[Sidefx-houdini-list] Building a particle collision system
hoknamahn at gmail.com
Mon Aug 16 04:52:46 EDT 2010
Another reason to use SDFs is their accuracy in terms of particle velocity
against sampling rate. In other words normaly when you use raytracing for
collisions you do oversampling for fast moving particles / collision
geometries just to make sure that nothing goes through each other. While for
volumetric objects it's not a big trouble because collisions performed in
volume, not on the surface so usually there is no need to use oversampling
or sampling rate can be much lower compare to raytracing solution.
On Mon, Aug 16, 2010 at 1:13 AM, Rangi Sutton <rangi.sutton at gmail.com>wrote:
> 2010/8/16 François Duchesneau <sidefx at trinix.ca>
> > I forgot to mention that I want to avoid SDF because even if I agree
> > that in many situations it's a great tool, in others it's just too heavy
> > to calculate to have a good resolution.
> > I started my collision system with this and then I had a shot where I've
> > got lots of rocks in a river separated by a great distance. The sdf had
> > to be huge. Thus I used many smaller SDF for each rock independently
> > merged but the Volume Sample Vop needs to know which primitive.
> Yep. you need to loop through the prims, searching for the volume with the
> lowest SDF value at that point... then collide against that one if it's
> negative. Or.. the first negative one you come across..
> We setup optimisations for this by also passing through a point cloud of
> SDF centroids, so we'd only sample the nearest few SDFs to the particle,
> pulling the prim numbers off the point cloud. When we had >100 SDFs as
> collision sources this sped things up considerably.
> This stuff is pretty easy to set up in vops, with some simple sop nets
> preparing everytthing... gives you a tonne of control.
> That's where I started to think about how fast and reliable Maya can do
> > collision on the fly without pre-baking. It should be like that in
> > Houdini too.
> Yeah, regular houdini particle collisions are slow and buggy compared.
> However, getting to a final shot using every trick available, the houdini
> build still wins out in my books.
> I'd love to see a better straight particle collider working in houdini!
> Sidefx-houdini-list mailing list
> Sidefx-houdini-list at sidefx.com
More information about the Sidefx-houdini-list