Developing for ROS can be less than painless. Locking your OS to a specific Ubuntu version just to develop ROS applications can make it a pain to use in the day-to-day, leading many to consider a VM or even another computer. Here, I propose a new solution using Docker with X11 passthrough and GPU support. As a bonus, you can incorporate it directly into your VSCode workspace. Here’s how.

Getting Setup

There’s a few initial steps we have to do to get everything running. Don’t worry, it won’t take too long.

Project GitHub: Here

Step 1 - Install the shell utilities

First of all, we have to install the docker-ros shell script from the GitHub repository:

cd /tmp
git clone https://github.com/JacisNonsense/docker-ros

cd docker-ros

# Only run the next line (rm -r) if you're replacing an old version
rm -r ~/.docker-ros

cp -r shell/ ~/.docker-ros

Step 2 - Making the script executable from anywhere

Next up, change your ~/.bashrc (or ~/.zshrc if you’re using ZSH)

Put these lines at the end:

export UID=${UID}
export GID=${GID}

source ~/.docker-ros/ros.sh

# OPTIONAL: Isolate the default HOME for the docker container if you don't want to passthrough your own.
ROS_DOCKER_HOME=path/to/my/isolated/home

Step 3 - Install the NVIDIA Docker runtime if you plan to allow GPU acceleration

If you’re not using a NVIDIA GPU, you can skip this step.

You will need to install nvidia-docker on your system to allow docker to use the runtime. This is an official runtime released by NVIDIA themselves.
The instructions for that are here: nvidia-docker GitHub

Running from the Command Line

There are multiple ways to run this script from the command line. I’m going to outline the basic options here, but you can inspect the script yourself to see all the options should you need them.

By default, the ros script will do the following for you:

  • Detect NVIDIA acceleration, and use the nvidia-docker2 runtime (you must install it first!)
  • Setup X forwarding
  • Create a new container image, passing through your local user and $HOME
  • Passthrough your current directory to /work via docker bind mount
  • Make the container interactive (-it --rm)

Interactive (Default)

$ ros <version>

Where <version> is one of the following:

  • kinetic, melodic - Aliases to kinetic-desktop-full and melodic-desktop-full
  • kinetic-ros-core, kinetic-ros-base, kinetic-robot, kinetic-perception, kinetic-desktop, kinetic-desktop-full
  • melodic-ros-core, melodic-ros-base, melodic-robot, melodic-perception, melodic-desktop, melodic-desktop-full

For example:

$ ros melodic
user@host:/work$

Launching a program directly

$ ros <version> <script>

For example:

$ ros melodic rviz

Specifying your own image

If you’ve built your own Docker image as a child of the jaci/ros image (for example, to install your own packages), this option is for you:

$ ros <version> --image yourname/yourimage:version

Setting up VSCode

You can use this image to develop applications for ROS in VSCode natively! You can use this to run your VSCode instance in the context of the Docker Container. This has a ton of benefits, the prime one being that you get intellisense from the ROS environment (i.e. code completion for C++ and Python).

Note that the VSCode part of this project does not forward X11, so it’s recommended to run any ROS applications from the ros script and use the VSCode installation purely for development.

Make sure you’ve setup the shell script above! VSCode ROS integration relies on the modifications in your .bashrc/.zshrc (you may have to log out and log in again after installing the shell scripts)

Step 1 - Install the Remote - Container extension

In the VSCode extension marketplace, find the Remote - Container extension and install it. This will allow us to open VSCode in the context of a docker container.

You can also do this by running the following in VSCode (CTRL + P)

ext install ms-vscode-remote.remote-containers

Step 2 - Apply the container configuration to the project

In your project root directory, copy the .devcontainer folder from ~/.docker-ros/.devcontainer. You can also find this folder on the GitHub

Step 3 - Open your project in the context of the container

Open your command palette (CTRL + SHIFT + P) and select Remote-Containers: Reopen Folder in Container. VSCode will build a new container and open the editor within the context of the container, providing C++ and Python intellisense with the ROS installation.

That’s it! You’re good to go!