Carla

We use carla as our simulator. However it is really annoying to setup carla inside docker. So, we document the procedure about how to make it work.

Step by step

Update nvidia driver version to 550.

Since there is known bug in lower version, just upgrade to latest one.

sudo apt install nvidia-driver-550

Upgrade nvidia-container-toolkit.

Please refer to the document in: https://docs.nvidia.com/datacenter/cloud-native/container-toolkit/latest/install-guide.html However, some network problem make me turn to another way. Download the .deb file from github and install every package manually.

# Install them in order.
sudo dpkg -i libnvidia-container-dev_1.17.3-1_amd64.deb
sudo dpkg -i libnvidia-container1_1.17.3-1_amd64.deb
sudo dpkg -i libnvidia-container-dev_1.17.3-1_amd64.deb
sudo dpkg -i nvidia-container-toolkit-base_1.17.3-1_amd64.deb
sudo dpkg -i libnvidia-container-tools_1.17.3-1_amd64.deb
sudo dpkg -i nvidia-container-toolkit_1.17.3-1_amd64.deb

Rebuild the docker.

FROM carlasim/carla:0.10.0
USER root

ENV NVIDIA_DRIVER_CAPABILITIES=all
ENV NVIDIA_VISIBLE_DEVICES=all

RUN apt-get update \
    && apt-get install -y \
    libxext6 \
    libvulkan1 \
    libvulkan-dev \
    vulkan-tools

# You must copy the one in your local computer into current directory.
# cp /usr/share/vulkan/icd.d/nvidia_icd.json .
COPY nvidia_icd.json /etc/vulkan/icd.d
USER carla

Run

docker run --privileged --gpus all \
  --net=host -e DISPLAY=$DISPLAY  \
  -v /tmp/.X11-unix:/tmp/.X11-unix:rw \
  carlasim/carla:0.10.0-fixed /bin/bash \
  CarlaUnreal.sh \
  -windowed -ResX=600 -ResY=600 --carla-rpc-port=22912 -quality-level=Low