Lane detection

[jruths]

We need a node to process sensing data (image and maybe lidar?) to produce individual lanes.

• Ideally these lanes would be annotated with lanes going in the same direction as the vehicle and lanes going the opposite direction. (using yellow vs white lines)
• Using road markings for turn/straight arrows to indicate other information about the lane would be interesting too.

This node may be quite related to the Driveable surface segmentation node.

[stepan-antonyuk]

09/09/2024 Update:

• Status: Researched lane detection algorithms
• Next Steps: I will start implementing a lane detection algorithm outside of the navigator to check if it is suitable to use for us
  • This is the algorithm I will be implementing first: https://github.com/matang28/ros-lane-detection
• Projected Completion: September 21, 2024
• Update: I have read through many algorithms and picked the ones that work with our equipment, fit our task, and give quality output. Below, you can see the algorithms I selected with some information about them.
  • Using a camera or stereo camera
    • https://github.com/isarlab-department-engineering/ros_dt_lane_follower?tab=readme-ov-file
      • Already uses ros.
      • Uses python.
      • It looks for white, yellow, and red lines in the image.
      • return error in pixel from the middle of the lane.
      • I am not sure how fast it is.
    • (My favorite) https://github.com/matang28/ros-lane-detection
      • Explains usage in ros.
      • Uses python
      • It can find curved lanes.
      • Calculates departure from the center.
      • Publishes to ros topics.
      • Can set threshold(from the center of the road).
      • I am not sure how fast it is.
    • https://github.com/AbangLZU/LaneNetRos
      • This is a huge project, so it may be hard to integrate into our stack.
      • Doesn't include ros implementation
      • Uses python
      • Uses model to detect lanes.
      • Allows to train the model.
      • I am not sure how fast it is.
    • https://github.com/ragibarnab/ros2-semseg-lane-detection?tab=readme-ov-file
      • This is WIP, but it can help understand how lane detection works
      • Already uses ros2
      • I am not sure how fast it is.
  • Using Lidar. There are some algorithms, but I would prefer working with images
    • https://github.com/kaist-avelab/K-Lane
    • (This one is interesting) https://github.com/kwh950724/lidar_lane_detector
  • While looking through the lane detection algorithm, found lane detection method for lane level localization
    • Implementing lane detection method for lane level localization purposes autowarefoundation/autoware#4702

[stepan-antonyuk]

09/16/2024 Update:

• Status: Implemented lane detection algorithms on local computer
• Next Steps: Make sure that the program is good for navigator use and implements a simple ros node
• Projected Completion: September 25, 2024
• Update: Have a working code that finds lanes and calculates displacement from the center, which is useful for the PID controller

[stepan-antonyuk]

09/23/2024 Update:

• Status: Waiting for the camera to be installed on the car to start adjusting parameters in the code.
• Next Steps: Adjust parameters once the camera is installed, and implement a simple ROS node.
• Projected Completion: September 30, 2024
• Update: I reviewed the code and understood how it works. Some intermediate information can be extracted and could be useful for other tasks. I currently have a working code and will begin adjustments after the camera is installed. I also started looking into other algorithms and repositories that could be useful for multiple lane detection and intersection management.

[stepan-antonyuk]

09/30/2024 Update:

• Status: Waiting for the camera to be installed on the car to start adjusting parameters in the code.
• Next Steps: Adjust parameters once the camera is installed, and implement a simple ROS node.
• Projected Completion: October 5, 2024 (depending on the camera installation time)
• Update: I have been experimenting with the code and now have a good understanding of how it works and how to extract useful intermediate information. ROS node integration will begin once the camera is installed and parameter adjustments are complete.