Docker installation#

Info

Since this page explains Docker-specific information, it is recommended to see Source installation as well if you need detailed information.

Prerequisites#

Git

For NVIDIA Jetson devices, install JetPack >= 5.0

How to set up a development environment#

Clone autowarefoundation/autoware and move to the directory.

git clone https://github.com/autowarefoundation/autoware.git
cd autoware

If you want to use ROS 2 Humble, use the humble branch.

git clone https://github.com/autowarefoundation/autoware.git -b humble
cd autoware

You can install the dependencies either manually or using the provided Ansible script.

Note: Before installing NVIDIA libraries, confirm and agree with the licenses.

CUDA

Installing dependencies manually#

Installing dependencies using Ansible#

Be very careful with this method. Make sure you read and confirmed all the steps in the Ansible configuration before using it.

If you've manually installed the dependencies, you can skip this section.

./setup-dev-env.sh docker

You might need to log out and log back to make the current user able to use docker.

How to set up a workspace#

Warning

Before proceeding, confirm and agree with the NVIDIA Deep Learning Container license. By pulling and using the Autoware Universe images, you accept the terms and conditions of the license.

Create the autoware_map directory for map data later.
```
mkdir ~/autoware_map
```

Launch a Docker container.

For amd64 architecture computers with NVIDIA GPU:

rocker --nvidia --x11 --user --volume $HOME/autoware --volume $HOME/autoware_map -- ghcr.io/autowarefoundation/autoware-universe:latest-cuda

If you want to use ROS 2 Humble:

rocker --nvidia --x11 --user --volume $HOME/autoware --volume $HOME/autoware_map -- ghcr.io/autowarefoundation/autoware-universe:humble-latest-cuda

If you want to run container without using NVIDIA GPU, or for arm64 architecture computers:

rocker -e LIBGL_ALWAYS_SOFTWARE=1 --x11 --user --volume $HOME/autoware --volume $HOME/autoware_map -- ghcr.io/autowarefoundation/autoware-universe:latest-cuda

For detailed reason could be found here

For more advanced usage, see here.

After that, move to the workspace in the container:

cd autoware

Create the src directory and clone repositories into it.
```
mkdir src
vcs import src < autoware.repos
```
Update dependent ROS packages.

The dependency of Autoware may change after the Docker image was created. In that case, you need to run the following commands to update the dependency.
```
sudo apt update
rosdep update
rosdep install -y --from-paths src --ignore-src --rosdistro $ROS_DISTRO
```
Build the workspace.
```
colcon build --symlink-install --cmake-args -DCMAKE_BUILD_TYPE=Release
```
If there is any build issue, refer to Troubleshooting.

How to update a workspace#

Update the Docker image.

docker pull ghcr.io/autowarefoundation/autoware-universe:latest-cuda

Launch a Docker container.

For amd64 architecture computers:

rocker --nvidia --x11 --user --volume $HOME/autoware -- ghcr.io/autowarefoundation/autoware-universe:latest-cuda

If you want to run container without using NVIDIA GPU, or for arm64 architecture computers:

rocker -e LIBGL_ALWAYS_SOFTWARE=1 --x11 --user --volume $HOME/autoware -- ghcr.io/autowarefoundation/autoware-universe:latest-cuda

Update the .repos file.
```
cd autoware
git pull
```

Update the repositories.

vcs import src < autoware.repos
vcs pull src

Build the workspace.

colcon build --symlink-install --cmake-args -DCMAKE_BUILD_TYPE=Release

Troubleshooting#

Here are solutions for a few specific errors:

cuda error: forward compatibility was attempted on non supported hw#

When starting Docker with GPU support enabled for NVIDIA graphics, you may sometimes receive the following error:

docker: Error response from daemon: OCI runtime create failed: container_linux.go:349: starting container process caused "process_linux.go:449: container init caused \"process_linux.go:432: running prestart hook 0 caused \\\"error running hook: exit status 1, stdout: , stderr: nvidia-container-cli: initialization error: cuda error: forward compatibility was attempted on non supported hw\\\\n\\\"\"": unknown.
ERROR: Command return non-zero exit code (see above): 125

This usually indicates that a new NVIDIA graphics driver has been installed (usually via apt) but the system has not yet been restarted. A similar message may appear if the graphics driver is not available, for example because of resuming after suspend.

To fix this, restart your system after installing the new NVIDIA driver.

Docker with NVIDIA gpu fails to start Autoware on arm64 devices#

When starting Docker with GPU support enabled for NVIDIA graphics on arm64 devices, e.g. NVIDIA jetson AGX xavier, you may receive the following error:

nvidia@xavier:~$ rocker --nvidia --x11 --user --volume $HOME/autoware -- ghcr.io/autowarefoundation/autoware-universe:galactic-latest-cuda-arm64
...

Collecting staticx==0.12.3
Downloading https://files.pythonhosted.org/packages/92/ff/d9960ea1f9db48d6044a24ee0f3d78d07bcaddf96eb0c0e8806f941fb7d3/staticx-0.12.3.tar.gz (68kB)
Complete output from command python setup.py egg_info:
Traceback (most recent call last):
File "", line 1, in
File "/tmp/pip-install-m_nm8mya/staticx/setup.py", line 4, in
from wheel.bdist_wheel import bdist_wheel
ModuleNotFoundError: No module named 'wheel'

Command "python setup.py egg_info" failed with error code 1 in /tmp/pip-install-m_nm8mya/staticx/
...

This error exists in current version of rocker tool, which relates to the os_detection function of rocker.

To fix this error, temporary modification of rocker source code is required, which is not recommended.

At current stage, it is recommended to run docker without NVIDIA gpu enabled for arm64 devices:

rocker -e LIBGL_ALWAYS_SOFTWARE=1 --x11 --user --volume $HOME/autoware -- ghcr.io/autowarefoundation/autoware-universe:latest-cuda

This tutorial will be updated after official fix from rocker.

Tips#

Non-native arm64 System#

This section describes a process to run arm64 systems on amd64 systems using qemu-user-static.

Initially, your system is usually incompatible with arm64 systems. To check that:

$ docker run --rm -t arm64v8/ubuntu uname -m
WARNING: The requested image's platform (linux/arm64/v8) does not match the detected host platform (linux/amd64) and no specific platform was requested
standard_init_linux.go:228: exec user process caused: exec format error

Installing qemu-user-static enables us to run arm64 images on amd64 systems.

$ sudo apt-get install qemu-user-static
$ docker run --rm --privileged multiarch/qemu-user-static --reset -p yes
$ docker run --rm -t arm64v8/ubuntu uname -m
WARNING: The requested image's platform (linux/arm64/v8) does not match the detected host platform (linux/amd64) and no specific platform was requested
aarch64

To run Autoware's Docker images of arm64 architecture, add the suffix -arm64.

$ docker run --rm -it ghcr.io/autowarefoundation/autoware-universe:galactic-latest-cuda-arm64
WARNING: The requested image's platform (linux/arm64) does not match the detected host platform (linux/amd64) and no specific platform was requested
root@5b71391ad50f:/autoware#