mismatch between pr2 pose in rviz vs reality

[Roxana Leontie]

Hi all, 

I am working on a project with our PR2 and I noticed that the robot constantly has the arms a bit misaligned after the startup calibration. Also when testing we constantly see that the robot thinks it is in a pose that is not in. For example in the attached picture the robot thinks the end effector is at a location, but the point cloud overlay shows it is at a different location. 

Is there a way to test the joint calibration or improve it? Any idea what might be wrong if not this?

Thank you,

Roxana 

PR2 with ros groovy. 

[Lasse Einig]

Dear Roxana,

your arm calibration is probably off. Did you run the full calibration
routine with the checkerboard (we never tested it in groovy, but it
should still work there)?

Best,
Lasse

--

Lasse Einig

[Roxana Leontie]

I've run a full calibration. But I still see this difference in the point cloud between the kinect and the robot model in rviz. Any idea why that might be? I've done the sanity check too, and the  wide_stereo/points point cloud matches the model reasonably well.  Would it be possible that the head_mount_kinect_ir_link is just at the wrong location? 

Thank you,

Roxana 

[Michael Görner]

In this case: are you sure the calibration included the kinect mount?
At least for one step of the calibration, there is an extra step
required to achieve this (See tutorial).
http://wiki.ros.org/pr2_calibration/Tutorials/Calibrating%20the%20PR2

We can't use the upstream calibration procedure on our PR2 due to
hardware changes, so I can't provide further insight here, sorry.


Michael / v4hn

--
Michael Görner, M.Sc. Cognitive Science, PhD Student
Universität Hamburg
Faculty of Mathematics, Informatics and Natural Sciences
Department of Informatics
Group Technical Aspects of Multimodal Systems

Vogt-Kölln-Straße 30
D-22527 Hamburg

Room: F-315
Phone: +49 40 42883-2432
Website: https://tams.informatik.uni-hamburg.de/people/goerner/

[Roxana Leontie]

Yes I did launch openni and extra kinect windows did appear during calibration.