AWSIM
AWSIM copied to clipboard
tier4
Docker container for AWSIM
Is there any plans to package the simulator inside a docker container?
It would be great if the simulator can be run as a container, especially together with prebuilt containers of Autoware.
Thanks for the suggestion. Not planned yet, but a good idea. I too would like to simplify the environment building process. Please give me some time to consider if I'm missing something.
I tried running running AWSim from within the autoware docker container, but was not able to get it to work. See this autoware discussion post for more details.
Is the recommended way to run autoware + AWSim to install autoware from source rather than running inside of a docker container?
@tleyden Sorry for the delay in replying.
Is the recommended way to run autoware + AWSim to install autoware from source rather than running inside of a docker container?
Yes, That is correct. We have not yet checked out running AWSIM with Docker. If you follow this tutorial, you should be able to run AWSIM in a stable. https://tier4.github.io/AWSIM/GettingStarted/QuickStartDemo/
@doganulus @tleyden Could you try this? https://github.com/tier4/AWSIM/issues/81
Why does #81 need to install AWSIM on an Autoware container? Why not an Ubuntu base ROS installed (and its other dependencies if any)?
@mackierx111 Could you please open this issue again?
The issue #81 is not the solution for the feature here. This is about packaging the AWSIM as a container.
I just find a way to run AWSIM in docker container: (Much thanks for the great job done by rickyjames35 to solve the vulkan problem)
-
Download AWSIM and unzip the downloaded file
-
Create an image containing both vulkan and ros humble with the follow dockerfile
FROM ros:humble-ros-base-jammy
# Needed to share GPU
ENV NVIDIA_DRIVER_CAPABILITIES=all
ENV NVIDIA_VISIBLE_DEVICES=all
RUN apt-get update && \
export DEBIAN_FRONTEND=noninteractive && \
apt-get -y install \
pciutils \
vulkan-tools \
mesa-utils \
ros-humble-rmw-cyclonedds-cpp
RUN echo "export RMW_IMPLEMENTATION=rmw_cyclonedds_cpp" >> ~/.bashrc
- Run a container with volume containing unzipped AWSIM folder
docker run --gpus all --runtime=nvidia -it --rm -e DISPLAY -v /tmp/.X11-unix:/tmp/.X11-unix -v <path to you AWSIM folder>:/AWSIM <your image name> /bin/bash
- In container, launch AWSIM
/AWSIM/AWSIM_demo.x86_64
I just find a way to run AWSIM in docker container: (Much thanks for the great job done by rickyjames35 to solve the vulkan problem)
- Download AWSIM and unzip the downloaded file
- Create an image containing both vulkan and ros humble with the follow dockerfile
FROM ros:humble-ros-base-jammy # Needed to share GPU ENV NVIDIA_DRIVER_CAPABILITIES=all ENV NVIDIA_VISIBLE_DEVICES=all RUN apt-get update && \ export DEBIAN_FRONTEND=noninteractive && \ apt-get -y install \ pciutils \ vulkan-tools \ mesa-utils \ ros-humble-rmw-cyclonedds-cpp RUN echo "export RMW_IMPLEMENTATION=rmw_cyclonedds_cpp" >> ~/.bashrc
- Run a container with volume containing unzipped AWSIM folder
docker run --gpus all --runtime=nvidia -it --rm -e DISPLAY -v /tmp/.X11-unix:/tmp/.X11-unix -v <path to you AWSIM folder>:/AWSIM <your image name> /bin/bash
- In container, launch AWSIM
/AWSIM/AWSIM_demo.x86_64
I am very happy to see this solution, but I tried it on wsl2 of win11, and there is no error, but it also failed to start. May I ask what environment you are using it in?
I just find a way to run AWSIM in docker container: (Much thanks for the great job done by rickyjames35 to solve the vulkan problem)
- Download AWSIM and unzip the downloaded file
- Create an image containing both vulkan and ros humble with the follow dockerfile
FROM ros:humble-ros-base-jammy # Needed to share GPU ENV NVIDIA_DRIVER_CAPABILITIES=all ENV NVIDIA_VISIBLE_DEVICES=all RUN apt-get update && \ export DEBIAN_FRONTEND=noninteractive && \ apt-get -y install \ pciutils \ vulkan-tools \ mesa-utils \ ros-humble-rmw-cyclonedds-cpp RUN echo "export RMW_IMPLEMENTATION=rmw_cyclonedds_cpp" >> ~/.bashrc
- Run a container with volume containing unzipped AWSIM folder
docker run --gpus all --runtime=nvidia -it --rm -e DISPLAY -v /tmp/.X11-unix:/tmp/.X11-unix -v <path to you AWSIM folder>:/AWSIM <your image name> /bin/bash
- In container, launch AWSIM
/AWSIM/AWSIM_demo.x86_64I am very happy to see this solution, but I tried it on wsl2 of win11, and there is no error, but it also failed to start. May I ask what environment you are using it in?
