In the ROS node cone_sorter.py, the system maps cone coordinates to a global frame and uses a threshold to prevent multiple registrations of the same cone. It ignores any new cone detections within 1 meter of an already registered cone. This functionality can be found here: cone_sorter.py#L28.
The underlying assumption here is that the upstream cone detection algorithm accurately identifies cone coordinates as soon as they appear in the system's view. However, this is obviously not always true, the detection accuracy actually improves as the vehicle approaches the cone. Due to this gradual increase in accuracy, the initial coordinates recorded can be significantly off, leading to inaccuracies in the path planner's outputs.
In the ROS node
cone_sorter.py, the system maps cone coordinates to a global frame and uses a threshold to prevent multiple registrations of the same cone. It ignores any new cone detections within 1 meter of an already registered cone. This functionality can be found here: cone_sorter.py#L28.The underlying assumption here is that the upstream cone detection algorithm accurately identifies cone coordinates as soon as they appear in the system's view. However, this is obviously not always true, the detection accuracy actually improves as the vehicle approaches the cone. Due to this gradual increase in accuracy, the initial coordinates recorded can be significantly off, leading to inaccuracies in the path planner's outputs.