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

[david@csworkbench.com]

Switching back and forth between integration schemes has been a design
goal in the whole proccess.  That way you can have the more accurate
integrator until you run into some major crash, and iterate through those.
 It would be a fairly simple matter to count contact joints in your
NearCallback, and switch to the iterated solver while that's over a given
threshold.  Transitions are seamless in all the tests I've run.

David

>
>> This is a danger that ODE already has, though.  Geoms can
>> cheerfully pass through each other if they're going fast
>> enough, are thin enough in one dimension, and your collision
>> detection steps are infrequent enough, requiring hacks with
>> rays and ccylinders for bullets, etc.
>
> Another good reason for fixed timesteps incidentally (at least keeps the
> problem predictable).  I vote for cloth - it would be _very_ nice having
> it integrated directly into ODE, and interacting with bodies.
>
> Either way though, way to go on the iteration front David - looking
> forward to giving it a whirl.  Is it possible to switch between the two
> methods mid-simulation for comparison?  Failing that, a compile time
> switch would be nice.
> --
> gl
>
> _______________________________________________
> ODE mailing list
> ODE@q12.org
> http://q12.org/mailman/listinfo/ode