Simulation of drones reading QR codes in Ware-houses

The repository can be accessed at https://github.com/webvenky/qr_scanning_drones

System Requirements

1. Install ROS and gazebo in the system I have used Ubuntu 14.04, ROS-indigo and Gazebo 7. Refer to http://wiki.ros.org/ROS/Installation for instructions.
2. Install libquirc Download the package libquirc –> https://github.com/dlbeer/quirc (I will also include the package separately along with the code.) Installation instructions are available in the README file. You can test the library by ./qrtest qr_image_name.jpg

Running the simulation

Make sure that you have run catkin build after you have placed the folders in the src folder in the src folder of your catkin workspace.

World: The description of the world is given in wall_qr3.world file.

Models: The models for the world file is given in the QR_Simulation/models folder. Place the models in the ~/.gazebo/models folder. The models contains models for several warehouse objects and the marker tags. If you want to change the marker tag, just replace the picture in the materials/texture folder in the marker models.

UAV: Firefly hexrotor is used as the platform for the simulation. I have used the model from the rotors_simulation package. (github_link) The properties of the hexrotor is given in the rotors_gazebo/resource/firefly_with_camera2.yaml file. It is important to account for the mass of additional payloads (such as gimbal and camera) for accurate vertical positioning. Other parameters for the lee_position_controller (such as velocity gain and angular rate gain ) are given in the rotors_gazebo/resource/lee_controller_firefly.yaml file.

Use this line to open the launch file. roslaunch rotors_gazebo mav_qr_code_scan2.launch. This loads up the world with the hexrotor at (0.0, -5.0, 0.0) as per the lines:

  <arg name="x" default="0.0"/>
  <arg name="y" default="-5.0"/>
  <arg name="z" default="0.1"/>

The simulation at the beginning looks like the image below:

The simulation is paused when it launches. When you press play, the UAV flies to the position (-2.5,-2.9,2.5) as per the line:

 <node name="waypoint_publisher" pkg="rotors_gazebo" type="waypoint_publisher" output="screen" args="-2.5 -2.9 2.5 90 1"/>

Lee position controller is utilized to position the UAV at the desired location. This controller is not a path planner. So use this controller only for minor changes in the position.

Navigation of UAV through waypoint publisher

The global path planning can be performed using a waypoint publisher node. Here the waypoint_publisher_aisle node is used to generate waypoints along a straight line at specified intervals. For example:

rosrun rotors_gazebo waypoint_publisher_aisle -2.5 -2.4 8.5 -2.4 20 2.5 90 __ns:=/firefly

generates 20 equally spaced waypoints between (-2.5,-2.4) and (8.5,-2.4) at an altitude of 1.0m. The UAV will also have a yaw orientation of 90 degrees. The source code of the waypoint publisher node is in the rotors_gazebo/src/waypoint_publisher_aisle.cpp file.

The 2-axis gimbal stabilizer

The gimbal will stabilize the camera to look at the horizontal direction irrespective of the camera roll or pitch.

This node is launched as part of the launch file.

<node name="firefly_cam_hold_horizontal2" pkg="venky_controls" type="cam_hold_horizontal2" args="-robot firefly" />

The source code of the gimbal stabilizer node is in the venky_controls/src/cam_hold_horizontal2.cpp file. The description of the gimbal and the camera properties (such as focal-length, resolution) is given in rotors_description/urdf/firefly_camera_2axis_gimbal.xacro file. The PID controller parameters are given in the venky_controls/config/camera_2axis_gimbal_controller_firefly.yaml.

Reading the QR Code

The QR code can be read using the read_qr_code node:

rosrun venky_controls read_qr_code -robot firefly

The node reads the image input from the topic /firefly/camera1/image_raw. This node utilizes the quirc library. The library decodes the qr code from the image input and prints the data. The library is capable of decoding multiple qr codes in a single image.

Depth Information

The depth information can be visualized using the node image_view:

 rosrun image_view image_view image:=/camera2/depth/image_raw

The parameters of the depth camera (libgazebo_ros_openni_kinect.so plugin) can be found in the file: rotors_description/urdf/firefly_camera_2axis_gimbal.xacro.

Video of the simulation: