[ODE] quickStep doesn't work for me.

Eric Froemling eric at ricore.net
Wed May 19 08:47:02 MST 2004


On May 18, 2004, at 10:34 PM, Russ Smith wrote:

>
> to improve the situation with motored joints i have improved the
> quickstep method to use warm starting - which means that the previous
> solution that was found for the constraint forces will be used as the
> starting point for the next solution. on my ant simulation this helps
> the situation a LOT - now only 20 iterations are needed for reasonably
> good stability. this increases the size of the joint structure by
> dReal[6] though.

Noting this, is it possible to perfectly restore the state of the 
solver at a given point with quickstep?  The method i've been using 
with plain old dWorldStep() is to create all bodies and constraints in 
their starting positions and then setting their positions and 
velocities to stored values for the target time.  Would this new bit of 
data need to be saved/restored as well to keep the sim arriving at the 
same results?

-Eric



More information about the ODE mailing list