Maintainer: Ingot Robotics
There have been previous efforts to support Turtlebot2 (based on Yujin's Kobuki mobile base), but none of them are docker-based nor supporting ROS2 Iron. This repository is our attempt to address this. We followed and then edited the instructions for the release-1.0.x for the Kobuki https://kobuki.readthedocs.io/en/release-1.0.x/software.html, and we took inspiration from the ROS 2 Bouncy Turtlebot2 repo: https://github.com/ros2/turtlebot2_demo/tree/master.
The repository contains a dockerfile, three udev rules files, and two packages to run a Turtlebot2 using a Hokuyo URG-04 as the navigation sensor and an Intel RealSense as a depth camera.
Additional navigation sensors may be added if requested, and pull requests are always appreciated.
After building up the Turtlebot2 and installing Ubuntu 22.04 on the robot computer, install Docker (docker.io), git, and any other favorite packages. We use a separate laptop for controlling the robot, so the robot computer also needs Open-SSH server and tmux.
Clone this repo, and from that directory, the dockerfile can be built with the command
docker build -t ingot/turtlebot2-ros-iron:desktop -f turtlebot2_ros2.dockerfile --build-arg from_image=osrf/ros:iron-desktop --build-arg parallel_jobs=4 .
The dockerfile has default values for the base image (ros:iron) and parallel build jobs (8), but the command above overrides the defaults. Adjust as fits your needs.
*Note: as of 15-Nov-2023, the official Docker Hub ROS 2 Iron image (ros:iron) has an issue with the sgml-base package. We will try to update this repo with a work-around, but for the time being, stick with the larger OSRF desktop image (osrf/ros:iron-desktop)
The robot computer needs device udev rules installed in /etc/udev/rules.d/. The Kobuki udev rule comes from the Kobuki organization: https://raw.githubusercontent.com/kobuki-base/kobuki_ftdi/devel/60-kobuki.rules. The Hokuyo udev rule can be found many places online. The RealSense udev rule comes from the IntelRealSense organization: https://github.com/IntelRealSense/librealsense/blob/master/config/99-realsense-libusb.rules.
After the docker container is built, it can be launched with the command
docker run -it --device=/dev/ttyACM0 --device=/dev/kobuki --device=/ttyUSB0 -v /dev:/dev --device-cgroup-rule "c 81:* rmw" --device-cgroup-rule "c 189:* rmw" ingot/turtlebot2-ros-iron:desktop.
If you are not running any of the Kobuki utilities like kobuki-version-info, you can omit the --device=/dev/kobuki switch. The RealSense grabs multiple video devices, and we will look to improve the access to them so the -v /dev:/dev can be replaced.
If you will run rviz on a separate machine, adding --network=host is a docker networking work-around to allow containers on separate machines but the same local network to communicate.
Inside the docker container, first source the overlay: source install/setup.bash, then launch the Turtlebot2 with ros2 launch turtlebot2_bringup turtlebot2_bringup.launch.py
The Turtlebot2 can be controlled by a separate instance of the same container. To run this instance, drop all of the --device switches from the docker run command above. Once in the container, source the overlay, and then run a keyboard or joystick control node, such as ros2 run kobuki_keyop kobuki_keyop_node or ros2 run teleop_twist_keyboard teleop_twist_keyboard. Alternatively, rviz2 and the Navigation2 stacks allow the use of naviation goals on the map.
If you want to interact with the Kobuki base without bringing up the full Turtlebot2 stack, launch the Kobuki with ros2 launch kobuki_node kobuki_node-launch.py and run keyboard teleoperation with remapping for the velocity commands:
ros2 run kobuki_keyop kobuki_keyop_node --ros-args --remap /cmd_vel:=/commands/velocity or ros2 run teleop_twist teleop_twist_keyboard --ros-args -r /cmd_vel:=/commands/velocity.
Note that for MacOS environment, you'll need to add the flag --platform linux/amd64 for all docker commands to specify non-arm variant