[ODE] penalty method

[Pierre Terdiman]

> The stiffness of the differential equations and the fact that the penalty
> forces rely solely on the position of the objects (not the velocities)
> demand that you use an integrator that does not amplify its own errors.

Who said the penalty forces didn't depend on velocities ? Mine do.

> Wu
> uses implicit Euler as integrator, which he claims keeps the system from
> exploding. Also, care should be taken not to add energy to the system by
> placing exerted springs on interpenetrating contacts, so either you have
to
> determine the exact time of collision, or you'll have to allow that
objects
> are interpenetrating to some extent.

Indeed, some interpenetration is allowed. That's good or bad, depending on
the way you see it. On the bad side, well, it might be visually unpleasant
if it's too obvious. On the good side, the system doesn't break or explode
when it happens. And it does happen.

Anyway I'm not sure the magic physics bullet has been found yet, each method
has its share of painful problems..

Pierre