[ODE] newbie joint question

[Piotr Obrzut]

Hi Jon,

W Twoim liœcie datowanym 2 grudnia 2004 (23:16:40) mo¿na przeczytaæ:


JW> I think these are the reasons nobody answer your question:

JW> Your diagram makes no sense in isolation (what are the
JW> joints? What does it look like initially?)

JW> Your description makes no sense (what is your goal? how do
JW> you set it up? what is the actual movement of the bodies? what is
JW> the expected movement?)

JW> And why do you think this is related to joint stability? If I
JW> can see anything from your diagram, it's that some bodies are at
JW> least a meter away from where you probably want them -- although
JW> you don't specify what you are actually drawing, and how. This is
JW> likely a setup/correctness problem, not a stability problem.
JW> Stability problems are when things explode totally, or when they
JW> never quite settle.

JW> Cheers,

on this diagram you can see Cal3D (it is "character animation library))
model (which name is Cally, it is takenfrom Cal3d sample).
I'm trying to control this model with ODE (for now i turned off totaly
Cal3d control on this model). For this purpose the only thing which
ineterest me from entire Cal3d is its skeleton, which consist of bones
( All bones have transform expresed in its own coordinate system, the
only way to move some particular bone is by this transform). I've made
a dynamic body for this model wchich consist of ODE rigid bodies and
joints. Every rigid body correspond to some particular bone, body's
position is featured by its global transform (related to world
directly). So when i want to get apropriate bone transform i simply
must calculate the transform from body which refer to "child" bone to
body which refer to "parent" bone and it is done. Now i want to
constraint all relative transforms form  "child bodies" to its "parent
bodies". For a test I've just made n-1 (n is a quantity of Cal3d bones
and rigid bodies) ODE ball joints, and I atached to them
"child-parent" rigid bodies pairs. The joints properties was: Anchor
transorm "=" 0; Global CFM = 0.01, ERP = 0.8; When I've added [0,20,0]
force on the second body Cally model (after some time) looked like
this: http://mathrick.org/files/joint_error.JPG (on start model was in
standing position - for some reasons i can't for now place it on some
ftp :( ); I was thinking that is
stability problem b/c as you can see from picture it looks like the
"child-parent" transforms translation length was changed (that
shouldn't hapend when this bodies are conected with ball joint-
it should constraint translations), I was reading a lot about joints
errors so i was thinking that it is b/c of them.

Goal: I want to constraint mentioned "child-parent" relative position
of given rigid bodies; The constraints effects should be as much
predictable as they can (when i'm using a bal or hinge joints i expect
that they wouldn't change its rel. translation length, or the max
error position will be easy to estimate) and the should look cool :).











-- 
greetings,
 Piotr Obrzut