[ODE] Re: center of mass

Nagymathe Denes denes at invictus.hu
Sat Nov 5 00:45:46 MST 2005



 Hi,

> Lastly, I never really looked into geomTransformgroup, but what was it 
> supposed to do, host multiple geoms that have the same offset transform? 
> Perhaps it has been eliminated because it doesn't offer anything above the 
> normal geomtransform?

 It hosted multiple geoms with _different_ transforms; just the thing we
 usually do with multiple geomtransforms, one for each transformed geom. So
 in the usual case where someone attaches n geoms (embedded in n
 geomtransforms) to 1 body, 1 geomtransformgroup would be enough instead.

 But the greatest idea of it was to make itself attachable to other
 _geomtransformgroups_; this way we could create hierarchical geom trees
 built on bodies!
 This, possibly with some further simple extensions has many advantages, 
from
 speeding up collision queries (the geomtransformgroup itself can have its
 own AABB or even an OBB that can be checked just like a usual box, 
filtering
 out the embedded geoms better than a sub-space) to animated objects (a 
whole
 subtree of geoms can easily be transformed by changing only the rel. pos 
and
 R of their parent geomtransformgroup within its parent ~), etc.

 If anyone would like to use more sophisticated, or even concave collision
 shapes than the built-in primitives (or just a few of those),
 geomtransformgroups could be _very_ useful. (We could also build these geom
 hierarchies using subspaces as nodes and that could filter out some geoms
 too, but it's different.)

 btw, the Bullet library also have a potential, but using it with trimeshes
 doesn't seem too simple yet. :o(

 Denes





More information about the ODE mailing list