[ODE] Rigid Body Dynamics Dissertation

[Adam D. Moss]

Ignacio García Fernández wrote:
 >>That's interesting.  I'm also using StepFast and I can rarely stack
 >>15 regular cubes squarely atop each other without them slowly oozing
 >>into wonkiness and falling over within about ten seconds.  Could
 >>you share what parameters you used for stacking hundreds of boxes
 >>(friction, etc), what mass they had, length of timestep, number
 >>of iterations, etc?
 >
 > Hi, I've seen some papers published in journals such as Phis. Rev. E
 > talking about granular systems, such as sand, that model the system
 > as hundreds of small irregular objects. The problem of achieving a
 > stable behavior is very hard. The most common solution is to stop the
 > integration of the positions of all the objects that are resting. Then
 > the problem becomes in deciding which are resting :-D

Well, I understand.  :)  I can certainly make my boxstacks
be unsimulated right from the start, until they get bumped into.

I do use a hand-rolled 'auto-sleeping', but it doesn't kick
in on the boxstacks because they start wibbling and falling over
right from the start.

On the other hand, having the stacks asleep by default keeps them
stable of course (the speed and accuracy of the simulation step become
practically irrelevant, since they're simply not simulated), *but*
then it's possible to nudge blocks out from the bottom of the pile
without waking the upper ones, leaving them floating unsimulated in
mid-air -- unless the sleeping blocks are set up to initially be
slightly interpenetrating and are re-awakened on the slightest awake-
body contact (whereapon my factory building falls down if licked by
a mouse ;)).

It's all really an array of hacks to achieve the brief illusion of
stability in a regular boxstack just several blocks high.  Hmm.  It
just occurred to me that a horizontal surface (i.e. a box's bottom
face) resting on another horizontal surface (plane or another
horizontal box face) in the presence of vertical gravity is a
(uniquely?) special but commonly-occuring case with an unusually
high expectation of stability from the user.  Maybe there's room for
some special-casing stability tricks for this occurance in ODE itself.

--Adam
-- 
Adam D. Moss   . ,,^^   adam at gimp.org   http://www.foxbox.org/   co:3
Consume Less, Live More