[ODE] Rewinding time / backward time stepping
Thomas Harte
thomasharte at lycos.co.uk
Tue Sep 3 13:16:01 2002
This message is in MIME format. Since your mail reader does not understand
this format, some or all of this message may not be legible.
--=_NextPart_Caramail_0279431031084143_ID
Content-Type: text/plain; charset="us-ascii"
Content-Transfer-Encoding: 7bit
I am trying to greatly simplify the mesh on mesh collision problem by being able to
move backward in time. The process will only ever be to have a particular point, to move
forward from it and then possibly discard that forward motion and return to the original
setting. I am wondering about the best way to achieve this.
This is the full definition of dWorldStep from the docs :
"void dWorldStep(dWorldID, dReal stepsize);
Step the world."
So, am I allowed negative step sizes? Assuming I am, does anyone care to speculate on
how much error I would get through stepping forward t units in time, then stepping back
the same amount?
Even if I am allowed negative step sizes, I appreciate that using them is probably a bad
idea when I can simply backup the original state, and then if necessary restore it. Which
leads me to wondering about the best implementation of that. So, some questions :
Is there already a means for backing up a state, then restoring it? An analogue to the
state and matrix stacks in OpenGL, for example.
If there is not, is there a way of polling for every body in a world, allowing me to write
entire world state backup and restore functions as a black box?
Or, at the end of the day, am I looking at finding my own method of keeping track of
every body in a world (which, of course, I almost certainly already have, but this will be
application specific) and iterating through them in my own way? If so them I am likely to
write my own functions to encapsulate all world functions and will almost certainly end up
with unnecessary memory usage as I keep a list of all bodies in parallel to the ODE
internals. Which is the main reason I want to avoid this if possible.
-Thomas
______________________________________________________
Check out all the latest outrageous email attachments on the Outrageous Email Chart! - http://viral.lycos.co.uk
--=_NextPart_Caramail_0279431031084143_ID--