Autonomy Levels (my thoughts)
Safety interventions seem self-explanatory (any required human interventions to prevent an accident).
Nuisance interventions (car gets stuck somewhere / stranded, needlessly delays traffic, too hesitant, can't find parking garage exit...) These may be one of the most difficult problems of all considering all of the possible edge cases that seem to require human levels of cognition)
Level 1: Works with Supervision:
- Seems easy (relatively
to achieve
- Safety interventions < 1 / 100 miles (although Elon simplifying it to red lights and stop signs gives me pause).
- Nuisance interventions < 1 / 100 miles
Level 1.5 - Works very well for 10k -50k miles at a stretch (this becomes a dangerous area when humans will just figure it's perfect because it never seems to make a mistake and stop supervising).
Level 2: Unsupervised
- Much more difficult
- Safety interventions < 1 / 5M miles (massive leap)
- Nuisance interventions < 1 / 100 miles ???
Level 3 :Regulator approved
- Technologically no change from Level 2
Level 4 : No driver (this is obviously required for TAAS, and may be surprisingly difficult)
- Safety Interventions < 1 / 5-10M miles
- Nuisance interventions < 1 / 10k miles ?? This is an area Waymo has found surprisingly difficult, and I am slightly concerned Elon has not addressed). I'm just guessing 1 / 10k miles. Even this often occurring would still annoy others (hopefully not too much) but with remote operators to help out maybe this many nuisance interventions required would allow deployment?
Thoughts?
