WATonomous Rosbags

Writing down a list of rosbags that we can use.

All of the rosbags in ROS2 are located under /mnt/wato-drive2 (you will see these if you SSH into Trpro, Thor, or any other of our server cluster machines). The ones we are interested in are:

• /mnt/wato-drive2/nuscenes_mcap/ros2bag - contains rosbags from the Nuscenes dataset
• /mnt/wato-drive2/rosbags2 - contains our own rosbags that we have recorded in the past with Bolty, our previous car

These are mounted for you automatically inside the data_stream Docker container:

  data_stream:
    build:
      context: ..
      dockerfile: docker/infrastructure/data_stream/data_stream.Dockerfile
      cache_from:
        - "${INFRASTRUCTURE_DATA_STREAM_IMAGE:?}:${TAG}"
        - "${INFRASTRUCTURE_DATA_STREAM_IMAGE:?}:main"
      target: deploy
    image: "${INFRASTRUCTURE_DATA_STREAM_IMAGE:?}:${TAG}"
    volumes:
      - /mnt/wato-drive2/nuscenes_mcap/ros2bags:/home/bolty/ament_ws/nuscenes
      - /mnt/wato-drive2/rosbags2:/home/bolty/ament_ws/rosbags2

Feel free to explore the rosbags.

Commands

Bring up the data_stream and foxglove container by running watod up data_stream foxglove, and you can play data by running some of the commands below:

Streaming through Foxglove

If you are streaming data through Foxglove over ssh, you should lower the speed at which the rosbag is being played to ~0.2. There is a lot of data being transmitted, and Foxglove will end up dropping lots of messages.

For Nuscenes

watod run data_stream ros2 bag play -l ./nuscenes/NuScenes-v1.0-mini-scene-0061/NuScenes-v1.0-mini-scene-0061_0.mcap

For our bags

The main juicy bags are the rosbags2/year3/test_track_days/W20.4/traffic_light1 to traffic_light4

Very good overview of the traffic lights, and various signs in Waterloo roads → good for Traffic Sign + Traffic Light Detection

ros2 bag play -l ./rosbags2/year3/test_track_days/W20.4/traffic_light1

Contains the following topic

• `/camera/center/image_color
• /camera/left/image_color
• /camera/right/image_color

ros2 bag play -l rosbags2/year3/test_track_days/W20.4/traffic_light2

ros2 bag play rosbags2/year3/test_track_days/W20.4/traffic_light3

ros2 bag play rosbags2/year3/test_track_days/W20.4/traffic_light4

Car driving through the roads (1 camera, 1 Lidar) → good for Road Segmentation

 ros2 bag play --start-offset 50 ./rosbags2/year3/test_track_days/W20.0/public_road_to_TT

• /camera/right/image_color
• /velodyne_points

There are some more test_track_days, W20.1 and W20.2 that seem really good for like cyclist and pedestrian datasets, not sure if it is going to work super well though.

People holding a traffic light that switches between different states (1 camera, LiDAR) → good for Traffic Light Detection

ros2 bag play -l ./rosbags2/year3/test_track_days/aug_10/Compressed/traffic_light_ped_odom/traffic_light_ped_odom.db3

• /camera/right/image_color
• /lidar_merged_visualized
• Question traffic_light_ped vs. traffic_light_ped_odom, are they the same?