Penalty Method Collision

[raigan2]

Hey,
Just came across this paper: http://www-robotics.usc.edu/~drumwrig/tvcg.pdf

There are some interesting bits such as using PID forces rather than simply PD. It seems weird though that this is presented as "stable" -- without an implicit solver like David Wu uses, isn't it only conditionally stable?

Also, do academics not know about iterative impulse-based methods other than Guendelman?

raigan

[bone]

From the introduction:

We use the word stability within this article to refer
to resistance to perturbation, as in the dynamical systems
sense, rather than numerical stability.

And then they go onto say they can't prove stability blah blah something or another blah.

[raigan2]

i guess i should stop my bad habit of always skipping past the intro and background sections

[bone]

Your point is still valid; anyone who is remotely interested in physics simulation will read the title and assume they mean numerical stability.