Franka Handeye Calibration

This repo is tested with Ros2 Humble Ubuntu 22.04 & FR3 & Realsense 455, which can reallize eye_on_base or eye_on_hand calibration with Aruco QR code. One has to manually control the franka moveit to move the robot, then sample 6 or more data to acquire relatively precise transpose result. About ~1.5cm deviation, which can be manually correct during using.

Related Repos: easy_handeye2, ros2_aruco, libfranka, franka_ros2, librealsense, realsense_ros.

Renference Repo: ar4_hand_eye_calibration. Appreciate for this author's work.

Ros1 Neotic Franka Handeye Calbration: Handeye Calibration Hardware: Franka1, realsense d435, aruco

Preparation

1. Find an aruco marker on this websit, and attach it to the franka end-effector. If you don't want to change the code, this repo use aruco marker information:

   Dictionary: 5x5(50, 100, 250, 1000);
   Marker ID: 250;
   Marker size, mm: 50
   

   If you change the information, remember to change the corresponding part in config/aruco_parameters.yaml

2. Install Ros2 humble

3. Install libfranka and franka_ros2. Remember to choose the corresponding version from here of your robot firmware.

4. Install librealsense & realsense_ros from source (Remember to choose the corresponding version. This repo uses librealsense: v2.55.1,realsense_ros: 4.55.1). Or use command:

   sudo apt install ros-iron-librealsense2* ros-iron-realsense2-*
   

Installation

Check the Aruco and Easy_Handeye git version, incase they update their repo to Ubuntu24

mkdir -p ~/calib_ws/src
cd ~/calib_ws/src

git clone https://github.com/ChengYaofeng/franka_handeye_calibration_ros2.git

git clone https://github.com/JMU-ROBOTICS-VIVA/ros2_aruco.git

git clone https://github.com/marcoesposito1988/easy_handeye2

rosdep install -iyr --from-paths src

colcon build

Error & Solutions

1. ModuleNotFoundError: No module named 'tf_transformations'

   sudo apt-get install ros-humble-tf-transformations
   

2. [aruco_node-1] AttributeError: module 'cv2.aruco' has no attribute 'Dictionary_get'. Did you mean: 'Dictionary'?

   The opencv-version is to high. pip uninstall opencv-python. Directly use the system version.

Execution

Activate FR3 Moveit, adjust the command to your own robot ip.

ros2 launch franka_fr3_moveit_config moveit.launch.py robot_ip:='172.16.0.2'

In a new terminal, launch the calibration file. You need to put the camera to capture the aruco marker on the robot before running the following command. Otherwise you'll meet the error.

ros2 launch franka_handeye_calib calibrate.launch.py 

Then use the rviz to move the robot 6 or more times to acquire relatively precise transpose result. About ~1.5cm deviation, which can be manually correct during using.

The final results will saved in ~/.ros2/easy_handeye/calibrations/fr3_calibration.calib