Autoware_class20

Self-Driving Cars with ROS 2 and Autoware

Lectures

MDs and PDF slides from here
Videos here

Building ADE

Requirments

Ubuntu 18.04
Docker.ce, installion.
Nvidia-container-runtime.

Installation

Follow installtion here

Notes

After docker installtion, docker and ade start command should be started with root privileges so you should do:

sudo groupadd docker 
sudo usermod -aG docker $USER
reboot

for more info, Check this link

Disable secure boot from BIOS -> to avoid problems to Nvidia dirivers installtion
Do unset ADE_DISABLE_NVIDIA_DOCKER and delete ADE_DISABLE_NVIDIA_DOCKER line from ~/AutowareAuto/.aderc -> to avoid Rviz problems

Demo

First implemention of ADE by checking 3D perception stack

lecture 1

Autonomous Stack

Autoware.Ai is like the skelton for Autowar.Auto. It contains the main software algorithms but supports ROS1.

Autowar.Auto is the next generation with more robust, clean and reliable code and supports ROS2.

LGSVL simulator with ROS2 and Autoware.auto

First LGSVL simulator implmentation with ADE

Note

Bridging with Autoware.Auto

LGSVL uses conventions which are not directly aligned with ROS 2 conventions. The full list of behaviors the lgsvl_interface implements is:

Converts control inputs with CCW positive rotations to the CCW negative inputs the LGSVL simulator expects
Provides a mapping from VehicleControlCommand to the RawControlCommand LGSVL expects via parametrizable 1D lookup tables To run the lgsvl_interface, enter the following in a new terminal window:

$ ade enter
ade$ source /opt/AutowareAuto/setup.bash
ade$ ros2 run lgsvl_interface lgsvl_interface_exe __params:=/opt/AutowareAuto/share/lgsvl_interface/param/lgsvl.param.yaml

Object Detection Stack

Autoware.Auto uses a classical object detection stack:

Simpler and more robust
Geometric and optimization-based methods for: ○ Ground filtering ○ Clustering ○ Shape Extraction

Test

Documentation: -https://autowarefoundation.gitlab.io/autoware.auto/AutowareAuto/lgsvl.html

https://www.lgsvlsimulator.com/docs/vehicles-tab/#how-to-add-a-vehicle
lgsvl-sensors.json file for vehicle configuration : https://gitlab.com/autowarefoundation/autoware.auto/AutowareAuto/-/blob/master/lgsvl-sensors.json

Running command for Object detection: ade$ source /opt/AutowareAuto/setup.bash ade$ ros2 launch autoware_demos
lidar_bounding_boxes_lgsvl.launch.py

Note:

For the lgsvl bridge with ROS2 connection, ros2-lgsvl-bridge package should be cloned and build inside ~/adehome and then restart ADE

Camera Perception - Lane detection and Ground Projection

Hands-on lab on Camera perception to detect and Project road lanes using ROS2 and Autoware ADE

Repo: https://gitlab.com/ApexAI/autowareclass2020/-/tree/master/code/src/08_Perception_Camera

Test:

enter ADE

ade start
ade enter

go to project repo
source install/setup.bash
sudo apt-get install libopencv-dev (optionally, if you don't have already libopencv-dev installed )
open 4 terminals ( inside ade ) and source install/setup.bash ( inside project repo)
terminal 1 : init. data_loader node ros2 run lane_detection_data_loader lane_detection_data_loader_node
terminal 2 : init. projection_node ros2 run lane_detection_projection lane_detection_projection_node
terminal 3 : init. visualization_node ros2 run lane_detection_visualization lane_detection_visualization_node
terminal 4 : open RViz rviz2

then setup RViz to visualize image and marker topics

AmarYasser/Autoware.auto-class2020