I run it on ubuntu 2204 and not test on wsl2
I just find a way to run AWSIM in docker container: (Much thanks for the great job done by rickyjames35 to solve the vulkan problem)
- Download AWSIM and unzip the downloaded file
- Create an image containing both vulkan and ros humble with the follow dockerfile
FROM ros:humble-ros-base-jammy # Needed to share GPU ENV NVIDIA_DRIVER_CAPABILITIES=all ENV NVIDIA_VISIBLE_DEVICES=all RUN apt-get update && \ export DEBIAN_FRONTEND=noninteractive && \ apt-get -y install \ pciutils \ vulkan-tools \ mesa-utils \ ros-humble-rmw-cyclonedds-cpp RUN echo "export RMW_IMPLEMENTATION=rmw_cyclonedds_cpp" >> ~/.bashrc
- Run a container with volume containing unzipped AWSIM folder
docker run --gpus all --runtime=nvidia -it --rm -e DISPLAY -v /tmp/.X11-unix:/tmp/.X11-unix -v <path to you AWSIM folder>:/AWSIM <your image name> /bin/bash
- In container, launch AWSIM
/AWSIM/AWSIM_demo.x86_64
Recently tried with this Dockerfile and docker arguments and it is failing, also with autoware-universe:prebuilt image and it is the same. Has anyone tried again nowadays ?
I would like to have a standalone AWSIM container to work as a sidecar with Autoware runtime container and perhaps with others. My latest container was:
FROM ubuntu:jammy as builder
ARG VERSION=v1.2.0
ARG AWSIM_VERSION=$VERSION
ARG AWSIM_RELEASE_PATH=https://github.com/tier4/AWSIM/releases/download/${AWSIM_VERSION}/AWSIM_${AWSIM_VERSION}.zip
ARG AWSIM_ADDITIONAL_MAPS=https://github.com/tier4/AWSIM/releases/download/${AWSIM_VERSION}/nishishinjuku_autoware_map.zip
RUN apt-get update && export DEBIAN_FRONTEND=noninteractive \
&& apt-get -y install \
wget \
unzip \
&& apt-get autoremove -y && apt-get clean -y && rm -rf /var/lib/apt/lists/*
RUN wget -qO /tmp/awsim_release.zip ${AWSIM_RELEASE_PATH} && \
unzip /tmp/awsim_release.zip -d /awsim && \
rm /tmp/awsim_release.zip
FROM nvidia/cuda:12.2.0-runtime-ubuntu22.04 as runtime
ARG VERSION=v1.2.0
ARG AWSIM_VERSION=$VERSION
# RUN apt-get update && export DEBIAN_FRONTEND=noninteractive \
# && apt-get -y install \
# sudo \
# && apt-get autoremove -y && apt-get clean -y && rm -rf /var/lib/apt/lists/*
# RUN groupadd awsim -g 1000 \
# && useradd -ms /bin/bash awsim -g 1000 -u 1000 \
# && printf "awsim:awsim" | chpasswd \
# && printf "awsim ALL= NOPASSWD: ALL\\n" >> /etc/sudoers
RUN apt-get update && export DEBIAN_FRONTEND=noninteractive \
&& apt-get -y install \
libvulkan1 \
x11-apps \
x11vnc \
xvfb \
&& apt-get autoremove -y && apt-get clean -y && rm -rf /var/lib/apt/lists/*
COPY --from=builder --chown=root:root /awsim/AWSIM_${AWSIM_VERSION} /opt/awsim
# USER awsim
# WORKDIR /home/awsim/
ENV DISPLAY :0
ENV XDG_RUNTIME_DIR /tmp/runtime-user
ENV PULSE_SERVER unix:/run/pulse/native
ENV RMW_IMPLEMENTATION=rmw_cyclonedds_cpp
CMD ["/opt/awsim/AWSIM_demo.x86_64"]
It builds but couldn't test it recently with a computer with Nvidia hardware. Last time I had got segmentation faults. I think some initialization work is needed for x11 forwarding mechanisms in the eventual entrypoint.sh.
