Uh oh, another Tesla autopilot crash fatality

[botsapper]

The first of its kind, a fatality in Autopilot mode. RIP

https://www.teslamotors.com/blog/tragic-loss

[redpoint5]

That's a problem with tractor trailer design. They aren't made with the safety of other vehicles in mind, as the trailer is often just at about head height for most passenger cars.

Interesting to see that there have been fewer fatalities per mile in autonomous mode than for standard driving. I figured this would eventually be the case, but didn't realize it was already that good.

[Hersbird]

I think their autonomous miles with no fatalities isn't that great when only compared to every other car on the road numbers. How many fatalities per mile are there in 0-5 year old $50,000+ sedans? They basically are trying to say, "Hey at least it's safer then an 11.4 year old Camary or Accord." because that is the average car, and I bet those cars in isolation actually beat the Tesla claim. Me thinks 30% above average fatality rate is not what a $75,000+ car shopper is going to be doing a lot of bragging about.

PS, Now consider this was a dry, clear, daylight, perfectly straight road, with no alcohol or speed involved. What is the average fatality per mile in those conditions? Again, without the autonomous feature activated I have heard Tesla bragging about being the safest car ever on the road. I bet this means you are 10 or even 100 times more likely to die if you use the autopilot. In their defense, one data point is not really useable, but they shouldn't try and use it either.

[Hersbird]

I guess the driver here was the same driver that recorded the near miss with the utility truck posted here before. Makes me wonder if he wasn't a little to dependant on a driving aide. I thought the utility truck incident was an easy thing for a human to spot, evaluate, and avoid without the drama assigned to the "near miss" of the autopilot oblivious to dangers until the last moment. Don't get me wrong, I'm all for this driverless car stuff, just pointing out Tesla should do less bragging about it, and humans need to be more aware of the limitations. The pioneers always take the arrows.

[oldtamiyaphile]

Originally Posted by redpoint5

Interesting to see that there have been fewer fatalities per mile in autonomous mode than for standard driving. I figured this would eventually be the case, but didn't realize it was already that good.

I'm guessing that in general, high end luxury cars have a lower fatality rate than the national average, which will include things like barely legal old jalopies on bald tyres.

Sample size of one proves nothing of course.

[MetroMPG]

Originally Posted by Hersbird

Now consider this was a dry, clear, daylight, perfectly straight road, with no alcohol or speed involved. What is the average fatality per mile in those conditions?

That was my first thought as well. Fatality rate on freeways/highways, where autopilot is likely almost exclusively used, is going to be dramatically lower than on other roads.

[Ryland]

I heard that the auto pilot system cut accidents by 50% in Tesla's.
In this case it sounded like the Tesla was driving into the sun as well, so a human driver wouldn't have done anything different.

[Pooft Lee]

Neither Autopilot nor the driver noticed the white side of the tractor trailer against a brightly lit sky, so the brake was not applied

This grabs my attention.

1) I wasn't there, but some amount of time passes before a jacknifing trailer makes it into your lane from the other side of the freeway. He could have panicked, he could've had a terrible reaction time, but I'm not convinced the driver was watching the road. Brightly lit sky is nicely included by the lawyers as some explanation, but the brakes were never applied in a fatal head on collision?

2) Who is familiar with the model S's sensor array? Am I wrong in thinking color and contrast doesn't matter at all here? The whole white trailer thing feels like a cop out. At some point before or after the high speed convertible conversion, the car should have done something if they wanted to call it autopilot.

Its fairly likely the driver wasn't present, and missed one of the most critical moments of his life because he was letting the car pay attention for him. If so, its very unfortunate and a little humbling. Regardless, the Tesla never knew what hit it, either because of a reflection and some sunlight, or because Tesla didn't anticipate obstacles above the hood. If I remember right, another driver tried to sue them because he left his car to self park (which he's totally liable for) and it repeatedly bumped into the back of a tractor trailer. Tesla is learning some harsh lessons, but just like every other manufacturer, they aren't going to let us know if they really got it wrong

My $0.02

[wdb]

NYT has more. The guy was a tech geek and a big Tesla proponent.

http://nyti.ms/2992FVh

[Hersbird]

Originally Posted by Ryland

I heard that the auto pilot system cut accidents by 50% in Tesla's.
In this case it sounded like the Tesla was driving into the sun as well, so a human driver wouldn't have done anything different.

The guy had a dash cam plus the Tesla's camera should store some am out of time before a collision. I bet you could see the semi just fine. The problem was the driver wasn't watching and the car thought it was an overhead sign.