There are several levels of abstraction involved in piloting and driving.
- Signals representing the state of the vehicle and its environment originating from multiple transducers1
- Latched sample vectors/matrices
- Boundary events (locations, spectral features, movement, appearance and disappearance of edges, lines, and sounds)
- Objects
- Object movements
- Object types (runways, roads, aircraft, birds, cars, people, pets, screeches, horns, bells, blinking lights, gates, signals, clouds, bridges, trains, buses, towers, antennas, buildings, curbs)
- Trajectory probabilities based on object movements and types
- Behaviors based on all the above hints
- Intentions based on behavior sequences and specific object recognition
- Collision risk detection
Moving from interpretation to control execution ...
- Preemptive collision avoidance reaction
- Horn sounding
- Plan adjustment
- Alignment of plan to state
- Trajectory control
- Skid avoidance
- Skid avoidance reaction
- Steering, breaking, and signalling
- Notifications to passengers
What, if any, levels of higher abstraction can be sacrificed? Humans, if they are excellent pilots or drivers, can use all of these levels to improve pedestrian and passenger safety and minimize expense in time and money.
Footnotes
[1] Optical detectors, microphones, strain gauge bridges, temperature and pressure gauges, triangulation reply signals, voltmeters, position encoders, key depression switches, flow detectors, altimeters, radar transducers, tachometers, accelerometers
