This repository contains the code and supporting files to run TurtleBot 2 demos using ROS 2. Due to reliance on existing Linux-only code and dependencies, these demos are intended for use only on Linux (that could change in the future).
The following instructions are for ROS Bouncy, if you are using ROS Ardent please refer to these instructions.
This demo assumes that you have an Orbbec Astra depth camera. Extra work would be required to use the Kinect or Asus Xtion Pro. Without an Astra, you can still do joystick teleop. For instructions on how to setup your turtlebot please see Setup your turtlebot2
First, install ROS2 from binaries following these instructions
Then install the turtlebo2 demo specific packages:
export ROS1_DISTRO=melodic
sudo apt install ros-bouncy-turtlebot2* ros-$ROS1_DISTRO-kobuki-ftdi
This assumes that you have ROS Kinetic installed (or at lease have the ros apt repository in your sources)
First, install ROS2 from source following these instructions
Then get the turtlebo2 demos specific code:
cd <YOUR_ROS2_WORKSPACE>
wget https://raw.githubusercontent.com/ros2/turtlebot2_demo/bouncy/turtlebot2_demo.repos
vcs import src < turtlebot2_demo.repos
export ROS1_DISTRO=melodic # or kinetic if using Ubuntu Xenial
sudo apt-get install --no-install-recommends -y libboost-iostreams-dev libboost-regex-dev libboost-system-dev libboost-thread-dev libceres-dev libgoogle-glog-dev liblua5.2-dev libpcl-dev libprotobuf-dev libsdl1.2-dev libsdl-image1.2-dev libsuitesparse-dev libudev-dev libusb-1.0-0-dev libyaml-cpp-dev protobuf-compiler python-sphinx ros-$ROS1_DISTRO-catkin ros-$ROS1_DISTRO-kobuki-driver ros-$ROS1_DISTRO-kobuki-ftdi
Reason for each dependency:
ros-$ROS1_DISTRO-kobuki-driver
: our ROS 2 kobuki driver builds on top of this package (and its dependencies)ros-$ROS1_DISTRO-kobuki-ftdi
: we use audev
rule from this packageros-$ROS1_DISTRO-common-msgs
: to support use of theros1_bridge
, we need the ROS 1 messages available (TODO: document use of the bridge to view depth images and other stuff)
For resource constrained platforms we will split the build into 2 steps to make sure not to overflow the memory
colcon build --symlink-install --packages-skip cartographer cartographer_ros cv_bridge opencv_tests ros1_bridge turtlebot2_amcl turtlebot2_drivers turtlebot2_follower turtlebot2_cartographer turtlebot2_teleop vision_opencv
Now the resource intensive packages and the ones depending on ROS1 packages:
source /opt/ros/$ROS1_DISTRO/setup.bash
colcon build --symlink-install --packages-select cartographer cartographer_ros turtlebot2_amcl turtlebot2_cartographer turtlebot2_drivers turtlebot2_follower turtlebot2_teleop
Go grab a coffee (or a meal if you compile on ARM)
If you installed from binary you'll need to download the udev rule by hand:
wget https://raw.githubusercontent.com/ros2/ros_astra_camera/ros2/56-orbbec-usb.rules
cd ~/ros2_ws/src/ros2/ros_astra_camera
sudo cp 56-orbbec-usb.rules /etc/udev/rules.d
sudo cp `rospack find kobuki_ftdi`/57-kobuki.rules /etc/udev/rules.d
sudo service udev reload
sudo service udev restart
If installed from Debian packages
source /opt/ros/$ROS1_DISTRO/setup.bash
source /opt/ros/bouncy/setup.bash
If installed from source
source /opt/ros/$ROS1_DISTRO/setup.bash
source ~/ros2_ws/install/local_setup.bash
You'll need to do this step for every terminal you use for these demos
This is a classic teleoperation demo where the robot can be driven around using a gamepad controller. This demo has been tested with logitech controllers and uses RB
as a deadman, the left joystick for driving forward/backward and the right joystick for rotation.
Try the launch file:
launch `ros2 pkg prefix turtlebot2_teleop`/share/turtlebot2_teleop/launch/turtlebot_joy.py
Or, run the nodes separately:
ros2 run turtlebot2_drivers kobuki_node
ros2 run joy joy_node
Note: this demo assumes that your controller is in D mode (switch on the back) and that the MODE led is on.
This demo uses the astra camera to detect blobs in the depthimage and follow them Try the launch file:
launch `ros2 pkg prefix turtlebot2_follower`/share/turtlebot2_follower/launch/turtlebot_follow.py
Or, run the nodes separately.
ros2 run turtlebot2_drivers kobuki_node
ros2 run astra_camera astra_camera_node -- -dw 320 -dh 240 -C -I
ros2 run turtlebot2_follower follower
This demo is using Google cartographer to build a map of the environment. The resulting map can be visualized in RViz using the ros1_bridge (more information below).
Try the launch file:
launch `ros2 pkg prefix turtlebot2_cartographer`/share/turtlebot2_cartographer/launch/turtlebot_carto_2d.py
The created map can be visualized in Rviz on a remote computer by using the dynamic bridge that converts messages between ROS1 and ROS2. This assumes that you have a ROS2 dynamic bridge on your system.
Setup your sources as explained on the setup sources section and then run
sudo apt update && sudo apt install ros-bouncy-ros1-bridge
Build your ROS2 workspace as explained in these instructions.
Terminal A:
. /opt/ros/$ROS1_DISTRO/setup.bash
roscore
Terminal B:
. /opt/ros/$ROS1_DISTRO/setup.bash
. <YOUR_ROS2_WORKSPACE>
ros2 run ros1_bridge dynamic_bridge
Terminal C:
. /opt/ros/$ROS1_DISTRO/setup.bash
rosrun rviz rviz
Topics you can visualize in Rviz:
- the map on the topic
/map
- the transforms on the topic
/tf
- the depth images on the topic
/depth
- the laserscans on the topic
/scans
Note: ROS Bouncy uses cartographer 0.3.0 that doesn't publish an occupancy grid on the /map
topic anymore.
See the AMCL demo README
What's happening here compared to the ROS 1 versions of these demos? Well, it's 100% ROS 2, with no bridge or shim. We took 4 different approaches in building the different pieces:
- Kobuki driver: we wrote a new, very small rclcpp node that calls into the existing kobuki driver packages, which are organized to be roscpp-independent. In this case, we're building on top of ROS 1 packages, but they don't use
roscpp
or other parts of the ROS 1 middleware, so we're just using them as supporting libraries. - Astra driver: we forked and ported the existing ROS 1 package (there's no roscpp-independent package separation).
- Joystick driver: we wrote a simple rclcpp node from scratch (Linux-only for now).
- Follower node: we created a new package into which we copied and then ported the ROS 1 follower nodelet.
As we start migrating more code to ROS 2, we'll discover more about these kinds of techniques and arrive at some best practices that we can recommend for similar projects.