Frame Transforms

The extension produces foxglove.FrameTransforms messages that define the coordinate frame hierarchy for the ego vehicle. These frames are used by Lichtblick's 3D panel to correctly position camera views and sensor data relative to the vehicle.

Frame hierarchy

global
  └── ego_vehicle_bb_center        (vehicle bounding box center in world coords)
        └── ego_vehicle_rear_axle  (rear axle, offset from BB center)
              └── virtual_mounting_position  (SensorData only: sensor mount point)

Frame names

Frame	Constant	Description
global	OSI_GLOBAL_FRAME	World coordinate system (root frame)
ego_vehicle_bb_center	OSI_EGO_VEHICLE_BB_CENTER_FRAME	Center of the ego vehicle's bounding box
ego_vehicle_rear_axle	OSI_EGO_VEHICLE_REAR_AXLE_FRAME	Rear axle position (translated from BB center)
virtual_mounting_position	OSI_SENSORDATA_VIRTUAL_MOUNTING_POSITION_FRAME	Sensor mounting position (SensorData only)

Transforms by message type

GroundTruth / SensorView

Produces 1–2 transforms:

1. Global → BB Center: Uses host_vehicle.base.position and host_vehicle.base.orientation
2. BB Center → Rear Axle (conditional): Uses host_vehicle.vehicle_attributes.bbcenter_to_rear (pure translation in body frame). Only produced if bbcenter_to_rear is present.

SensorData

Produces 1 transform:

• Rear Axle → Virtual Mounting Position: Uses mounting_position.position and mounting_position.orientation

Incomplete SensorData support

The SensorData converter does not produce the Global → BB Center and BB Center → Rear Axle transforms. This means the virtual mounting position frame has no parent chain back to global. See issue #147.

Host vehicle resolution

The host vehicle is identified by message.host_vehicle_id.value, then located in the moving_object array via a single .find() call. The resolved object is passed directly to the transform builder functions.

Fallback chain:

1. Use GroundTruth.host_vehicle_id if present
2. Fall back to SensorView.host_vehicle_id if the GroundTruth value is missing
3. Warn if both are present but disagree (divergence detection)
4. Return empty transforms if neither source provides an ID

All warnings are emitted via emitAlert() so they appear in the Lichtblick panel UI.