I was able to run AWSIM using the following commands:
Download AWSIM and maps for AWSIM by following https://tier4.github.io/AWSIM/GettingStarted/QuickStartDemo/
git clone [email protected]:autowarefoundation/autoware
cd autoware
./setup-dev-env.sh docker
./docker/run.sh --map-path /home/mitsudome-r/autoware_map/nishishinjuku_autoware_map/ --workspace /home/mitsudome-r/Downloads/AWSIM_v1.2.0 /bin/bash
# inside docker container
cd /workspace
sudo ./AWSIM_v1.2.0.x86_64
AWSIM failed with segmentation fault if I run it without sudo. Also, I still get the following error messages (probably due to some access issues to GPU devices), but I was still able to launch it without segmentation faults.
MESA: error: Failed to query drm device.
glx: failed to create dri3 screen
failed to load driver: iris
failed to open /dev/dri/card0: No such file or directory
failed to load driver: iris
@mitsudome-r I tried the same
git clone [email protected]:autowarefoundation/autoware
cd autoware
./setup-dev-env.sh docker
./docker/run.sh --map-path /home/mitsudome-r/autoware_map/nishishinjuku_autoware_map/ --workspace /home/mitsudome-r/Downloads/AWSIM_v1.2.0 /bin/bash
# inside docker container
cd /workspace
sudo ./AWSIM_v1.2.0.x86_64
with sudo but I am getting segmentation fault.
Can you try nvidia-smi inside and outside of the container ? And also what is your exact GPU model and its driver in use ?
@oguzkaganozt Here's the information
GPU: NVIDIA GeForce RTX 3060 Mobile
nvidia-smi inside container:
Fri Mar 29 00:08:27 2024
+---------------------------------------------------------------------------------------+
| NVIDIA-SMI 545.23.08 Driver Version: 545.23.08 CUDA Version: 12.3 |
|-----------------------------------------+----------------------+----------------------+
| GPU Name Persistence-M | Bus-Id Disp.A | Volatile Uncorr. ECC |
| Fan Temp Perf Pwr:Usage/Cap | Memory-Usage | GPU-Util Compute M. |
| | | MIG M. |
|=========================================+======================+======================|
| 0 NVIDIA GeForce RTX 3060 ... On | 00000000:01:00.0 On | N/A |
| N/A 54C P8 11W / 115W | 57MiB / 6144MiB | 18% Default |
| | | N/A |
+-----------------------------------------+----------------------+----------------------+
+---------------------------------------------------------------------------------------+
| Processes: |
| GPU GI CI PID Type Process name GPU Memory |
| ID ID Usage |
|=======================================================================================|
| 0 N/A N/A 2150 G /usr/lib/xorg/Xorg 45MiB |
+---------------------------------------------------------------------------------------+
outside container:
Fri Mar 29 00:07:28 2024
+---------------------------------------------------------------------------------------+
| NVIDIA-SMI 545.23.08 Driver Version: 545.23.08 CUDA Version: 12.3 |
|-----------------------------------------+----------------------+----------------------+
| GPU Name Persistence-M | Bus-Id Disp.A | Volatile Uncorr. ECC |
| Fan Temp Perf Pwr:Usage/Cap | Memory-Usage | GPU-Util Compute M. |
| | | MIG M. |
|=========================================+======================+======================|
| 0 NVIDIA GeForce RTX 3060 ... On | 00000000:01:00.0 On | N/A |
| N/A 60C P3 17W / 115W | 57MiB / 6144MiB | 2% Default |
| | | N/A |
+-----------------------------------------+----------------------+----------------------+
+---------------------------------------------------------------------------------------+
| Processes: |
| GPU GI CI PID Type Process name GPU Memory |
| ID ID Usage |
|=======================================================================================|
+---------------------------------------------------------------------------------------+
I also tested with a machine with GeForce GTX 1070 as well, and it worked.
Could you try running the run.sh command with sudo and see if that changes anything as well?
It is now running with the new docker containers and AWSIM version 1.2.1. With these instructions:
git clone [email protected]:autowarefoundation/autoware
cd autoware
./setup-dev-env.sh docker
./docker/run.sh --map-path /home/mitsudome-r/autoware_map/nishishinjuku_autoware_map/ --workspace /home/mitsudome-r/Downloads/AWSIM_v1.2.0 /bin/bash
# inside docker container
cd /workspace
sudo ./AWSIM_v1.2.0.x86_64
Our solution is here: https://github.com/bounverif/containers/tree/main/awsim-desktop
It is based on ghcr.io/selkies-project/nvidia-egl-desktop.