[ODE] Help with LCP internal error

[Bob Dowland]

> In my experience, the most common cause of these insoluble problems is
> conflicting or redundant constraints (either joints or contacts). 

definitely, although, given the scenario - 

>I have 4 bodies, and 5 geometries, distributed among one space and one 
world....

One wonders where these nonsense joints/contacts can be coming from.

In the absence of other types of "joint" bad contacts can appear if
	(discreet) cd is picking up on the intersections too late
	response is not stiff enough to halt/zero penetration velocities
	cubes aren't getting enough contacts to support - "coming to rest" 

So check your contacts (of course) but have a quick shufty at these other things. If you're still stuck after that try four spheres first (an excellent approximation to a cube - in the absence of erhh, cubes..), then try boxes when things are looking a little saner.

:)

Bob



> -----Original Message-----
> From: Jeffrey Smith [mailto:jeffreys at softimage.com]
> Sent: 01 November 2004 15:26
> To: Donny Viszneki; ode at q12.org
> Subject: RE: [ODE] Help with LCP internal error
> 
> 
> Donny Viszneki (smirk at thebuicksix.com) wrote:
> > I get the following error 4 times from ODE:
> > 
> > ODE Message 3: LCP internal error, s <= 0 (s=nan)
> 
> This means, roughly, that the LCP (linear complementarity problem)
> solver has been presented with an insoluble problem, and has failed.
> In my experience, the most common cause of these insoluble problems is
> conflicting or redundant constraints (either joints or contacts).  I
> suggest you manually weed through the contacts generated by ODE, and
> remove duplicate or redundant contacts.  This will not only eliminate
> most of these LCP problems, but also speed up your simulation.
> 
> Jeff
> 
> _______________________________________________
> ODE mailing list
> ODE at q12.org
> http://q12.org/mailman/listinfo/ode
>