How Drunk Is Your Car

The criteria used to evaluate driver performance in this Australian study of the effects of alcohol on drivers might be usefully applied to autonomous car candidates as a kind of “car-ing” test(Oh gawd):

http://www.tac.vic.gov.au/road-safety/statistics/summaries/drink-driving-statistics/effects-of-alcohol

I think we can give rough ratings on each of the criteria based on the current state of the art of sober human driving and automated driving.

Divided Attention

For fully autonomous vehicles this should be no problem, since the human occupant could devote all their attention to tasks not related to driving. For vehicles less than fully autonomous I think this might be a worse problem than it is for manual drivers in some cases, but the stated case of looking for house numbers would, of course, be a non-issue for a computer-guided car.

Vigilance

Humans are pretty terrible at this, and automatic braking systems are already saving lives.

Tracking

Humans do better at this than computers in non-standard environments, but computers really love concentrating on one thing.

Perception

Autonomous cars are probably hiding their deficiency in this with better reaction times. Computers will probably get better at this, but I think they will remain inferior to superior humans for a very long time. This is important, and will make or break the robot car industry’s attempts to transform transportation.

Vision

Sensors are evolving fast. Human eyes get worse with age.

Psychomotor Skills

Computers all the way, drunk or sober.

Reaction time

people are limited here. Computers might be, too, but their limits are much faster.

Drowsiness

Computers don’t sleep.