This is a sample ROS2 workspace for IARC Mission 9 and TAC Challenge 2022

Overview

Previously, we have had trouble that code builds, but only on one spesific computer. By building within a Docker image, we make sure that all dependencies are included and work - not just on one PC, but on all relevant computers.

Because we use a NVIDIA Jetson AGX for Mission 9, and likely a similar platform for TAC, the Dockerfile has to work for both amd64 and arm64 architectures.

Useful features

• Pre-configured development container for simple development in VSCode
  • Works on most platforms including Ubuntu, Mac (Intel & M1) and Windows
  • Automatic formating and linting of C++ using clang-format and clang-tidy
  • Works in both Docker root and rootless modes (just change remoteUser in .devcontainer/devcontainer.json)
• Github Actions make sure that the code builds on all systems
  • Automatic builds for amd64 and arm64
  • Ensures that tests pass with colcon test
• Uploads docker image to GitHub Container Registry
  • By default turned on for main and devel branches

Using the image

Prerequisites

Install and run Docker

Install and run Docker. It is available for most platforms, including Ubuntu, MacOS (Intel & M2) and Windows.

On Linux, it is recommended to run Docker in rootless mode.

Login to GitHub Container Registry

NB! Some of the docker images used in the build process are private for the AscendNTNU organization. In order to pull the images, you need to authenticate to GitHub Container Registry using your Ascend GitHub account. See the following guide on how to authenticate: Authenticating To the Container Registry

Developing in VSCode

This requires installing the Remote - Containers extension.

NB! If using docker on a Mac, follow the same instructions as for root-mode. I.e. uncomment the following line in .devcontainer/devcontainer.json:

// If running Docker in root mode, uncomment the "remoteUser" line.
// If using Docker rootless, leave it commented out.
// More info: https://aka.ms/vscode-remote/containers/non-root.
"remoteUser": "ros",

If using VSCode, simply open the development container : CTRL + SHIFT + P ... "Remote-Containers: Rebuild and Reopen in Container"

Running multiple ROS2 containers together

If you want to run multiple ROS2 containers that should be able to communicate, they need to be added to the same Docker network. To do this:

1. Create a docker network: docker network create ros2-network-name
2. Add the container to the network, this can be done either
   • for the devcontainer, by uncommenting the network-line in devcontainer.json
   • when starting the container by adding the option --network=ros2-network-name to the run command
   • while the container is running, by running docker network connect ros2-network-name CONTAINER.
   • in a docker compose file by adding network: ros2-network-name

Developing with this template

Requirements

• docker/Dockerfile must contain a stage named deploy that features a fully built ROS2 workspace on both amd64 and arm64 architectures.
  • If the program is only ever supposed to run on one of the architectures, remove the the unused architecure file from TARGET_PLATFORMS in .github/worklfows/build-and-test.yml and .github/worklfows/build-and-upload.yml

Installing dependencies

Any additional dependencies should be installed in the Dockerfile. If there are buld stages that take a long time, consider uploading the image to our GitHub Container Registry, so that others can download a pre-built image.