Tesla Model X autonomously crashes into building, owner claims | Hacker News Reader

Tesla Model X autonomously crashes into building, owner claims

computerworld.com

42 pointsFoggydog68 comments

I'm very skeptical about it. A lot of accidents happen when people confuse the brakes and acceleration paddle. Furthermore, the person who had the accident claimed the break was unresponsive. This could further mean she confused the paddles and instead of breaking, she accelerated even more.

I'm happy to hear, though, that the person was not heavily injured (she had burns on her arms but I don't think they are of the second or third degree because she would have mentioned that). Those accidents can be quite dangerous (especially in fast cars like the tesla).

I'm looking forward for the black box reporting

Where did you get your information on the danger of a crash with a tesla vs a regular car? They surpassed every car in collisions tests and the concerns about the battery going up in flames have been fixed since the car was established.

He said "fast cars like the Tesla". He didn't say they were especially dangerous.

Since they said "fast cars like the tesla" I assume they're referring to the acceleration. This sort of accident has much more potential for mayhem in a Model X P90DL that does 0-60 in 3.2 seconds as compared to, say, a Prius which will do 0-60 in 10 seconds.

> Where did you get your information on the danger of a crash with a tesla vs a regular car? They surpassed every car in collisions tests and the concerns about the battery going up in flames have been fixed since the car was established.

I'm not saying that tesla is not safe, in fact I'm saying quite the opposite. Tesla is a fast car and if not properly handled, all fast cars are dangerous. Since the 75D's aren't delivered yet, I'd expect it to be a 90D or P90D. The go from 0 to 60 in 4.8 seconds. I believe they are fast then any other SUV out there (except prototypes, maybe), when in ludicrous mode.

So, I'd expect that this tesla did not crash with 10mph into the wall. I'd expect a Porshe with similar acceleration to have much more serve damages to the driver. This, once again, shows how safe a tesla is.

my bad!

I thought the same. Also, a new car feels odd even if you have a clean record. Don't like it when such stories make it into the news before anything substancial is known about the accident.

If news organizations waited for substantial information before publishing, there would be far fewer news organizations and far less news. While this would be great for society, it would be terrible for them....

Given that (1) pressing the wrong pedal is a fairly common accident, (2) the driver was new to Model X, and (3) the Model X has "one pedal" driving, it's extremely likely that the driver thought she was pressing the brake.

Contrary to what the husband says in the article, this problem happens to drivers of all ages.

One pedal driving? How does that work? When you take your foot off the accelerator it brakes? What about the clutch?

When not driving in autonomous mode, yes, taking your foot off the accelerator causes the car to activate regenerative braking.

And it's an electric vehicle, there's no clutch.

Electric cars have no clutch. The Tesla is a single-gear transmission.

Is it hard to get used to a car with no clutch?

I drive a manual car, and have driven an automatic on occasion. I found it quite simple in most occasions, but when you need to slow down suddenly, the instinct is still to use a clutch and brake.

most cars don't have clutches these days.

... in america. Most cars sold in Europe are still manuals, although automatics are slowly taking over there, too.

There's also robotic transmissions mostly on Japanese cars that are "automatic" but do have an internal clutch requiring you to take your foot of the gas to switch gears, I've used a car with one and it did promote more civil driving since you could not just bottom out the gas.

I drive manuals (diesels) normally, and had driven automatics a couple of times. My one drive in a Model S was a little anxious at first but I got used to it very quickly. You just have to drive responsibly: focus 100% on driving until you're genuinely used to it.

It's a much simpler way of driving though, as long as you focus it's easier.

EVs work more like an automatic transmission car, but they actually has no gears to change.

Yep most of them use a single speed fixed ratio transmission usually with only a reverse mode gear switch.

If I press both accelerator and brake pedals in my Model S simultaneously, it beeps at me and says 'multiple pedals pushed', then ignores me until I only press one.

That sounds like the worst possible solution. Why doesn't it just stop?

why would it not just apply brake?

Another relatively common "accelerator" problem is a floor mat that rests on top of one or both pedals.

That is some terrible reporting. Who knows what will come of this accident, but the one referenced where the guy claims the car ran into a trailer on it's own was pretty well refuted by Tesla's black-box style monitoring.

It raises the question of how to prove the trustworthiness of the event recording. I suppose an audit of their recording mechanisms could be enough, but it would be nice if the permanent recorder was made by a third party.

