particle collision techniques

[whodini]

I've been exploring collisions between points simulated from a pop network and colliding with geometry. I've tried several techniques and they all give different results and collision accuracy. What's the best way to get accurate collisions between points and geometry without having a very heavy sim?

Some Findings:

My first attempt is a simple collision using the collision pop node to have my points collide with geometry. This produced the best looking collisions without having to do too much. However, the collisions become really slow to cook when the collision geometry is very dense and high res.

Next, I've tried using dop network with the appropriate nodes (pop object, pop solver) to have a pop simulation within dop context and I used the static solver and static object dop nodes to acheive volume based collision. The collisions worked but it's harder to control bounce and friction (points jitter and slide about) and the collision volume needs to be very high res when working with odd shaped geometry.

Lastly, I've tried doing volume based collision using a VDB volume I generated for my geometry and having my static object reference that volume. This worked but VDB thickens the collision area and it's still difficult to get the particles to have correct friction and bounce. So I still prefer using the collision pop method for now.

[rayman]

Another way is using dop bullet solver with rbd point configure object with representation as sphere geometry . This way you have complete control of mass/friction/etc, plus self collision for free.

To source rbd as particles each frame just put $SF in creation frame.