This is an edit to the Easy Hand-Eye calibration program to create an automatic calibration program for the HIRO Sawyer robot.
This utilizes the eye-on-base implementation of the program using an Intel RealSense camera mounted on the ceiling above Sawyer.
This tutorial is picks up from the basic ROS setup outline in the HIRO wiki. If you have not yet setup your workspace using that guide, please do so first!
These are the same basic steps for implementation of Easy Hand-Eye.
- clone this fork of the repository into your workspace - this fork is different than the master version of easy_handeye, so if you have the original version, you will need to delete that fersion first:
cd ~/catkin_ws/src # replace with path to your workspace
if you have the original version, remove it from your workspace first:
rm -r easy_handeye/
clone the HIRO fork of the repository:
git clone https://github.com/HIRO-group/easy_handeyee
installing other dependent packages (feel free to skip any packages you already have in your workspace - the only package in the program that is different from the master is the easy_handeye):
git clone https://github.com/pal-robotics/aruco_ros # aruco_ros package
git clone https://github.com/IntelRealSense/realsense-ros # package for using Intel camera
git clone https://bitbucket.org/traclabs/trac_ik/src/master/ # inverse kinematics package
satisfy dependencies
cd .. # now we are inside ~/catkin_ws
rosdep install -iyr --from-paths src
- build
catkin build
The calibration is built on the principle of a 3D grid. The poses of the upper left and the lower right corners of the Intel RealSense's view are used to define the grid.
The y- and z-axes are split into three sections, and the x-axis is split into four sections. The calibration starts at the lower left corner of the grid. There are a total of 36 cells in the grid and the program moves Sawyer through each, sampling a random position within the bounds of the x, y, and z ranges of each cell as well as a random set of values of pitch, yaw and roll (rotation and tilt of the end effector). Once the end effector reaches the semi-random position, the program attempts to take a sample snapshot of the arUco marker's position.
If the marker is consistently readable to the camera (the camera can detect the orientation of the marker stably), the program takes a sample snapshot of the marker/end effectors position and saves it for calibration. If the marker is unstable or the camera is unable to read the marker in the position, the program will pause for a few seconds, print an error, and move on. The program need not take a sample of all positions to produce an accurate calibration, but a significant reduction in successful samples or a reduction in a certain area of the grid will affect calibration accuracy.
- Connect to the HIROLab_5G wifi network
- Plug the Intel RealSense camera cable into your machine
- Turn on Sawyer by pressing the power button on the computer at the base of the robot
If this is the first time you are running the calibration or you have changed the position of the camera above the Sawyer, you will need to intialize the boundaries of the grid by recording the new reference positions of the upper left and lower right corners of the camera's view.
For all terminal tabs used:
- Enter your intera workspace:
cd catkin_ws
./intera.sh
- Easy Hand-Eye runs on Python 2. To check your python version, run:
python
If the version shows 3.X, you can set up a temporary Python 2 environment
- Launch the sawyer_realsense launch file:
cd src/easy_handeye/easy_handeye/launch/
roslaunch sawyer_realsense_handeyecalibration.launch
This should launch Rviz Easy Hand-Eye Config and RQT Easy Hand-Eye Perspective. These are the GUIs used to perform a manual calibration with Easy Hand-Eye.
- Open another terminal tab (making sure to enter into the intera workspace and run Python 2) and show the Intel RealSense camera image:
rosrun image_view image_view image:=/aruco_tracker/result
- Navigate to the automatic calibration and run the script:
cd easy_handeye/rqt_easy_handeye/src/rqt_easy_handeye/
./automatic_calibration.py