Crazy idea but what about a camera placed in a way that Tesla would capture footage of the brake and accelerator. Store up to X minutes/hours of that would come in helpful when combined with the other data they collect.

Sounds like an unnecessarily complex alternative to just logging the pedal state the same as you would a keypress.

The problem with just logging like that is in the reliability of the logs - what if there's a defect in the position sensor? The pedal (and thus the logs) may report throttle at 100% and breaks at 0% even if it was the other way around.

What is complex about logging video footage with sensor overlay (or still shots)? Plus it gives you evidence that any jury can understand.

Yeah, but what if the pedal state was bugged?

Look, a carpet snagged on pedals a few years ago and that was sufficient to cause a recall.

http://www.consumerreports.org/cro/news/2011/02/toyota-annou...

That's the level of service I'd expect from any major auto-company. Just because the pedal was down doesn't necessarily mean that we've got driver-error going on here.

New car, 5 days old, seems from the onset that it is driver error as someone was in the vehicle behind the wheel. New vehicles are a major change to routine. Speculation... the data might reveal that the wife lied to cover crashing a 5 day old car.

This story will certainly be more interesting once Tesla publishes the data from their side.

The story is already interesting because Tesla does this, and no other car company seems to.

Tesla monitors your driving behavior so that they can save face and blame the driver if things go bad. I don't know of any other car company that does this.

Those monitors are installed to protect the company image, not you.

Its like you're buying a $100,000 piece of spyware. The car isn't your possession. Its sensor data belongs to Tesla the company, and not you. Despite your purchase of the vehicle. And that's assuming that Tesla's secret black-box technology is actually legitimate, and that they're not making stuff up / hiding the truth from us.

This is true, and while I do despise the invasion of privacy, and everything that comes with owning a $100,000 piece of spyware, I do appreciate that barring conspiracy and a flagrant disregard for the law someone somewhere has some sort of technical data unlikely to be vulnerable to human memory/eyewitness accounts related to the crash.

Tesla has also convinced me that it is the type of company whose employees that would not seek to hide something data it correctly clearly supported. That's a completely unverifiable (and most likely incorrect) assertion on my part, and of course a result of brand-cultivation on theirs, but still some companies just can't claim that at all, and I think Tesla has a leg to stand on still.

> someone somewhere has some sort of technical data unlikely to be vulnerable to human memory/eyewitness accounts related to the crash.

I can agree with this. Having this information recorded should be a standard feature for cars of this price-range.

However, the information should be in the customer's control (which should also be fairly passed onto Tesla through a subpoena if necessary).

Its a basic tenant of "ownership". If you don't have primary control of the data, you really don't own it.

True, and when Tesla doesn't release a report on an incident like this it will be telling.

As for the legitimacy of their recorders, I think it would be best to have it audited etc. That said, doing anything shady with it is a risk way bigger than any potential rewards, so I'm not that suspicious. It would be worse for their credibility than VW's emission cheating if they were falsifying their recordings and then using them to blame customers.

This is why I hope that Lyft/Uber/? purchase and run their autonomous fleets as planned. I can keep my gas engine car for longer trips while taking advantage of super low cost robot transport for day to day stuff.

I dunno why you are getting downvoted for saying this. Seems like a very legitimate concern.

I believe you can opt-out of the data sharing if you want, though nobody does. In any event, I doubt Tesla is storing this data just to protect their company image. It may be a nice side benefit, but they have repeatedly said that the data is for training their automatic driving systems.

Literally EVERY single one of these accident reports have been followed by Tesla reporting the driver log data on their blog.

EVERY SINGLE ONE.

Look, I get not trusting reporters in a lease of a review model (especially when said reporters were rather shady to begin with). But... we're onto customers and the general public now. The fact that Tesla is still engaging in this behavior is concerning to me.

And more interesting when the air bag controller is dumped by a neutral third party. That's not under Tesla's control; it's a Takata unit, self-contained, with no network connection. When the air bags fire, the controller saves the last 8 seconds of vehicle speed, brake light switch, engine speed, throttle position, acceleration, and seat belt use. It's basic data for crash analysis.

There are companies which do this and analyze the info, such as Failure Analysis Associates and Crash Forensics. Bosch makes a readout tool for this.[1] If the vehicle isn't too badly damaged, the data can be retrieved by plugging into the OBD port. Worst case, you have to get the airbag controller out of the vehicle (it's a tough device, designed to survive a crash) and connect it to the readout tool.

[1] http://www.crashdatagroup.com/

Or the wife might be telling the truth, but be mistaken about what happen. It's fairly common for people to press the accelerator by accident thinking it's the brake, then press it harder in a panic since it's clearly not working. The result is exactly as described here, with the car suddenly accelerating forward and the driver insisting that they were hard on the brakes but they didn't work.

There are plenty of vidoes on YouTube of cars "suddenly accelerating to maximum speed, jumping a curb and slamming into things". They all do it accidently. People will blame others for their errors. Bad journalists will pick up the shit for page views.

Tesla is famous for publishing the logs. The owner should be aware of this before making claims. It would have been prudent to keep this quiet till Tesla had a chance to respond. Now if the logs prove it was driver error Tesla will publish the logs and all this outrage would be over nothing. On the other hand if this was an error in Tesla's system it is in Tesla's best interest to be transparent and publish the logs and an explanation on how it intends to fix this.

Either way a public outrage was not required unless the owner wanted to ensure he got a new car even if it was the driver's fault. By claiming ""I don't' know if they're going to acknowledge if it was their fault," Puzant said.", the owner wanted to cast doubt on Tesla's integrity and also give himself an "out" if Tesla determined it was driver's fault.

I would suspect that if Tesla has a rock solid way to determine it was the drivers fault they will fight that till the ends of the earth. (Of course the same could happen even if not the case but that would be a conspiracy and probably less likely for practical reasons).

Considering that the owner's wife said that she tried the brakes but they weren't working I am pretty sure there will be brake usage logs coupled with associated acceleration/deceleration data. That should tell us if the brakes were applied and if they were working. The acceleration logs will also show if the acceleration pedal was engaged at the time or not. I am assuming that an auto acceleration log is different from a hardware pedal usage log. If not we can't determine if the acceleration was due to software or hardware.

Sure but by the same token in theory if a brake was applied it should work and in this case the car owner is saying it didn't work (hence my video footage comment elsewhere).

Given that this type of accident happens all the time in all types of cars, why is it that the car owner's memory is preferred to a sensor? Because you're basically saying that ALL cars have the same flaw. It's a lot more likely that humans are fallible.

Tesla is quick to publish data that shows driver error, I wonder if they will be as quick to do so if it actually is a software issue. My guess however is that any refusal to publish the black box data will be taken as an admission of guilt.

I get the impression that the media is just itching for an actual software error so they can start a FUD firestorm about SDCs.

I don't think they quickly publish data to show data error, but just to refute claims that their product is deficient or dangerous. This might be what you mean but they aren't looking to disparage customers but just to maintain the integrity of their brand to false claims.

> It's not like she's a 90-year-old person who's going to press the gas pedal instead of the brake

While there is a correlation between age and difficulty driving, I don't know how this is a valid defense. In moments of stress and danger we're all prone to making mistakes.

> It's a mile and half away from our house

Many accidents happen close to home due to your perceived feeling of knowledge and safety in the environment.

Looks like an answer is already out:

http://electrek.co/2016/06/06/tesla-model-x-crash-not-at-fau...

No idea about that source but supposedly tesla has made a statement. Similar statement also at this site:

http://www.teslarati.com/tesla-model-x-crash-result-human-er...

Turns out my skepticism was right after all, the driver did hit the wrong paddle: http://electrek.co/2016/06/06/tesla-model-x-crash-not-at-fau...

Discussion: https://news.ycombinator.com/item?id=11851817

The state of journalism today is pathetic. Some random person claims something that is extremely unlikely and it gets run as a story. No verification of events or even a comment from Tesla. They likely published intentionally before Tesla's comment because they knew it would be "Our logs show the driver did something incredibly stupid and it had nothing to do with auto-pilot" and this would become a non-story. Instead we have this clickbait bullshit that unjustly harms Tesla's reputation.

"He said his wife had not activated any self-driving features at the time of the crash."

He said it himself, the electronics were off, his wive fucked up.

They are claiming that the features were off - but malfunctioned and 'worked' anyway. (eg: a mishap by Tesla)

As with most others here, I believe it was user error. These sort of accidents happen - it's just "unique" this time since they are claiming it was Tesla's fault.

