Skip to content

Latest commit

 

History

History
116 lines (86 loc) · 3.93 KB

awsim_simulation.md

File metadata and controls

116 lines (86 loc) · 3.93 KB
Raw

AWSIM simulation

A. Preparation

Before trying AWSIM simulation, you have to once exit the Docker container by exit command.

Required PC specs

  • CPU: 6cores and 12thread or higher
  • GPU: RTX2080Ti or higher

Download AWSIM binary

sudo apt update
sudo apt install libarchive-tools -y
cd $HOME
wget -qO- https://github.com/tier4/AWSIM/releases/download/v1.1.0/AWSIM_v1.1.0.zip | bsdtar -xvf-
chmod +x ./AWSIM_v1.1.0/AWSIM_demo.x86_64

Download a map

wget -qO- https://github.com/tier4/AWSIM/releases/download/v1.1.0/nishishinjuku_autoware_map.zip | bsdtar -xvf-
mv nishishinjuku_autoware_map ~/autoware_map

B. Run simulation

Terminal 1

Launch a Docker container Note that when you first launch docker container, it will take 3 hours.

rocker --nvidia --x11 --user --privileged --network=host --volume $HOME/autoware_map --volume $HOME/AWSIM_v1.1.0 --volume /tmp -- ghcr.io/tier4/online:humble-awsim-stable-prebuilt-cuda

Launch Autoware

export ROS_LOCALHOST_ONLY=1
export RMW_IMPLEMENTATION=rmw_cyclonedds_cpp
export RCUTILS_COLORIZED_OUTPUT=1

source /autoware/install/setup.bash
ros2 launch autoware_launch e2e_simulator.launch.xml \
vehicle_model:=sample_vehicle \
sensor_model:=awsim_sensor_kit \
map_path:=$HOME/autoware_map/nishishinjuku_autoware_map

If you feel it consumes memory too much, you can turn off Displays > Map > PointCloudMap on Rviz panel. map_pointcloud_off

You can see the map appears.

Terminal 2

Launch a Docker container

rocker --nvidia --x11 --user --privileged --network=host --volume $HOME/autoware_map --volume $HOME/AWSIM_v1.1.0 --volume /tmp -- ghcr.io/tier4/online:humble-awsim-stable-prebuilt-cuda

Launch AWSIM

export ROS_LOCALHOST_ONLY=1
export RMW_IMPLEMENTATION=rmw_cyclonedds_cpp
export RCUTILS_COLORIZED_OUTPUT=1

cd $HOME
source /autoware/install/setup.bash
./AWSIM_v1.1.0/AWSIM_demo.x86_64

You can see unity is launched.

C. Enjoy self-Driving simulation

  1. Launch AWSIM and Autoware according to the steps described earlier in this document.

  2. Move "Time Scale" bar squared by pink in the picture from × 1.00 to × 0.50 in order to reduce rendering delay.

  3. The Autoware will automatically set its pose estimation as presented below.

  4. Set the navigation goal for the vehicle.

  5. Optionally, you can define an intermediate point through which the vehicle will travel on its way to the destination. The generated path can be seen on the image below.

  6. Enable self-driving.

    Launch a Docker container on the another terminal.

    rocker --nvidia --x11 --user --privileged --network=host --volume $HOME/autoware_map --volume $HOME/AWSIM_v1.1.0 --volume /tmp -- ghcr.io/tier4/online:humble-awsim-stable-prebuilt-cuda

    Now you can start the ego vehicle driving by clicking the AUTO button on OperationMode in AutowareStatePanel. Alteratively, you can manually start the vehicle by running the following command:

    export ROS_LOCALHOST_ONLY=1
export RMW_IMPLEMENTATION=rmw_cyclonedds_cpp
export RCUTILS_COLORIZED_OUTPUT=1
source /autoware/install/setup.bash
ros2 topic pub /autoware/engage autoware_auto_vehicle_msgs/msg/Engage '{engage: True}' -1

    Finally, the self-driving simulation should start!