[ros-users] IK with KDL

[Herman Bruyninckx]

On Tue, 9 Nov 2010, Jeroen Willems wrote:

> I have a question regarding the inverse kinematics using KDL.
>
> I would like to use carttojnt of the inverse kinematics solver.
> The problem is I have to use a destination frame as input.
> I'm using something similar to a elbow manipulator so you can imagine I only want a reference position.
> The arm only can have 2 poses to a reference position, to get the rotational part of the frame I actually have to get the joint positions to calculate it (correct me if I'm wrong).
>
> So
> 1. Am I right?
> 2. Could I use KDL to compute the IK or should I use an analytic solution?

The major issue with your problem is that "IK" does not make too much sense
for a 2DOF system: when you don't take precautions the input frame will
never lie in the span of your 2DOF, so you will have make a "projection"
from the full 6D space onto your 2D subspace. KDL has algorithms to do
that, but they make choices about how to project that might not correspond
to your intuition. But that's inevitable.

So, please try to formulate exactly what kind of input you would like or
need to have.

Herman