search

ros-naoqi / nao_robot

BSD 3-Clause "New" or "Revised" License
53 stars 60 forks source link

r_wrist frame typo #25

Closed letrend closed 8 years ago

letrend commented 8 years ago

There is obviously a typo in the nao_description/urdf/naoVXX_generated_urdf/nao_robot.xacro files in versions XX=32/33/40/50, causing the coordinate frame of r_gripper to have an offset in rotation.

Current length:

<child link="l_gripper"/>
<origin rpy="0 0 0" xyz="0.05775 0 -0.01213"/>

Actual length:

<child link="l_gripper"/>
<origin rpy="0 0 0" xyz="0.05775 0 -0.01231"/>

Note the -0.01231 The dimensions can be checked on Aldebarans page

vrabaud commented 8 years ago

@suryaambrose, @lsouchet, please confirm / infirm.

mikaelarguedas commented 8 years ago

Please note that the V32 values are fine but 33/40/50 have the same typo.

sbarthelemy commented 8 years ago

I confirm the typo, I found the same inside aldebaran's library from which the urdf files are generated. Now that I fixed it there, I suppose generating the urdf files again should propagate the fix to ros.

vrabaud commented 8 years ago

@sbarthelemy thx. Can you please send me the newly generated URDFs so that I fix that bug ? Thx. Otherwise, you can make a PR yourself using those instructions: https://github.com/ros-naoqi/naoqi_bridge/blob/master/naoqi_tools/scripts/README_generate_urdf.py.rst

vrabaud commented 8 years ago

btw, @letrend , how did you find that bug ? did it impact any of your experiments ?

letrend commented 8 years ago

@vrabaud jup, made transformations from gripper to torso frame impossible. at first this seemed to solve our issues, but we noticed that the coordinate frame of the right gripper was still weiredly twisted. So there might still be something wrong about the coordinate frames. we couldnt figure out what, however. eventually we used visual tracking of the endeffectors to compensate for this.