TF and RobotDK frame transformation differences
arsh09 opened this issue · comments
Hello,
I am using the GP25 description support package from this repository and I seem to be getting different rotation matrix when looked at the tool0 frame in base frame using TF. Please see the attached file. The frame transformation reported from RoboDK and the one with the TF has considerable differences between them.
I also tried to create my own URDF from CAD model of GP25 arm. However, as there is very less information where the actual frames are defined by Motoman, it is difficult to assign them (talking about the /base frame especially).
I have also tried to use lookup transformation with different euler angles conventions (XYZ, ZYX etc), for some poses, it gives correct results but for other, they are completely off.
Thank you
The URDF model is correct. So are the values in RoboDK. The Quaternion values displayed are also correct, so that would confirm that the model is correct. I've tried various positions in RViz and in my MotoSim environment and the quaternion values are matching. Your issue seems to be in the conversion from the rotation matrix or quaternion to RPY (Euler ZYX). There are various online tools to check rotation conversion, you can confirm that your issue is with the RPY calculations. I'm not sure where that conversion is coded for what you are displaying on the screen.
Hello,
As you can see in the tf_echo, the RPY are different than the ones in RoboDK.
But you are correct, Quaternions seem to match always. I will use them.
Edit: for anyone who might face this issue, the problem was in the precision of quaternions reported from RoboDK. RoboDK interface was rounding the q1...q4 to 5 decimal places which when converted to Euler (or even if copy pasted as is, in a desired/target pose in MoveIt commander yields very different results).