PDA

View Full Version : Dynamics - Groups and CTRL Key?



wmcole
02-16-2008, 10:35 AM
I think I need to set the crash test dummy (hieracrhical parts rigged for IK w/o bones) as the collision controller. It will collide with various portions of the interior of a vehicle (also a hierarchical object, no IK). I first tried to set the seat belt of a 3-point harness (a ***** to model, BTW) with Cloth Dynamics and set the non-fixed connection point surface for collision detection. Then I started at the top of the dummy heirarchy setting each of its layers to with FX Collision in the object layer's Properties Dynamics tab.

In setting up the 32 layers of the dummy to be the collision controller, I noticed the <new group> option which is not well documented.

1. How do I create a group, edit a group, delete a group etc.? Could I, for example, put all 32 layers of the dummy into a group and somehow just set the entire group as a collision controller with one set of parameters (instead of having to copy and past the settings 32 times over)?

2. In trying a test calculation for this 600 frame animation, it became obvious after 80 frames that the seat belt was flying right through the dummy and distorting all over the place for no apparent reason. The little status panel says you can stop the calculation with the ctrl key. However, pressing it appeared to have no effect other than to cause the program to hang after a few more minutes. How do you abort a dynamics calculation in progress?

Any other suggestions on where to look for an appropriate tutorial or discussion on how to set up this animation for dummy collision with the seat-belt (and its cloth deformation -i.e. limited stretch and shape change due to side-to-side or up/down dummy motion) and collision with other parts of the interior (headliner, seat, steering wheel etc...) would be greatly appreciated.

Dodgy
02-16-2008, 03:40 PM
Okay, here's how I would do it.

Make the dummy all one object, spreading it's parts out a bit, and then adding a bone to each part, so each bone only influences that part, and on frame 1, pulling the dummy back together using the bones. That way you can add CollFX to just the one object, and not have to deal with every part separate.

Make a low poly proxy for the seat belt, which would consist of just a strip of quad polys forming the rough shape of the belt. This makes calcs a lot quicker and more robust usually. You can then make the high poly belt its child and use Metalink.

You may want to use a poly reduction tool on the dummy itself to get a low poly proxy for collision, to speed things up.

ctrl can be a bit dodgy, especially if you aren't using the FX browser to run the calc. Use that, as it handles switching tasks around much better than just pressing calc.

Dodgy
02-16-2008, 03:48 PM
See here for more details:
http://www.lightwiki.com/Dynamics

wmcole
02-18-2008, 12:34 PM
Dodgy,

Thanks for the quick feedback.

I went for a "quick and dirty" workaround: I scrapped dynamics, keyframed dummy movement (always going to be the approach to show a particular facet of actual vehicle testing) and used bones to pin the seatbelt attach points and then deform the seatbelt in accordance with dummy motion. I made use of some shortcuts by keframing the dummy, then saving its hip motion file to import as the deforming bones motion file. Then it was a slight adjustement here and there to keep the belt polys from falling inside the dummy.

HOWEVER - I want to rebuild this thing with dynamics to be sure I understand how to do it as that would have "expanded business" implications down the road. So I am going to have to ponder your suggestions to get a full understanding (gotta do some reading on Metalink) before trying it out.

Thanks again for the help.

stib
02-19-2008, 04:51 AM
Definitely don't use real models for dynamic collision objects. Apart from making the calculations absolutely glacial, they tend to make them go wrong. So make some low poly proxies for all your dynamic object and meta-link them in.

Doing everything with dynamics seems like a good idea at first, but only until you realise that they don't really work.