Regarding Test Data Details?

[Former user]

Dear Beeson

you mention "calling IK implementation for target pose using nominal joint values as seed".

Does the "10,000 random samples" in your validation contains random nominal q_seed ?

What can be difference in results using random q_seed.

[Patrick Beeson]

1) This is not a bug, so it got ignored much longer than if it was a question emailed directly to me.

2) The 10,000 samples are created using Forward Kinematics from randomly sample joint values.

3) "Nominal" seed in the joint values halfway between joint limits (often all zeros)

4) Seeds can highly influence the solutions found, especially on > 6-DOF chains.