The value of a system is relative to whether you actually use it. The more annoying you make it, the less likely customers are to use it, so you lose out on any safety benefits of using it. Overly aggressive eye tracking is an annoyance, while non-aggressive tracking... what's the advantage of that over wheel torque?
IMHO, any reasonable autonomy system evaluation should be simply the formulae:
Satisfaction Score = O
Safety Score = P * S
... where O is surveyed user satisfaction, P is the percentage of time a user is likely to have it engaged, and S is how much safer they're likely to be while having it engaged (can be negative). Using any other formula to evaluate autonomy systems seems nonsensical to me. You could try to come up with formulae to estimate P and S, but IMHO the best way is just to collect real-world data on both (the best data on S would have to come from the manufacturer, although you could approximate it independently by looking at how frequently the driver felt the need to take over from the system, plus how frequently the system disengaged on its own, with each multiplied by a risk factor for measured driver inattentiveness at the moment the user took control)
Last edited:
