Ball 🥎 Schubser 🤖

A robot 🤖 that pushes a ball 🥎 (into a goal 🥅).

Citation and link to our paper: Bracke, A., Gosein, D., Kürschner, M., Mayer, J., & Seefeld, J. (n.d.). Development of an Autonomous Robot for Detecting and Collecting Objects (p. 12) [Working Paper] URN: urn:nbn:de:bsz:953-opus-1006

Startup Instructions

1. Clone this repo
2. Inside .env add MASTER_IP=<your-ip>
3. Run docker-compose up -d --remove-orphans
4. Start Turtlebot and connect to via SSH ssh ubuntu@<your-ip> with default password: turtlebot
5. Set export ROS_MASTER_URI=<your-ip>
6. Run sh ~/launch.sh
7. Run sh ~/cam.sh

Simplified (beta)

1. Clone this repo
2. Execute export BIP=<your bots ip> && export MIP=<your pc ip>
3. Execute start-everything

Raspberry Pi 🐢 bot Preparation

To prepare the image for the Raspberry Pi follow these instructions: https://emanual.robotis.com/docs/en/platform/turtlebot3/sbc_setup/#sbc-setup

After the micro SD card is ready, boot it from your Turtlebot, connect via SSH: ssh ubuntu@<TURTLEBOT IP> (default password: turtlebot) and do the following:

1. Run: sudo usermod -aG root ubuntu to make the user root (see Lessons Learned).

2. Add your network to the /etc/netplan/50-cloud-init.yaml or prepare and copy it via SCP: scp 50-cloud-init.yaml ubuntu@<TURTLEBOT IP>:/etc/netplan/50-cloud-init.yaml. Also refer to this template.

3. Edit: nano ~/.bashrc to add the below variables and: source ~/.bashrc.

   export ROS_MASTER_URI=http://<REMOTE PC IP>:11311
   export ROS_HOSTNAME=<TURTLEBOT IP>

4. Install Raspberry Pi Cam and dependencies with the following commands:

   sudo apt install libraspberrypi-dev libraspberrypi0 libpigpiod-if-dev ros-noetic-compressed-image-transport ros-noetic-camera-info-manager ros-noetic-diagnostic-updater
   cd ~/catkin/src
   git clone https://github.com/UbiquityRobotics/raspicam_node
   catkin_make

5. Run: export TURTLEBOT3_MODEL=Burger to set the Turtlebot model.

6. Start without LIDAR sensor: roslaunch turtlebot3_bringup turtlebot3_core.launch.

7. Start the Cam via:

   rosparam set cv_camera/device_id 0
   rosrun cv_camera cv_camera_node

For step 5, 6 and 7 you can also use the provided Makefile and run make start-everything.

Detection Node

Download weights for yolov4 from: https://github.com/AlexeyAB/darknet/releases/download/darknet_yolo_v3_optimal/yolov4.weights and move the weights into detect/yolo_node/weights.

Dependencies:

• tensorflow
• pandas
• opencv-python

Lessons Learned

• The default user ubuntu on the Turtlebot image does not have the proper tty permissions. The problem could be solved by adding ubuntu to root user group, by executing: sudo usermod -aG root ubuntu, as the normal group used for that called dialout was not set for tty.
• ROS machines must have a resolved DNS name as they communicate with each other, see: http://wiki.ros.org/ROS/NetworkSetup.
• Router should not be connected to the internet in order to not disturb the connection
• Running the detection on the Turtlebot causes extrem delays:

1/1 [==============================] - 6s 6s/step
x1                    250
y1                    256
x2                    354
y2                    360
class_name    sports ball
score            0.693495
w                     104
h                     104
Name: 0, dtype: object
1/1 [==============================] - 6s 6s/step
x1                    250
y1                    255
x2                    354
y2                    360
class_name    sports ball
score            0.756736
w                     104
h                     105
Name: 0, dtype: object

Raspi cam

• ui: rqt
• list all packages: rospack list-names
• build packages inside ~/catkin_ws: catkin_make
• start master: roscore