[ODE] Re: Inside ODE

[Nate W]

On Tue, 12 Mar 2002, [iso-8859-1] François Beaune wrote:

> Hmmm, on second thoughts, I suppose it should not be a problem
> since in most cases relaxation (error reduction) is involved because
> of penetrating (thus moving) bodies. However, if you create a stable
> assembly of bodies (a wall made of bricks) which interpenetrate
> slightly each other and then run the simulation, the wall may crash,
> am I wrong ?

If you create interpenetrating objects, they behave as if they were made
of rubber and pressed together tightly.  When the simulation begins, they
just bounce apart, but it does not necessarily cause a simulation crash
(you don't get division by zero or not_a_number values).  With a bit of
trial and error, I found global CFM values that worked well for what I'm
doing, and I haven't seen a simulation explosion in a long time, even when
deliberately creating interpenetrating systems.

If this system has a flaw, it may be that it is too easy to make it
tolerate errors.  It's easy to make simulations where all of the joints
are slightly elastic.  They're very resistant to explosion, but if you
look closely sometimes the bodies don't line up perfectly with the
joints.  On the other hand, by changing the CFM you can make everything
line up perfectly but with increased risk of binding or explosion.

What kinds of systems are you intersted in simulating, if I may ask?

-- 

Nate Waddoups
Redmond WA USA
http://www.natew.com