Accuracy in cartesian space and units

[publicDaniel]

Hey guys! Currently I'm using your panda-arm in conjunction with ROS-Kinetic and MoveIt! - pathplanning with collision detection and execution works fine.

Inaccuracies in cartesian space

I started testing more with the robot and detected inaccuracies as I moved the robot in cartesian space. For instance, when I tell MoveIt! to move the robot e.g. up in Z-direction, the robot just doesn't move that up respecting the spezified tolerance of . John-HarringtonNZ seems to have a similar issue as he states in #53 "Calibrated DH Parameters".

Units

Some other questions rised as I tested different parameters for the collision thresholds, cartesian- and joint impedances.

You documented very nicely the units and borders of the cartesian impedance parameters.
Coordinate impedances are given in the unit - orientational impedances in the unit . Seems clear to me although different cartesian impedances do not affect the behavior of the robot.
I'm wondering about the unit for the joint impedances. I estimate the joint impedances are - like the cartesian orientational impedances - given in the unit . Is that right? Regarding the borders of the joint impedances - is the maximum value for joint impedances ? Higher values are not accepted by the FCI. Altering the joint impedances do indeed affect the behavior of the robot.

As you state in the documentation the parameters for the cartesian orientation collision thresholds (Roll, Pitch, Yaw) are given in the unit . It's not clear to me, why these values are not given in the unit .
During my testing I recognised that lower values than for the cartesian- and for the joint collision thresholds seem to have no effect on the robot. Are there also borders like there are with the impedance values in cartesian space?

Kind regards,
Daniel