SoCal man says computer on his Tesla froze, causing it to be stuck at 83 MPH | Hacker News Reader

SoCal man says computer on his Tesla froze, causing it to be stuck at 83 MPH

abc7.com

78 pointserehweb88 comments

I'm dumbfounded by people dismissing this article by saying that it wasn't "stuck" simply because the brakes worked. The turn signals didn't work, neither did hazards, the failure of which means the driver can't communicate to others that there is a major issue with his vehicle, nor that he is about to perform certain maneuvers such as lane changes or sudden slow downs.

It also suggests that the vehicle was maintaining a steady 83 MPH. Whether or not you can brake to slow it down and come to a stop isn't the point; the car is insisting it is going to drive at a specific speed unless you tell it otherwise, that's absolutely not OK and yeah, that's technically "stuck". You've also got the additional damage that occurs by hitting your brakes when the motor is still trying to make it go 83 MPH.

Yeah, interesting how people on HN sometimes really get stuck on technicalities...

The words we use matter, because they're how we build a shared understanding of the world.

I'm guessing, based mostly on how I feel, that people aren't so much annoyed at the negative Tesla news: people are mostly annoyed that the problem description in the title is misleading, and that you have to read the contents of the article with some care to understand what actually happened.

It's objectively sloppy/bad journalism, and I think the only question is "how much does it matter". I'd argue it matters a lot in the context of a news aggregator: we should be able to find a more accurate, less misleading, or less impressionistic piece of journalism on this subject.

This shouldn't be understood as a cry for "nothing to see here" or "fake news"... On the contrary, it still looks bad for Tesla. Let's just make sure we're focused on accurate wrongs.

We're pointing out the issue isn't nearly as bad as what the headline states. That doesn't imply there isn't a problem.

a confounding factor is whether he had cruise control active or not. Not ruling out that this could have happened as described however my understanding (and in my experience as a Tesla owner) is that the main screen only only for user feedback. The main screen crashes maybe once a month for me, and although everything works while it reboots, the indicator sounds don't play giving you the impression they’re not working.

If you have autopilot or cruise control on they’ll continue to operate even though you can’t see any feedback or hear the indicators.

I could see how someone could get into this situation and while panicking accidentally activate cruise control and think the car is stuck.

Cruise control should disengage when you hit the brakes for the first time and the car should begin to coast.

> The main screen crashes maybe once a month for me

Excuse me, but what the fuck? How is it possible that the multi-ton high-speed metal box with potential to kill people reboots once a month? If my car was rebooting once a month, I would definitely get rid of it. No feedback means I don’t know what my indicators show, which means they might be misleading to others and cause crashes.

To be clear, reboot != crash while driving. Sometimes I get in the car in the morning and the screen is black and thus I need to press the steering wheel buttons to reset. Or similar, sometimes the screen still works but a particular feature doesn't (no cellular connectivity, the map is frozen, songs don't play etc)

I've had a BMW i3 and a Chevy volt, and I'd say they had similar issues, however since the instrument cluster wasn't on the only display, it wasn't as notable of an issue.

In both cars, I've driven kilometres without working music/nav/whatever until I could pull over and reboot.

This screen is an infotainment screen. The driving systems are independent and still function normally if it crashes.

Fwiw this is a fucking terrible issue and has scared the fuck out of me when it happened.

I do have to add though, turn signals etc etc all worked. You just didn't hear the audible feedback of them.

I had brake failure in my 2014 Model S, completely unable to stop, and I sailed through a red light before regen slowed me down enough to turn off into a driveway.

I know you're thinking it was pedal error or something similar and I would definitely dismiss someone else's story the same way. It was not reproducible. But I am 100% certain that the brakes were not responding even using both feet on the brake. My best guess is that something was unexpectedly frozen given the outside temperature at the time.

[ I still later traded it in and bought a new Y though, lacking better EV options at the time. /shruggie ]

Wow, so the mechanics failed but the electrical part still worked. That's pretty shocking, it should have been the other way around every time. I'm trying to think of how the brake system could have been designed that this would even be possible, the only thing that I can imagine is that a dual piston master cylinder had one of the pistons freeze solid so that the whole system failed to move. But that should be pretty much impossible.

One of the disadvantages of regenerative brakes is that if you use them well that you will end up using your hydraulic brakes less than normal which can cause the pistons to seize (and which is an expensive repair if they don't free up by themselves), did the brakes function normally prior to this or did you not use them for a long time?

It hadn't been driven in a couple days, and it was slightly below 0 C. So I guess it's possible something froze, but it wasn't excessively cold out.

