Maya Nucleus

[Oscar Civit Flores]

Hi All,

I've recently found this in Autodesk's website:

http://area.autodesk.com/custom/?id=4547

It's completely based on particles and claims to be able to simulate "any" kind of Model: rigids, cloth, fluids... Sounds like a "Theory Of Everything" for physics based animation.

Just an excerpt:

[quote]
The key concept of the Maya Nucleus unified simulation framework is that the interaction of different elements can be accurately represented by a fairly simple computational model based on interacting particles. The heart of the framework treats the interaction of all elements as a system of particles that collide and exert forces on each other. The complex behaviors of dynamic elements like cloth, hair and water emerge from these simple rules.
[\quote]

What are your impressions on it? Is it just Verlet+Relaxation with tons of makeup (I mean "really well done", because Stam's work in other fields is impressive ) or actually a new paradigm?

And talking about particles+relaxation methods... which are their precise limitations? I've always thought them a "rough approximation" that gets worse (in accuracy/cost) as the number of particles increases (as propagation sequences grow very long, so the number of iterations required to achieve the same accuracy grows too)

Oscar

PS: It's Great that the forums are back

[raigan2]

Jos Stam presented a session titled "A General Dynamical Solver for Computer Graphics" at GAMMA 2006 http://www.montrealgamesummit.com/2006/

A friend who attended said Stam gave an overview+videos, and some interesting details on how solids, deformables, cloth, fluids, cables, etc are simulated.

Sadly I haven't yet been able to find anything online; if anyone here knows Jos, perhaps he could provide some materials?

[bart_k]

Quite an interesting interview. Not terribly technical, but it provided me with some insights.

URL in next post because of first post restrictions.

[bart_k]

http://features.cgsociety.org/story_cus ... ry_id=4026

[raigan2]

also, am I the only one who finds his new particle-based approach very similar to Muller/Jakobsen?

[Oscar Civit Flores]

Interesting link, thanks!

The method sounds very simple... but handling such dense meshes interactively must be a challenge, specially with continuous self-collisions.

I'm curious abount how well particle-based fluids will perform, I've always thought eulerian fluids to be superior for CG purposes.

The "attracting mesh" idea is very useful, by the way... I used something similar in my tiny-little cloth simulation system and it gives you a degree of control over the resulting simulation that would be very hard/expensive to achieve with "real" physics only.

Oscar

[Dirk Gregorius]

What is the "attracting mesh" idea and how and for what did you use it in your cloth system?

[Oscar Civit Flores]

Hi,

The attracting mesh in my system is a skinned mesh where each vertex attracts its corresponding particle with a user-defined coefficient [0,1] which I call "rigidity". In general I use it to control the global shape of a deformable surface/cloth or to avoid large deformations in specific areas.

A good example would be a knight with a cape performing fast combat movements. To avoid the cape ending up blinding the knight and making him look stupid, you would usually want it to remain straight down close to the shoulders/neck near the attachment area and make it increasingly free towards the unattached end. So you model the cape, you skin it to the knight's skeleton, and the you set up the "rigidity" coefficients so that they decrease from shoulders to the opposite border.

Whith some per-particle tunning you can even simulate some deformable surfaces without any inter-particle constraint (although in presence of collisions the surface would move very unrealistically and break)

All in all it's just a hack without any kind of mathematical or moral justification, but it gives you a lot of control

Oscar