Vision-based Lane Keeping Module

Requirement

• Python 3.X
• Ubuntu 18.04
• ROS Melodic
• Vision Sensor (from SVL or Real Camera)
• Actuator (from SVL or Real Car)

Installation

sudo apt-get install python3-pip -y
pip3 install rospkg scikit-build opencv-python

sudo apt-get install ros-melodic-rosbridge-suite ros-melodic-autoware-msgs ros-melodic-nmea-msgs -y

cd svl/setup
python3 -m pip install -r requirements.txt --user .

Usage with SVL Simulator

1. Download the SVL Simulator and execute simulator.

2. Click Open Browser, which will automatically show wise.svlsimulator.com website. Sign up and sign in with your account.

3. Navigate to the Clusters tab and make your cluster.

4. Navigate to the Vehicles tab and click Lexus2016RXHybrid. Click buttons next to Sensor Configurations. Clone Autoware AI configuration to make new sensor configuration. Make sure bridge is set to ROS.

5. Change configuration of Main Camera as below and save. You can copy asset ID when you click ID icon.

6. Navigate to the Simulations tab and click Add New button. Fill the blank as below.

   • Simulation Name: As your wish
   • Cluster: a cluster you made in 3
   • Runtime Template: API Only

7. Run rosbridge node

   roslaunch rosbridge_server rosbridge_websocket.launch
   

8. Click Run Simulation.

9. Change ASSET_ID in API script (svl/script/1_cubetown_base.py) into value you copied in 5 and run the script. If the count of topic increases in bridge tab (plug icon), SVL connected with bridge successfully.

   cd svl/script
   python3 1_cubetown_base.py
   

10. Copy config/config-svl.yaml to config/config.

11. Run LKAS node.

    python3 lkas.py
    python3 vehicle_cmd_publisher.py
    

Demo