[ODE] ODE with non-penetration constraint

Nguyen Binh ngbinh at gmail.com
Wed Jul 11 21:00:59 MST 2007

On 7/11/07, Huang Rubby <rubbybaby at gmail.com> wrote:
> Thanks.
> So such penetration was coming from the "bad" solution of LCP solver,
> which
> will generate a bigger normal contact forces, in my case the normal
> contact force
> will be more than m*g,  to  bounce back such  penetration.
> Please point me out if my understanding was not right.

One thing in force-acceleration model: the normal contact force at the time
step in which collision occured is NEVER be the same as the load (i.e m*g in
your case).
The problem is we want to stop penetration "instantly". In order to do so, a
very large contact force (+infinity) need to be apply to stop the body
instantly. Imagine if the contact force = load, then body's velocity =
constant, means they will penetrate.
That's also a drawback of force-acceleration level model.

Impulse-velocity based model don't need very large impulse to prevent

If you still want to know more about impulse-velocity based simulation, go
here to read a set of lecture notes from Prof Jeff Trinkle. Note that he's
the first (along with Prof David Stewart) to proposed a feasible formulation
that can deal with exact Coloumb friction model. Note that ODE-style and
most game oriented LCP formulations do not follow Coloumb friction exactly.


As for the "soft" contact parameters - erp and cfm, how are they related to
> the
> LCP?
> Ruby

Try to read Russ's article on Gamedev gems. I think you can find it in ODE's
main site. You'll know how erp,cfm work. But note that erp,cfm is just ODE's
specific not a standard in physical simulation.
Binh Nguyen
Computer Science Department
Rensselaer Polytechnic Institute
Troy, NY, 12180
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://ode.org/pipermail/ode/attachments/20070712/c48cc7b4/attachment-0001.htm

More information about the ODE mailing list