[ODE] Re: LCP

[Russ Smith]

Sergey Zabaryansky (sergey at nephilims.com) wrote:
> I need to model multiply collision bodies. I think ananlitical model
> is the best for this, isn't it?

i don't understand what you mean.

> Both Baraff and Trinkle suggests to use Lemke method for simulation
> static & dynamic friction. But they also say that it may fail with
> second ray result. It seems like contradiction. They propose to use
> not reliable scheme, don't they? Have you had a deal with modeling
> static/dynamic friction in multibody system?

the lemke method will reliably solve a wider class of LCP problems than
the dantzig scheme (or other some other principal pivoting schemes).
so simulation-stability wise it's the better choice. but it's a bit
harder to optimize for high performance, which is why i went with the
dantzig method in ODE. lemke is more likely to work on a frictional
problem.

russ.

--
Russell Smith
http://www.q12.org