Tesla driver again dies in crash with trailer, Autopilot not yet ruled out | Hacker News Reader

Tesla driver again dies in crash with trailer, Autopilot not yet ruled out

electrek.co

42 pointsjbuzbee29 comments

When will the US require trucks to have side underride guards like the EU? These accidents happen even with human drivers.

What do these look like? Never seen one.

US truck: https://heavyvehicles.com.au/wp-content/uploads/fresightline...

EU truck: https://www.euractiv.com/wp-content/uploads/sites/2/2014/12/...

On the US truck, there is an open space under the trailer, where the engine of a car can go in so the trailer decapitates the passengers. On the EU truck, there are bars under the trailer to stop cars from going in there and letting the crumple zone do its job.

If you've actually seen those guards in real live, you'd realise they're not able to stop anything heavier than a bicycle hitting them. The supports are often thin bent sheet steel and the bars themselves soft aluminum extrusions. Enforcements of standards is very lax in most countries.

You're probably confusing the crash bars with similar looking but non-structural aerodynamic fairings, which are much more common in the US. The EU-compliant crash bars are an integral part of the trailer frame, not something bolted on.

I was researching vehicle safety for a new car purchase a few years ago and came across this video: https://www.youtube.com/watch?v=bT3G-kcKN70

It freaked me out and I stay as far away as possible from semis on the road now.

This happened last year to some people I knew. They were in an SUV, stopped in traffic behind a flatbed truck. A semi hit them from behind and shoved them under the truck in front. Family of four, killed instantly.

After that video came out, a bunch of trailer manufactuers improved their underride guard designs, and the newer ones are apparently much better at stopping cars: https://www.iihs.org/iihs/news/desktopnews/all-major-trailer...

It's going to be a long time before most trailers have the improved guards, but hopefully it will get better eventually.

Why is this problem not legislated out of existence?

Because of lobbyists.

Video of side underride guards: https://youtu.be/mrL7AUMT4To

I agree. That's the whole point of having a bumper. I remember hearing one reason for different car standards in the US and EU was the average height of a car is much taller in the US, so the bumper has to be higher. There were concerns that a subcompact would go right under an SUV. Semi trucks should be a way bigger target; I would think they're in use more than a car, the average speed would be higher (since they're used more on highways), and the body style is more standard and less diverse.

I'm doubtful any such measure would effectively prevent a fatality with a BEV skateboard design traveling at highway speeds vs. an effectively stationary perpendicular trailer.

Basically all the vehicle mass is near the ground in a relatively thin layer. That component isn't going to stop easily. But the lightweight rest of the car, interfering with the structural part of the trailer, will shear right off.

It would be interesting to see a crash test, for sure. I imagine the top of the car is still pretty firmly attached to the bottom for rollovers, etc. But even if it's not, as long as the passenger compartment is firmly built and stays together with its crumple zones, isn't it fine if the bottom skateboard goes flying away as long as the passengers are safely decelerated in one piece. Or do the feet go down into the skateboard?

In the worst case, at least the underrun skirt should help the collision avoidance systems avoid the collision in the first place.

> In this new accident, the vehicle again kept going for over 500 yards (half a km) before coming to a stop.

Disturbing.

Presumably if the roof was taken off, lots of components of the autopilot system would have been seen as unavailable.

In that case, I would design the system to apply the brakes quite hard (unless autopilot was disengaged by the user). Going to full emergency stop mode would clearly cause accidents in other more mundane cases like a malfunctioning sensor.

A deceleration of 1 m/s from highway speeds ends up with the car travelling 400 yards, which is broadly in line with what happened here.

That's reckless. If the car is effectively blinded, continuing to travel at highway speeds for nearly half a kilometer sounds like a terrible idea.

If you see a convertible strolling down the highway with a decapitated driver at the wheel, you know what happenened :/

Statically are you more or less likely to die using Tesla auto pilot?

I wonder if the number of hours drivers went without autopilot, the number with, and the number of fatal accident occurred with and without autopilot are public?

Tesla has published numbers stating that the accident rate with autopilot engaged is roughly half the rate with it disengaged - across their fleet, so same driver pool. https://www.tesla.com/blog/q3-2018-vehicle-safety-report

Note, this data is almost certainly biased since AP can only be engaged in certain circumstances and is mostly used on the highway, where accidents are much less likely overall.

> Note, this data is almost certainly biased since AP can only be engaged in certain circumstances and is mostly used on the highway, where accidents are much less likely overall.

I don't disagree but highway accidents have to be the most dangerous as well due to the speed as well as other factors. I'm not as worried about fender benders as I am a 3, 4, 5, 6.... car pileup on the highway.

The question of whether one is more or less likely to kill or injure other people when using Tesla autopilot is somewhat more important. This is not an easy question to answer, especially as there are not that many incidents, and also there may be some features of the autopilot package that enhance safety, while others might create more risk than they mitigate.

I assume you are unfamiliar with the strange NHTSA report, claiming improvements that seemed too good to be believed - and were: https://www.latimes.com/business/autos/la-fi-hy-tesla-nhtsa-...

Reducing that to a number is dangerous. What if it was 100% safe but every hour a random tesla would just blow into pieces. Would you ride it? What if it was 100 Teslas? What if it was a number just enough to be statistically safer than regular cars?

Likewise, failing to consider the overall number at all leaves us prone to discount significant progress that has a scary failure mode. If a system made me and my family 10x safer overall but every failure meant we were all gruesomely decapitated, I’d absolutely take it. Many people would argue strenuously against it.

You seem to be conflating absolute risk and relative risk.

In the same example, if my odds of getting into a car accident was 1 in 1000, and the Tesla 1 in 10000, I'd drive myself because fender benders and wipeouts are included in that 1 in 1000.

If we're talking 1 in 5 to 1 in 50, this changes the calculus significantly.

You seem to be conflating safety ratio with accident ratio.

I specifically predicated on "makes me and my family 10x safer overall."

https://arstechnica.com/cars/2019/02/in-2017-the-feds-said-t...

"When QCS director Randy Whitfield ran the numbers for these vehicles, he found that the rate of crashes per mile increased by 59 percent after Tesla enabled the Autosteer technology.

So does that mean that Autosteer actually makes crashes 59 percent more likely? Probably not. Those 5,714 vehicles represent only a small portion of Tesla's fleet, and there's no way to know if they're representative."

Reply on news.ycombinator.com