[ODE] Particle/cloth/skeleton (was: Re: Choleski factorization)

Joakim Eriksson jme at snowcode.com
Mon Mar 24 06:33:01 2003


> >However what goes againts using a verlet approch is that I think it
> >would be very hard to integrate the two systems.
> Verlet is just an integrator, you can do extacly the same 
> things by using an Euler integrator, you need just to project
> the particles velocities on a valid manifold.
> to be porecise only the particles velocity component along 
> the edge should be preserved.
> However by using verlet is faster as we work only with 
> positions. Xavier Provot did not used a verlet integrator.

That was a bit bad formulated on my part. I'm well aware that
it's just another integrator. What I was aiming at was that its
two diffrent systems to handle the same problem. 
If you solve the collision constraints using one system
(LCP constraint solver in this case) and the cloth's constraints
using another system (Verlet constraints in this case) you will
have problems because one system doesnt known about the constraints
in the other.
I explained this a bit better in a previous mail I sent to the list.

I have built this kind of system (Cloth using verlet constraints and
another constraint system for the collisions) and I havent been able
to solve the problem with objects that can pass through the cloth.
The quick solution have been to simply increase the collision mass
of the cloth so you never let the rigid body go to far into it.
It has worked so far but it would be nice to be able to garantee
that it can never happen.

/Joakim E. - http://www.planestate.net / http://www.snowcode.com