ColladaConverter

[JamesH]

I'm currently implementing Bullet in the C4 Engine to mirror a PhysX implementation that I've been working on, and I've hit a bit of a wall with regards to saving compound rigid bodies, ie made up from numerous convex hulls (generated using ConvexDecomposition)

At the moment I am using ConvexDecomposition on a test model of a column, it works fine decomposing at runtime but obviously it is slow. So I'm saving it out to a .dae file to load back in at runtime rather than decompose it every time. The problem I have is that the collada file that is generated of the saved compound rigid body does not contain multiple convex hulls that were generated, just a single one.

If I then run the same scene in PhysX (both are using the same ConvexDecomp code) and save a dae file out using NXUStream the file is completely different, but if I then load that file into the bullet scene it loads fine and all of the hulls are created correctly.


The image shows a yellow column that has the collision mesh generated at runtime and the red one loaded the file generated from NXUStream, Column-PhysX.dae - the Bullet debug is turned on.

My question is, do I have to do anything special when saving a btDynamicsWorld, I just create the world, create one rigid body (the compound one) and save it. I created my own class that inherits from ColladaConverter and override the ::createConvexHullShape() and ::createCompoundShape() to check that the convex hulls are being created when it loads and just one hull is created when loading the Bullet generated collada file, Column-Bullet.dae.

The two files are here if anyone wants to check them out.

COLLADAFiles.zip

(15.25 KiB) Downloaded 1034 times

Many thanks

James.

[Erwin Coumans]

Cool, that is interesting work, thanks for sharing.

There is probably some issue with the BulletColladaConverter compound export.
In particular, the COMPOUND_SHAPE_PROXYTYPE case in void ColladaConverter::buildShape, in Bullet\Extras\BulletColladaConverter\ColladaConverter.cpp.

I'll see if we can fix it. Obviously, if you fix it before us, we accept contributions.

Please keep us up-to-date and let us know if there are any issues,
Thanks,
Erwin

[JamesH]

I am going to go through it in the next couple of days and see if I can track the issue down. I'll let you know if I find the problem.

[Erwin Coumans]

The latest trunk and Bullet 2.73 RC3 has compound collision shape export in the ColladaConverter:
http://code.google.com/p/bullet/downloa ... 73-rc3.zip

Can you test this (or attached zipfile), and report if that works for you?

Thanks,
Erwin

[JamesH]

I modified the three changes identified in the GoogleCode's change-set (I didn't have time to merge in 2.7.3), and it works fine now thanks.

One observation from the outputted file, is that the hulls are tagged as <convex>, but looking at Chapter 6 of the COLLADA specs for 1.4 and 1.5, the Analytical shapes, should it be <convex_mesh> ?

Anyway, once again thanks for the fix.

[Erwin Coumans]

One observation from the outputted file, is that the hulls are tagged as <convex>, but looking at Chapter 6 of the COLLADA specs for 1.4 and 1.5, the Analytical shapes, should it be <convex_mesh> ?

Bullet doesn't produce the exported XML, it relies on COLLADA_DOM. Are you using Bullet's ColladaConverter, in combination with the COLLADA DOM & libxml from Bullet/Extras folder?
I just tried to convert the convex parts from Bullet/Demos/ConvexDecompositionDemo, and the output is fine:

<convex_mesh convex_hull_of="#RenderMeshBulletUnnamed-0-Geometry"/>

How can you reproduce this <convex> tag? Can you attached a zipped exported file?
Thanks,
Erwin

[JamesH]

Sorry Erwin, ignore my post, I meant <mesh> not <convex>, but I can see that there is a <convex_mesh> tag that refers to the previous <mesh> section.