[ODE] Re: LCP

[Russ Smith]

Sergey Zabaryansky wrote:
>
> Hello russ,
> 
>   I've just visited your web page and find that you interested in
>   simulation of rigid body physics. I'm interested in this field
>   too. I'm a game developer (I develop 3d actions). I've got a
>   problem with solving LCP when matrix A and vector B is formed with
>   Coulomb friction law. (There is no problems when it is not counted
>   because A is symetric and positive defined and task is easily
>   solved using Dantzig algo). Trinkle suggest to solve such problem
>   with Lemke method but 50% of requests to it fails with second ray
>   result. Have you exeprienced such problem? Maybe i've got a mistake
>   in Lemke algo implementation? Can you help me?
> 
> Sincerely yours,            www.venom3d.com, www.nephilims.com
>  Sergey Zabaryansky         mailto:sergey at nephilims.com

hi,

if you are using acceleration based equations of motion, then you can
indeed find friction problems that have many LCP solutions, or none at
all. some of baraff's papers give a good explanation of this problem.
i believe that when Trinkle suggests using lemke's method, he is
suggesting that it be used on velocity-based time-stepping methods
(because dantzig's method may fail on these problems even though
they have a unique solution).

russ.

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