Scenarios

Scenarios

Scenario files are stored in the root scenarios folder. These files are used to create a plan/describe a behaviour for data collection. Each file is comprised of four sections, each detailing a separate component of the data collection process. Once these sections are defined the run_scenario.launch can be used to execute the plan.

Each section is documented below and complete file example is presented at the bottom.

Please refer to default_config.yaml for an implementation example.

1) Context

The context field is a string of any length. This parameter can be thought of as a unique identifier to describe the type/location/time of your data collection plan.

context: "context-here"

When storage method is filesystem, all collected data is placed under a with the same name as context. When storage method is database the data is stored in a collection of name context.

2) Storage

The storage.method field can be one of two possible strings "database" or "filesystem". This parameter tells the data logger where to store the data.

If "filesystem" is set, data will be stored in the location defined by the storage.location parameter. This mode is useful for offline use, where a database isn't available.

A utility provided in convert.py to migrate the data to a database later on.

storage: 
    method: "filesystem"
    location: "default"

If "database" is set, data will be stored in the database under a collection named context. The database connection config must be specified so, a URI can be inferred from it the default is default_db_config.yaml.

storage: 
    method: "database"
    config: "default" 

3) Data

The "data" field is a n level dictionary. The values of the keys in this dictionary will be stored in the exported data. If the value is a string that is also currently a published topic, then the value will be populated with the respective ros topic information. In the example below the "/clock" topic information is stored in a ros_clock_topic container.

data: {
    int_example: 5,
    float_example: 4,
    str_example: "ID: example",
    list_example: ["hello", "world"],
    ros_clock_topic: "/clock",
}

A more complicated example below will group localisation data from multiple robot topics in a single localisation container.

data: { # Store the ros topic information in this key value container structure
    localisation: {
        laser_scan: "/scan",
        current_edge: "/current_edge",
        current_node: "/current_node",
        closest_node: "/closest_node",
        amcl: "/amcl_pose",
        robot_pose: "/robot_pose",
    },
}

4) Collection

This is the most complex parameter since it describes the data collection behaviour. The "collection" field is a dictionary containing some required and optional parameters depending upon which parameter is set for name.

The method sub-field defines the exploration behaviour. There are five possible options described below:

• action_server: Will wait until actionlib requests are received on collection.action_server_name. To test you can run to get a graphical request system rosrun actionlib axclient.py '''/collect_data'''.

collection:
  method: "action_server"
  action_server_name: "collect_data"

• action_server_video: Will wait until actionlib requests are received on collection.action_server_name with a request to start/stop data collection. In the action server request you can specify start|stop|true|false|t|f to start or stop capturing data every time wait_topic publishes a new message. The wait_topic field is string that must be a valid rostopic. To test you can run to get a graphical request system rosrun actionlib axclient.py '''/collect_data'''.

collection:
  method: "action_server_video"
  action_server_name: "collect_data"
  watch_topic: "/realsense_camera/color/image_raw"

• timer: Captures data every timer_delay seconds. The timer_delay field is integer. Ideal for when the robot is in continuous operation in unmapped environments.

collection:
    method: "timer"
    timer_delay: 3

• event: Captures data every time wait_topic publishes a new message. The wait_topic field is string that must be a valid rostopic. Ideal for capturing video sequences (of rgb frames for example)

collection:
    method: "event" 
    watch_topic: "/realsense_camera/color/image_raw"

Example file

The below yaml file is a complete example which will capture camera and robot odometry data when a request is received.

context: "documentation_example"

storage: 
    method: "filesystem"
    location: "default"

data: {
    localisation: {
      laser_scan: "/scan", current_edge: "/current_edge", current_node: "/current_node",
      closest_node: "/closest_node", amcl: "/amcl_pose", robot_pose: "/robot_pose"
    },
    rgb: {
      image: "/doc_cam/color/image_raw", intrinsics: "/doc_cam/color/camera_info"
    },
}

collection:
    method: "action_server"
    action_server_name: "collect_data"