RE: [Algorithms] Trouble with translating ragdoll joint rotation to animation joint rotation

[Jon W. <hp...@mi...>]

I'm assuming ODE gives you all the bodies in world space, assuming 
you get both the rotation matrix and the position (offset) and 
stuff them into a 4x4 matrix.

Does your animation system want bone-local matrices, or world-space 
matrices for bones? It's significantly easier if it wants bones; 
in that case, just feed the ODE matrices into the animation system 
(possibly with an offset for global->model space transform).

If you want parent relative matrices, then you have to do a little 
more work, but not a whole lot. Form the ODE 4x4, then multiply 
that with the inverse of the 4x4 for the parent, and set that as 
your bone matrix.

Note that the position part matters! If your skeleton doesn't allow 
bones to move (in addition to artuculate) then you have to separately 
take into account the actual points that the bones end up in, and try 
to run an IK-style solver to make them match the ragdoll -- it might 
be easier to just change the animation system at that point :-)

In general, you can't talk about the "axis" of a joint in an 
animation system, although you CAN talk about an "axis" for 
constraints, and for expressing constraints of that joint. Sometimes, 
you'll optimize to the point that you know a joint is always only 
rotational around axis A, and you only store the rotation angle, but 
if you want to match a dynamics system with the animation system, 
you probably need more expressive systems than that for good results.

Obviously, you also need to match your ODE skeleton as closely as 
feasible to the animation skeleton; you can remove bones by setting 
them to identity in the animation when rag-dolling, but the best 
results if the skeletons are not the same is to do the IK-like thing. 
(That's in fact what we do when tumbling our avatars)

Cheers,

			/ h+


-----Original Message-----
From: gda...@li...
[mailto:gda...@li...]On Behalf Of Megan
Fox
Sent: Friday, February 18, 2005 12:47 PM
To: gda...@li...
Subject: [Algorithms] Trouble with translating ragdoll joint rotation to
animation joint rotation


I'm having trouble with my ragdoll code, but it isn't the physics side
- ODE has that handled famously - the trouble I'm having is with
translating the rotation of the physical joint to the rotation of the
animation joint.  Specifically, getting the coordinate space of the
physical side of things to match the coordinate space used by the
joint in question.

In the physical world, I have my "root" body for a creature, which is
roughly equivalent to the root node in the animation system.  To
create, say, the waist joint, I make the "root" physical
representation resemble the hips, create another physical body to
resemble the chest, and joint the two.  To get the rotation between
the two, I would do chestMat * (hipMat ^ -1), and take the quaternion
of the result matrix.

This works fine for exactly one test matrix - an arm - because it just
so happens that the single joint of the arm has axis which align with
the object-space axis.


The trouble arrives when you look at the coordinate systems for the
hips and chest.  The torso ends up having axis which go xvec =
-objecty, yvec = -objectx, and zvec = -objectz - this is because in
the animation skeleton, there's a good 3 bones between the root and
hips (all of which liking to tweak the coordinate space - it ends up
that positive X always points "up" the bone towards the parent joint),
while in the physical skeleton it's just a straight connection.

I've tried canning it, by working through the pose (default joint
position) quaternions, quat1 * quat2 * ... etc up to the hips,
hard-setting that as an assumed hips rotation, setting the initial
orientation of the chest body to match the pose joint rotation of that
bone, then doing the rest of the joint rotation calc normally from
there... but even that didn't give me the results I was expecting - it
aligned the torso's axis with object-space axis, though the joint
movement past that was seemingly correct.

So, I assume my understanding of all of this is flawed on a basic
level, probably relating to my fairly weak understanding (self-taught
as I've needed it) of the application of matrices.  I could very much
use some help in figuring out where I've gone wrong.


Thanks,
-Megan


-------------------------------------------------------
SF email is sponsored by - The IT Product Guide
Read honest & candid reviews on hundreds of IT Products from real users.
Discover which products truly live up to the hype. Start reading now.
http://ads.osdn.com/?ad_id=6595&alloc_id=14396&op=click
_______________________________________________
GDAlgorithms-list mailing list
GDA...@li...
https://lists.sourceforge.net/lists/listinfo/gdalgorithms-list
Archives:
http://sourceforge.net/mailarchive/forum.php?forum_id=6188