If we don't develop due process for machines, they're going to be the scapegoat for everything.

...and I'm going to get away with murder.

The model X seems to attract a lot of issues. Too bad it also tarnish the reputation of the model S.

This is a very bold effort on Tesla's part. Fast deployments of autonomy code has never been tried before. There will inevitably be some casualties (even some deaths) in this process, but no other process can guarantee otherwise. The long term impact on public health is HUGE [1]. Some company just has to figure out how to weather the storm of (possibly irrational) criticism from the first such casualty. The company that can parse through this process and update their methods first will win.

[1] http://www.sciencealert.com/driverless-cars-could-reduce-tra...

Unlike the previous accident, which the Tesla S literally doesn't have the sensors to see the above-ground trailer... this Model X crashed into the side of a building. I'd like to assume that the below-chassis sensors would have been able to pick that up.

I'm more concerned about edge-cases like the Tesla S autonomous crash rather than cases like this one. My current assumption (pending evidence: we don't have any yet so we can only assume) is that the driver is the blame in this specific case.

We'll see when more data comes in.

----------

Unfortunately, I don't think it is moral for a car company to install sensors and use sensor data against the car's owner.

If I buy the car, then the whole car is allegedly mine: including these sensors that only Tesla can tap into. The fact that I don't have access to all the sensor data means that Tesla's definition of "ownership" is rather convoluted compared to a normal car.

There's a "remote access" setting you can turn off. I believe that cuts off Tesla's access, not just your own. Of course, roughly nobody actually turns it off, because everybody prefers the convenience of having it on.

> I believe that cuts off Tesla's access, not just your own

So no way to cut off Tesla's access to YOUR OWN CAR, without cutting out your own access?

And even then, you're not sure?

Yeah, remote access goes through their servers. You can still drive the car, you just can't use their smartphone app. I don't know why you'd expect anything else.

If you want to be sure, then pull the SIM card. It's a bit of work to get to it, but not too tough.

> Yeah, remote access goes through their servers. You can still drive the car, you just can't use their smartphone app. I don't know why you'd expect anything else.

Just because I open my computer up for OneDrive doesn't mean I am going to let Microsoft keylog all of my keyboard presses. Or Teamviewer (ignoring mistakes like hacks / software vulnerabilities).

The closest analogue is Cortana / Swype reading keystrokes to train the keyboard AIs on phones, which is clearly denoted in the terms of use and can be turned off individually.

Regardless, Tesla's policy and technology stack are designed to protect the company, not the customer. Lets put it this way, do customers get this Tesla black-box data available to them if they get into a mundane car crash on a normal road? To prove that it was the other driver's fault?

Or does Tesla normally keep this data to itself?

I totally agree that it should be a separate control. But I'm not at all surprised that it isn't, and if you're so upset about this access that you consider it immoral, losing the smartphone app should be a small price to pay.

I have no idea if we get this data in other circumstances. If I ever find out, I'll be sure to let people know. I hope I never have to find out, of course.

Sure it's moral - in fact, it's great. The question of morality only arises if there's asymmetric access to the data. In the event of a claim being made by the driver against the company, both sides should have complete access to the logs data to make their case more firmly. What shouldn't happen is ONLY one party being able to inspect the data.

But this is unlikely to happen in practice. A lawyer for the driver would undoubtedly subpoena Tesla for access to the data. And a reasonable judge would probably concur. IANAL and IANAJ, so the latter is a guess. :)

> But this is unlikely to happen in practice. A lawyer for the driver would undoubtedly subpoena Tesla for access to the data. And a reasonable judge would probably concur. IANAL and IANAJ, so the latter is a guess. :)

So what's the process here? You're going to start suing Tesla so that you hope you can officially subpeona them of the data you don't have access to?

Its asymmetric no matter what. The role of these logs should be reversed: the driver ought to have first-dibs at the logs to determine if he has a legitimate case to bring forth against Tesla.

Then Tesla should subpeona the data once they start getting sued. In any case, Tesla holding the data so that they have the information-advantage against their own customers is a bit fishy.

Yes, that's exactly what you do when you sue anyone and you believe they have data relevant to the case. It's called discovery. First you have to have a reasonable basis for doing so, and a claim that their car auto-drove into the wall is such a basis.

Reply on news.ycombinator.com