View on GitHub

tuw-cpsg.github.io

ROS with Docker

The following tutorial is a description of how I’ve got started with ROS.

Meanwhile, I use my own Dockerfiles to use ROS (e.g., run demos with Daisy) which saves some time (specifying options, e.g., environment variables, working directory, or installing packages) and uses x11docker for GUI tools (I strongly recommend!).

Installation

Install docker (see installation instructions for your platform on docker docs - About Docker CE and post-installation steps).

If you want to use ROS GUI tools build a custom docker image (there are no official desktop-full images?):

$ cat Dockerfile
FROM ros:melodic
RUN apt-get update && apt-get install -y \
    ros-melodic-desktop-full
$ docker build -t ros:melodic-desktop-full .

If you only use the CLI, simply pull the ROS docker image:

$ docker pull ros:melodic

Run ROS container:
```
$ docker run \
    -it \
    -v ~/catkin_ws:/root/catkin_ws \
    -v ~/.ssh:/root/.ssh \
    --network host \
    --env ROS_MASTER_URI=http://<your hostname>:11311 \
    --name ros
    ros:melodic[-desktop-full]
```
This command starts the ROS docker image ros in an interactive terminal mode (-i -t) and mounts the host catkin workspace ~/catkin_ws to the home directory of the running container. Note, mounting does not resolve symlinks!

The container uses the host’s network with no isolation (!) (as if the container processes would run on the host) using the host’s ssh configuration, /etc/hosts and hostname. The environment variable of the master URI is set, however, you might have a ros-entrypoint-script sourcing the workspaces anyway where you can add the export of the variable.

It is convenient to assign a name to the container:
- to start another shell in the same container (this is not the typical Docker way to do it – one process per container).
```
$ docker exec -it ros bash
```
- to stop the container.
```
$ docker stop ros
```
- to restart the container.
```
$ docker start ros
```

Usage Examples

Useful Docker commands

$ docker ps
$ docker ps -a
$ docker image ls
$ docker info
$ docker start <name>
$ docker exec -it <name> bash
$ docker stop <name>
$ docker container rm <name>

Inspect ROS nodes

On the host do:

Run the ROS docker container created before:

$ docker start ros
$ docker exec -it ros bash

Start some nodes in the container, e.g.:

$ source /opt/ros/melodic/setup.sh
$ roscore &
  :
$ rostopic pub -r 1 /test std_msgs/String "test"

Run a docker container executing rqt_graph:
```
$ docker run \
    --net=host \
    --user $(id -u) \
    --env DISPLAY=$DISPLAY \
    -v "/tmp/.X11-unix:/tmp/.X11-unix:rw" \
    --name rqt_graph \
    ros:melodic-desktop-full \
    rqt_graph
```
Use the host’s network as before to be part of the ROS network. The GUI shall be executed by the current user – not root (see also ROS Docker and GUI). Note that --user=$USER did not work for me. The name is unknown but the UID works. The display X server is shared.

Give the container a name to be able to restart it later, select the desktop-full image and execute rqt_graph.

You may create an alias for convenience (in .bashrc or similar):

alias docker_gui="docker run --net=host --user $(id -u) --env DISPLAY=$DISPLAY -v '/tmp/.X11-unix:/tmp/.X11-unix:rw'"

So you can later start a ROS GUI tool by:

$ docker_gui --name rqt_graph ros:melodic-desktop-full rqt_graph

You can then (re-)start rqt_graph by:

$ docker start rqt_graph

Some GUI apps still make troubles (e.g., rviz). You may try x11docker.

Controlling Daisy

On the host do:

Setup the network with Daisy. You will need to connect to Daisy’s network install openssh-server and exchange SSH keys (see network setup).

Create a folder for your ROS workspace and download the general-ros-modules repository including the tele-operation package:

$ mkdir -p ~/catkin_ws/src
$ cd ~/catkin_ws/src
$ git clone https://github.com/tuw-cpsg/general-ros-modules.git

Run ROS in docker container created before.

$ docker start ros
$ docker exec -it ros bash

In the container do:

Install p2os_msgs (replace melodic with your distro) which is needed to communicate with the robot driver (e.g., motor control). bash-completion is just for convenience.
```
$ apt update
$ apt install bash-completion
$ apt install ros-melodic-p2os-msgs
```

Initialize the ROS workspace (build and source):

$ source /opt/ros/melodic/setup.sh
$ cd ~/catkin_ws
$ catkin_make
$ source devel/setup.sh

Launch the teleop node to control Daisy:

$ roslaunch pioneer_teleop drive.launch \
    notebook:=<your hostname> \
    robot:=daisy robot-distro:=indigo robot-port:=/dev/ttyTHS1

References

Docker Documentation
docker docs - About Docker CE
ROS Docker and GUI – outdated but still useful instructions to run ROS with Docker
ROS Docker images – ROS images and examples
ROS workspace – how to create a ROS workspace
x11docker – the simple way to run GUI apps in docker
Daisy Demos with Docker – images to control Daisy with ROS on docker

2019-02-08 | Denise Ratasich