Skip to content

Troubleshooting#

Setup issues#

When installing CUDA, errors may occur because of version conflicts. To resolve these types of errors, try one of the following methods:

  • Unhold all CUDA-related libraries and rerun the setup script.

    sudo apt-mark unhold  \
  "cuda*"             \
  "libcudnn*"         \
  "libnvinfer*"       \
  "libnvonnxparsers*" \
  "libnvparsers*"     \
  "tensorrt*"         \
  "nvidia*"

./setup-dev-env.sh

  • Uninstall all CUDA-related libraries and rerun the setup script.

    sudo apt purge        \
  "cuda*"             \
  "libcudnn*"         \
  "libnvinfer*"       \
  "libnvonnxparsers*" \
  "libnvparsers*"     \
  "tensorrt*"         \
  "nvidia*"

sudo apt autoremove

./setup-dev-env.sh

Warning

Note that this may break your system and run carefully.

  • Run the setup script without installing CUDA-related libraries.

    ./setup-dev-env.sh --no-nvidia

Warning

Note that some components in Autoware Universe require CUDA, and only the CUDA version in the env file is supported at this time. Autoware may work with other CUDA versions, but those versions are not supported and functionality is not guaranteed.

Build issues#

Insufficient memory#

Building Autoware requires a lot of memory, and your machine can freeze or crash if memory runs out during a build. To avoid this problem, 16-32GB of swap should be configured.

# Optional: Check the current swapfile
free -h

# Remove the current swapfile
sudo swapoff /swapfile
sudo rm /swapfile

# Create a new swapfile
sudo fallocate -l 32G /swapfile
sudo chmod 600 /swapfile
sudo mkswap /swapfile
sudo swapon /swapfile

# Optional: Check if the change is reflected
free -h

For more detailed configuration steps, along with an explanation of swap, refer to Digital Ocean's "How To Add Swap Space on Ubuntu 20.04" tutorial

If there are too many CPU cores (more than 64) in your machine, it might requires larger memory. A workaround here is to limit the job number while building.

MAKEFLAGS="-j4" colcon build --symlink-install --cmake-args -DCMAKE_BUILD_TYPE=Release

You can adjust -j4 to any number based on your system. For more details, see the manual page of GNU make.

By reducing the number of packages built in parallel, you can also reduce the amount of memory used. In the following example, the number of packages built in parallel is set to 1, and the number of jobs used by make is limited to 1.

MAKEFLAGS="-j1" colcon build --symlink-install --cmake-args -DCMAKE_BUILD_TYPE=Release --parallel-workers 1

Note

By lowering both the number of packages built in parallel and the number of jobs used by make, you can reduce the memory usage. However, this also means that the build process takes longer.

Errors when using the latest version of Autoware#

If you are working with the latest version of Autoware, issues can occur due to out-of-date software or old build files.

To resolve these types of problems, first try cleaning your build artifacts and rebuilding:

rm -rf build/ install/ log/
colcon build --symlink-install --cmake-args -DCMAKE_BUILD_TYPE=Release

If the error is not resolved, remove src/ and update your workspace according to installation type (Docker / source).

Warning

Before removing src/, confirm that there are no modifications in your local environment that you want to keep!

If errors still persist after trying the steps above, delete the entire workspace, clone the repository once again and restart the installation process.

rm -rf autoware/
git clone https://github.com/autowarefoundation/autoware.git

Errors when using a fixed version of Autoware#

In principle, errors should not occur when using a fixed version. That said, possible causes include:

  • ROS 2 has been updated with breaking changes.
  • Your local environment is broken.
    • Confirm your .bashrc file, environment variables, and library versions.

In addition to the causes listed above, there are two common misunderstandings around the use of fixed versions.

  1. You used a fixed version for autowarefoundation/autoware only. All of the repository versions in the .repos file must be specified in order to use a completely fixed version.

  2. You didn't update the workspace after changing the branch of autowarefoundation/autoware. Changing the branch of autowarefoundation/autoware does not affect the files under src/. You have to run the vcs import command to update them.

Error when building python package#

During building the following issue can occurs

pkg_resources.extern.packaging.version.InvalidVersion: Invalid version: '0.23ubuntu1'

The error is due to the fact that for versions between 66.0.0 and 67.5.0 setuptools enforces the python packages to be PEP-440 conformant. Since version 67.5.1 setuptools has a fallback that makes it possible to work with old packages again.

The solution is to update setuptools to the newest version with the following command

pip install --upgrade setuptools

Docker/rocker issues#

If any errors occur when running Autoware with Docker or rocker, first confirm that your Docker installation is working correctly by running the following commands:

docker run --rm -it hello-world
docker run --rm -it ubuntu:latest

Next, confirm that you are able to access the base Autoware image that is stored on the GitHub Packages website

docker run --rm -it ghcr.io/autowarefoundation/autoware-universe:latest

Runtime issues#

Map does not display when running the Planning Simulator#

When running the Planning Simulator, the most common reason for the map not being displayed in RViz is because the map path has not been specified correctly in the launch command. You can confirm if this is the case by searching for Could not find lanelet map under {path-to-map-dir}/lanelet2_map.osm errors in the log.

Another possible reason is that map loading is taking a long time due to poor DDS performance. To address this issue, first enable localhost-only communication to reduce network traffic, and then tune DDS settings if the problem continues to occur.

Simply put, add the following settings to .bashrc and reboot the terminal. In many cases this is not a problem.

export ROS_LOCALHOST_ONLY=1
if [ ! -e /tmp/cycloneDDS_configured ]; then
  sudo sysctl -w net.core.rmem_max=2147483647
  sudo ip link set lo multicast on
  touch /tmp/cycloneDDS_configured
fi

Note

DDS configuration can be determined by running the following command.

echo $RMW_IMPLEMENTATION  // if Cyclone DDS is configured, this command will return "rmw_cyclonedds_cpp"

If that does not work or you need more information, read the following documents.

  1. Enable localhost-only communication
  2. DDS settings

Multicast is disabled#

If you get the error message selected interface "{your-interface-name}" is not multicast-capable: disabling multicast, run the following command to allow multicast.

sudo ip link set multicast on {your-interface-name}

Node performance degradation#

If you notice a decrease in the running performance of a node, such as issue2597, you need to check if your compilation instructions use Release or RelWithDebInfo tags. If not, recompile the project using the following instructions:

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