[Papervision3D] Collada Biped

[Jon Bradley]

On Feb 23, 2007, at 1:15 AM, jakelewis wrote:

> Hi Jon,
> In the code in the link you posted they are in fact performing the  
> step I suggested:
The point was that it's not necessary to calculate how far away from  
the bone you are, or do any additional calculations. I think I  
mentioned that this was the 'older' way of doing it, where you worry  
about where the vertex is in relation to the bone. You really  
shouldn't have to worry about that. By transforming the vertex  
directly by the matrix, it should be right, assuming your matrix is  
in the proper space. It also allows you to shove other  
transformations into the stack from other places easier - say a  
cluster or other deformation node.

> Quarternions. I guess this depends on the how the animation data is  
> exported within Collada. I imagine it can support multiple types.  
> Considering PV3D is many flash users first exposure to 3D, it does  
> seem a little harsh to up the ante to 4 dimensions, albeit 3 of  
> them imaginary.

Yea, totally agree. Quaternions are just easier to work with and much  
more powerful for gaming work.  They make short work out of things  
like aim and up vector constraints and motion on a sphere. For  
something like, say, a random rotation like a spinning asteriod, it's  
much more difficult to do that by quats. Speaking of which, I have a  
pretty decent Lua class I can hand off to everyone here if we want to  
have some nice random rotations on asteroids in a space game. :)

Quats are difficult to wrap your head around, but no more so than  
matrices, in my opinion. On import of the Collada data, you could  
quite faithfully convert the matrices into quats.

This is going to be some cool stuff...

cheers,

jon