Federal regulators says car makers “cannot wait for perfect” on automation

“Pre-market approval steps” could clarify responsibility in the future.


Read the whole story

 

[Jim Z]

Steveha7[/url]":3mkrpg6m]Is there really a demand for a driverless car? Driving is the fun part and is why I got a license. This may help handicapped people and if kids can ride in them without an adult. Other than that, I don't see much demand.

#1 the group of people who actually enjoy driving (i.e. enthusiasts) is a small percentage of the car-owning public (and they think they're way bigger and more important than they are.) the average person buys a car because they "need" one, not because they like driving.

#2 I'm an enthusiast (I even work in the industry) but I have to admit that the bulk of the driving I do (commuting) is shitty and I hate it. It's the kind of driving that makes even me hate a manual transmission at times. I would rather let my car deal with the traffic on the interstate then for me to sit there hating people left and right.

(I'd actually rather move to commuter rail, and save the driving for non-commuting time. hell, given where people tend to live and work in the Detroit area, you could run rail along all of the major interstates and state highways in the tri-county area and stand a good chance of connecting a hell of a lot of us from home to work, and v.v.)

 

[Jim Z]

JButler[/url]":2naiiyyc]

Rosekind said automation would “save people’s lives”

Again, is there real evidence proving/projecting this? Or are we just assuming things will work out? I'm talking about Level 3 and above where drivers are not expected to pay constant attention to cover for system limitations and bugs. I also highly doubt Tesla's Autopilot in current form reduces accidents.

it's going to rely on vehicle-to-vehicle (V2V) communication, so your car "knows" what surrounding cars are doing, and are going to do.

but even then, an autonomous car doesn't drive drunk, tired, isn't texting friends or taking selfies, isn't all pissed off and driving like an asshole, isn't some 16-year-old asshole who thinks the roads are race tracks, etc.

 

[Jim Z]

Dietz[/url]":1nm6g3c3]So, are these same regulators going to shield car companies from civil liability when imperfect automation causes driver fatalities?

Yep, didn't think so.

Liability is one of those interesting questions which will need to be answered. Who's responsible for the car? The owner? The manufacturer? What happens when an autonomous car and a human-controlled car collide?

the Tesla case is interesting, because AutoPilot is likely only a tertiary factor in the collision. just strictly going on the event as described, the fault for that crash seems to be (in descending order):

1) The truck driver for not yielding the right of way (I believe he was ticketed)

2) The Tesla owner for not being in control of his vehicle,

3) AutoPilot for not being able to "see" the broad side of a 40-foot trailer.

in Michigan, with a two-vehicle crash the investigating officer can assign fault to either party, or assign "shared fault" by noting both parties took a "Hazardous Action" contributing to the collision. Had this happened in Michigan (and had the Tesla owner survived) it's probable the truck driver would have been cited for a "2600.Failed to yield right of way," and the Tesla driver would have been popped for "2300.Fail to have car under control."

 

[Jim Z]

Thucydides411[/url]":uth8uzjm]

halse[/url]":uth8uzjm]If the preliminary assessment of a Phase 3 drug trial shows that X reduces fatalities by a factor of 2 relative to the current Y the trial is stopped as it is unethical to continue giving patients Y.

That is about where Tesla's autopilot currently is.

Tesla's autopilot actually is not known to increase safety. It might increase safety, but there's simply not enough evidence to demonstrate that with even marginal confidence.

One fatal accident in 130 million miles of driving on the highway is very similar to what unaided humans achieve. It's actually significantly worse than what other luxury sedans achieve. But this is just one accident, so the statistics are extremely uncertain. For all we know, Tesla's autopilot could double or halve the rate of fatal accidents. It would take more than an order of magnitude more road testing to know which is more likely.

very good point- according to NHTSA for 2014 (the most recent year they show) the rate was 1.08 fatalities per 100 million miles driven. but in fairness, that number in 2014 was a record low. Estimates say it went up for 2015.