Rbd Chain - On Character Neck

[stass3d]

hello all!!

I`ve a problem.

I trying to make a rig from a character with a chain on the neck.

Character body is deformed from skin and baked in OBJ sequence.

The chain has a big rings, for this reason i try to make a RBD dynamic.

I made a volume collision , with relative much divisions.

Chain is Fractured RBD Object

now, i have a problem, thet if body start to be deformed, chain penetrate throught the first ring,(constrained to the neck) and flyes out.

I try to make a sortly time step, now like a 0.005 and minimum substeps like a 8 ,

The simulation become very slowly - 1 frame per minute, but i still have this problem.

Which direction must i go, in oder to solve this problem?

[SpencerL]

In your RBD Fractured Obj make sure under the Collisions tab that Laser scan is unchecked so it will take into acount the "hole" in the chain link.

"In laser scan mode the volumetric representation is built by sending rays along the primary axes. Only the closest and farthest intersections are used. The space between these two points is classified as inside, and the rest outside. "

[mtucker]

You can see the volume used for collision detection by the RBD system by turning on the "Show Collision Guide Geometry" option on the Collisions page of your RBD objects. That will at least let you see if the volume is a good representation of your geometry. If it's not, you can try adjusting the other parameter on the Collisions page. But for rings, like SpencerL said, you'll definitely want to turn off Laser Scan.

Mark

[michael]

oh lord....don't let ben see this....

it's a tough problem to do this right now....

[bensimons]

Oh gee... There's no surf this morning, so i thought i'd check-out the odforce forum: first time in ages. Pffft. Why oh why?

It's a hard problem in DOP's right now. Trying to just throw all the geometry at one solver is not going to get you very far. I think the way forward is to "divide and conquer" the problem. There's two sorts of collisions going on: 1. Body & Chain, 2. Chain Links themselves (inter-chain). With a greatly simplified chain, if you solve the path for 1, then you should be able to solve 2 with spring constraints (links constrained to path by springs, and rbd for link to link).

Try a greatly simplified chain (think "connected tubes") for problem 1. Alternately, try a Wire Deformer(!). Set the width of the wire to the width of your chain link. Set weights to the points on your wire. It gets you close to a solution. Wire/Volume collisions are fixed now (8.0.433).

Be very careful to make all your values real world values! Check your units (mass, size) are sane. This is important. There's values which get squared (acceleration) and so you can't just scale stuff afterwards to fix things.

In the RBD Solver, The "Shock Propagation" value is important for a chain. Set it to a value greater than the largest number of free-hanging links in your chain.

Recently (~ 8.0.440) the minimum substeps setting on the RBD Solver was a little unreliable. try changing (make smaller) the Global Timestep parameter up at the DopNet->Simulation-Tab level. This is the 1/$FPS value. This is going to slow everything(!), but might get a better result. Just something to try.

The "Resolve Penetration" setting on the RBD Solver is what you'd hope would fix links which initially intersect. The trouble is it's not too smart about how to do this with a chain: fixing one link can cause a break further down the chain. It just fixes ("resolves") one link at a time. So, it's a good idea to check your chain is ok before you start. That's probably what's causing your link to "fly out".

A good way to start is consider a huge "pre roll". Set up your chain, all "nice", and sitting well above your character. Create a long pre-roll sequence where the chain free falls onto your character. Then ramp into your character animation, so that the intial forces are small. Ease it all into the point where frame-1 should be.

Make sure all your cooked geometry is being blended at the sub-frame level so there's no hard-jumps (causing huge accelerations). I use two FILE SOP's (reading $F and $F+1), with a BLENDSHAPES SOP so that you get nice-smooth motion for all sub-frames. You can use the same idea to get a nice-smooth pre-roll.

wow. I've probably said too much, but I think it's a good problem to hammer on. I started to make progress once I stopped thinking "this should just work", and began looking at "what works right now".

Oooh Surf's up! :-)

ben.