UBR-1 Calibration

This is used to calibrate the UBR-1 robot. You probably aren’t looking to do that since you probably don’t have a UBR-1 - but you might be here to see an example of how to use robot_calibration in ROS 2.

There are several config files of interest here:

• config/capture.yaml - The capture step configuration.
• config/calibrate.yaml - The calibrate step configuration.
• config/calibration_poses.yaml - The capture poses to use. These have been manually tweaked to be collision free when directly interpolating between adjacent poses - so that we don’t have to rely on MoveIt to run planning. The first pose uses a ground plane finder to align the head camera to the ground, and then every other pose uses the LED finder.

There are several entry points:

• launch/calibrate.launch.py - Launch file that I use 99% of the time to calibrate the robot.
• launch/calibrate_base.launch.py - Launch file for calibrating base track width, base rollout, and IMU gain.
• scripts/calibrate_from_bag - If the launch file has already been run, and I have a bagfile, but want to tweak some aspect of the calibration step, this is how to run it.
• scripts/manual_calibration - This runs robot calibration in manual mode, where you have to move the robot arm and head yourself. I find this is generally only useful for developing new feature finders. This is NOT a launch file, since launch would eat our input data and we would not be able to hit enter…

An additional aspect of the calibration of the UBR-1 is that the LED detection works much better if auto exposure and white balance for the Primesense camera are turned off. This is done by running the scripts/camera_reconfigure.py script, which calls the SetParameter service to disable these features.

There are additional files that are really maybe not great examples:

• config/checkerboard_2d - this directy contains alternate capture and calibration configurations for using the Checkerboard2d finder. It performs suboptimally to the 3d version for a Primesense style camera that is already aligned. This particular configuration doesn’t actually calibrate the robot, instead it simply finds the checkerboard transform to validate the pipeline.