Feature #142

Mechanical calibration, its application, evaluation, and verification

Added by Vladimir Smutny almost 5 years ago. Updated over 3 years ago.

Status:In ProgressStart date:03/27/2013
Priority:NormalDue date:
Assignee:Vladimir Smutny% Done:


Category:RobotsSpent time:-
Target version:-
Related Partner Organisation:CTU


This feature already started in autumn. Today we have measurement results from mechanical engineers and we have to apply them.
I suggest the following procedure:

1- Mr. Petrik implemented calibrated data into ROS. The differences between gripper positions according CAD data and calibrated data are up to 25 mm. This does not mean, that the robot is so bad. Where exactly are the sources of inaccuracy shall be investigated.

2- Let us use immediately calibration data for hand-eye calibration and for robot use (Wagner). The observed differences both good and bad shall be reported. E.g. robot calibration error in hand eye calibration is expected to drop. On the other hand one shall be careful during manipulation.

3- Let us make following experiment (Mr. Petrik): Use CAD data for transformations upto the base of the robor arm. Use calibrated data for the arm. Evaluate the difference of positions in the working space (e.g. above the table) using those two data sets. This experiment should describe the accuracy of the arm independently of the platform mounting and turn table. The experiment should be performed for both arms separately. Just looking into DH parameters we saw something like differences in the range of 0.1 mm upto 0.5 mm and some tenth or less of degrees.

4- Investigate the calibration results in the torso part (Smutny and Petrik), that is the inaccuracies in the turn table, platform, and mounting of arms to platform.


#1 Updated by Vladimir Smutny over 4 years ago

First loop of HW installation, planning, data capturing, image processing, and calibration is operational. We are not happy with the obtained parameters. The next round will include:
- high resolution camera,
- new calibration target,
- different set of calibration points.

#2 Updated by Vladimir Smutny over 3 years ago

The tasks listed above are done.
The Glasgow robot has to be calibrated. It seems that its inaccuracy is in the similar range like ours before calibration. Gerardo will investigate the possibility to calibrate the robot by Motoman or independent supplier.

We shall continue in the camera-target calibration procedure already started as well as calibration procedure by the Neovision granite target.
Smutny shall revive his effort together with Jan Heller.

Neovision shall adapt its camera-target mounting adapter for the second generation gripper.

Also available in: Atom PDF