[ODE] how to debug and fix NAN values

Benoit CHAPEROT b.chaperot at jstarlab.com
Tue Oct 30 06:34:31 MST 2007


I have found a bug in my code, and you may have the same bug in your code.
I had one friction coefficient (mu) per Geom. When two geoms collided, I was
averaging their friction coefficients to find the contact's friction
coefficient between the two geoms while creating the contact in the
Some geoms had an infinite friction coefficient, others had a finite
friction coefficient.
Infinite friction coefficient are treated as a special case in the ODE code.
When averaging finite and infinite friction coefficients, I was getting near
infinite friction coefficients, which were then not treated as special case
by ODE, and produced NAN values in the solver every now and then.


On 10/22/07, ode-request at ode.org <ode-request at ode.org> wrote:
> Message: 4
> Date: Mon, 22 Oct 2007 16:57:22 +0800
> From: Hampa <hampa at toribash.com>
> Subject: [ODE] how to debug and fix NAN values
> To: "ODE Mailing List ((E-mail))" <ode at ode.org>
> Message-ID: <34ABE95A-16D6-44CC-B16A-1268E4187DCB at toribash.com>
> Content-Type: text/plain; charset=US-ASCII; delsp=yes; format=flowed
> Hi,
> Every now and then my simulation produces NAN values, usually during
> collisions using hi velocity objects.
> In the debug build (lastets 0.9 svn checkout) I get asserts in:
> ODE INTERNAL ERROR 1: assertion "bNormalizationResult" failed in
> _dNormalize4() [../../include/ode/odemath.h]
> What is the recommended way to debug and fix this kind of problems?
> cheers,
> hampa
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://ode.org/pipermail/ode/attachments/20071030/2b5102c7/attachment.htm

More information about the ODE mailing list