2024-08-15 Racing Working Group Meeting

[pojenwang]

Administrative

Previous Meeting Minutes

Attendees

• Po-Jen Wang (AWF)
• Mitsudome-san (Tier.IV)
• David Walmroth (Open AD Kit / Pix-Moving)
• Atanasko

Minutes: Po-Jen Wang

Topics

Improved Autoware on F1tenth

• The current Autoware on F1tenth runs reduced Autoware stacks. Our goal is to expand the usage of Autoware packages for F1tenth and introduce mission planning and scenario planning so users can run more complicated scenarios (ex. Parking scenario)
• Utilizes the Autoware Rviz UI instead of F1tenth Rviz UI
• Since many F1tenth cars only use 2D Lidar instead of 3D Lidar, a tool was developed to convert the F1tenth 2D occupancy map into 3D pointcloud for setting up the vector map and obstacle visualization in Autoware Rviz UI
• Map converter
  -- Converts the .png occupancy map generated from slam_toolbox into the .pcd pointcloud data format
  -- The occupancy map has well-defined obstacle space(black), free space(white), undiscovered space(gray)
  -- Uses the light-weight libpng package to read the png image instead of the heavy OpenCV
  -- Uses the pcl package to handle all of the pointcloud operations
  -- The generated pointcloud comes with two components
  --- Ground plane: Down-sampled (configurable) and cover the entire map at z = 0. Required by the Autoware tool to generate the lanelet2 vector map (lanelet needs to be on or close to pointcloud data)
  --- Obstacle walls: Extrude the obstacle cells in the z-direction with user-defined height and density
  
• The pointcloud map is used to generate vector map (lanelet) and being imported into Autoware's Rviz UI for obstacle visualization
  
• top-down view in Autoware Rviz UI
  
• angled view in Autoware Rviz UI
  
• Setup the f1tenth_vehicle_description package with the f1tenth vehicle parameters. The vehicle 3D model in Rviz needs to be updated from the default Lexus model to the F1tenth RC car.
• Next steps
  -- Integrate F1tenth's Particle filter into Autoware's localization pipeline(replace ndt_scan_matcher). Merge the UI with Autoware UI (convert the real-time 2D scan into pointcloud and overlay them onto pointcloud map)
  -- Integrate and test Autoware planning pipeline

Autoware Training Vehicle (1/5 scale) Update

• A new simulator environment with uneven terrain is developed for off-road vehicle testing
• A temporary vehicle model is used and integrated into SODA sim. A detailed model integration is in progress
  

Other topics:

• AWF Work Group overview: https://autoware.org/join-a-work-group/