N up of the target primitive. Couldn't find topics about matching the orientation of rbd objects used in this way.

I've attached a hip file for reference.

Would be grateful if anyone knows how to solve it.

 

   

Wrong Orientation.hip

[fencer]

Try to brute-force calculate transformations, your "name" attribute is not correct for direct tranform pieces.

Wrong_Orientation_v02.hip

Edited April 7, 2023 by fencer

[Yader]

Here's a simplified version of your scene, where the lowres simulated objects are replaced by the highres objects. Most of the nodes should be self explanatory, if not - just ask

Wrong Orientation_yader_2.hiplc

 

 

[hellbydante]

Thanks @fencer and @Yader for the valuable help!

From my own mistake, I realise that the RBD solver interprets the randomized orientations as "rest" states and also not suing the correct attribute name. I understand how both solutions corrected the problem by either brute force with extract transform and establishing the rest states early.

Edited April 8, 2023 by hellbydante

[hellbydante]

@Yader After running the sim in the simplified version, the attribute randomize for "orient" seemed to affect the collision geometry causing some primitives to intersect. The orientation of the collision shapes are different than the incoming primitives. I've tried adding the rest node, before the RBDBullet solver but can't seem to solve it. Maybe I'm doing something wrong.

[Yader]

Interesting, here's another setup. The orient comes before copy and packing the geo, then using the proxy and geometry input (transform pieces is already handled by the rbdbulletsolver) the setup seems to work and gives correct results.

See attached scene:

Wrong Orientation_yader_4.hiplc

Edited April 8, 2023 by Yader

[hellbydante]

Thanks again @Yader!