Quaternions are much nicer for keyframed animations, no question.
But for IK solver Euler angles seem to be much more natural:
- They allow me specify constraints easily (as maximum/minimum rotation angles). Converting constraints to quaternion space seems quite clumsy...
- Joints with only 1 DOF (knees, elbows) have only 1 adjustable component in Euler angle based rotation, but 3(4) in quaternion, meaning that solver will probably be slower and noisier.
Or is it pseudoproblem, as (hopefully) most IK adjustments to legs and hands are small enough, so the lack of constraints would not be visible?
Any help would be welcome!