[ODE] Coincidental contacts

[Russ Smith]

> Coincidental contacts produce problems for the LCP solver.

the problem is not specific to the LCP solver: if you have coincidental
contacts where the normals also point in the same direction then the
system becomes closer to singular, which is to say that it's harder for
the factorizer to determine a unique solution. you can try increasing
CFM - but a better solution is to not generate the redundant contacts in
the first place.

> if two coincidental contacts involve the same two bodies, is this as
> bad as two coincidental contacts that involve three or four different
> bodies

contacts only occur between pairs of bodies: for two contact to be
coincidental, they must be at the same position between the same pair of
bodies.

russ.

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