SV: [ODE] Iterative solution

david@csworkbench.com david at csworkbench.com
Wed Mar 19 16:24:01 2003


For that matter, how would you know that the bottom box had moved into the
box above it, short of running collision detection every iteration, which
is not worth it unless you're actually using the iterations to improve the
accuracy of the system at the cost of speed (instead of vice-versa).  If
that's the case, you might as well just run ODE as is, but with a smaller
timestep....

On the other hand, I'm not against position updating.  Practically, it's
the best solution I've tried so far.  However, I haven't tried only
updating the velocities yet (as I'm still trying to get the updating
forces to work).  I'm just trying to experiment with as many schemes as I
can work out in my head, so I can pick the one that works best
practically. So far, none of the solutions I've tried have passed both the
box wall crash test and the test_boxstack test very well.  They're either
very jittery in test_boxstack or quite (unbearably) slow in test_crash.

David

>
>> If you don't have position update, what happens when a stack of boxes
>> first penetrates the ground?
>
> riiiiight ... i have been only thinking about systems that have the
> "correct it later" approach to interpenetration (i.e. ODE's current
> approach). i've not yet begun to think about the no-penetration case (so
> we're not really comparing apples to apples here!).
>
> russ.
>
> --
> Russ Smith
> http://www.q12.org/
>
> _______________________________________________
> ODE mailing list
> ODE@q12.org
> http://q12.org/mailman/listinfo/ode