Re: [Algorithms] Low pass filter on quaternions

[Tom P. <ga...@fa...>]

> The torque can now be defined as log(Qrel) * Cspring, where Cspring is
> the spring constant. You will need to damp out oscillations, by
> multiplying the relative angular velocity (also a 3-vector) by a damping
> constant and subtracting it from the torque.

I love springs as much as anybody, but I've never satisfactorily solved
the "spike" problem, which is to say if the target value is sufficiently
far away from the at-rest current value, the damping doesn't kick in until
you've already blasted off into space.  I don't remember my calculus well
enough to solve this myself, but Game Gems 4 (I think?) has an article on
iterative critically damped springs.  Unfortunately, critically damped
springs also tend to converge very slowly.

-tom!

--