I am mostly a one-pedal/regen driver, but the brakes had to have been working a few days prior because the car doesn't fully stop without using the hydraulic brakes to go from low speed -> stopped.

But after low speed testing in the parking lot I turned into I couldn't reproduce any failure so I barely had anything to report that would have been credible.

I also learned that you can't slam on the parking break (what I would have called an emergency brake in other cars!) because it's software-disabled above very low speeds.

> But after low speed testing in the parking lot I turned into I couldn't reproduce any failure so I barely had anything to report that would have been credible.

If it happens again, remember, Teslas collect and send an enormous amount of telemetry. It's very possible just with a time period they'd be able to look into it.

You may have water in your brake fluid then. Brake fluid is very hygroscopic you should totally exchange it very two years or so regardless of how full the container is. Once unstuck it would stay unstuck until the next freezing cycle.

You can have a cheap water contents test done for your brake fluid at almost any garage.

> so the mechanics failed but the electrical part still worked.

Electric cars use brake-by-wire. Push on brakes, computer processes signal, computer applies brakes. Glitch somewhere in that system

Not for braking past a specific point, it's just regular hydraulics once you get out of the range for regenerative braking.

Hmm, interesting. I didn't know that

Brembo is working on an all-electric braking system where the calipers are actuated by a servo system but afaik this has not been released yet and even then you'll need to meet or exceed the kind of performance you expect from regular hydraulic brakes to ensure that in an emergency situation that car can be stopped fast enough, which regenerative brakes can not normally provide (if only because the electrical system could fail).

What is "regen"?

Regenerative braking

When you lift your foot off the accelerator, the car starts to run the motors "in reverse", charging the battery with some of your momentum. So the car slows gradually even if the brake pedal is unresponsive.

Did your Model S have a camera pointed at you as a driver?

It's going to be so much easier to prove such cases when video evidence of the driver's reaction during the incident is available.

> regen slowed me down

good thing the battery wasn't full

I mean it's all in software right? So if a system crashes and there is a relay that tells the motor to spin, not spin (which is probably not how variable speed works) but follow me = then if that subsystem crashed (some exception is thrown) and the code cannot be recovered - then whatever relay state the vehicle was in may just stay that way.

ICE vehicles may be much more complicated, and we all know they're going away, but let's not dismiss these real issues.

It's amazing to me how there are so many Tesla Stan's on HN that any time any article gets posted that is in any way critical of Musk or Tesla it immediately gets flagged and booted from the site.

These articles are often inaccurate or misleading.

I'm not suggesting Tesla is beyond reproach (in fact I believe in the importance of high accuracy of criticism, so we can focus on the bad things that matter), but negative Tesla news does attract clicks, so it shouldn't be surprising that the quality of negative Tesla news tends to be lower than the quality of positive or neutral Tesla news.

Where "high quality positive news" is a tweet of a drone video, and "low quality negative news" is an LA Times article about ridiculous racism at Tesla.

The words you're putting in my mouth don't reflect my feelings or beliefs.

Don't put words in other people's mouths.

---

See: https://news.ycombinator.com/newsguidelines.html

> Be kind. Don't be snarky. Have curious conversation; don't cross-examine. Please don't fulminate. Please don't sneer, including at the rest of the community.

> Please respond to the strongest plausible interpretation of what someone says, not a weaker one that's easier to criticize. Assume good faith.

This is an interesting case but I wonder what the statistics are around this. Computers die just like cables snap. There will absolutely be unpreventable failures, but there really seems to be an interesting psychological component to the way we expect computers to work flawlessly where we are better at acknowledging the failures of mechanical components. Obviously the failure mode here was not ideal, so hopefully that can be improved.

If a throttle cable was jammed it would be the same outcome and at least for me, I can imagine it would make me feel less uneasy. Maybe it's the consistency of code that makes us expect that if one failure exists, more are likely?

No sources, but there were comments here on HN in the past year which pointed out non industrial grade (for the lack of a better word) computers on board a Tesla.

My opinion is that with mechanical components it is easier to make a distinction in quality. With computers, we tend to stick to GHz or how many nanometers some tinfoil is.

> non industrial grade (for the lack of a better word) computers on board a Tesla.

Such as the non-volatile memory chips that began to fail after irresponsible logging practices [0]

[0] https://hackaday.com/2021/02/11/tesla-recalls-cars-with-emmc...

Not sure which computers you mean, or what you mean by "industrial grade," so possibly I'm off-target, but cars could happily have some less reliable computers, for example in the entertainment center. Even stuff like high-level navigation could be run on the entertainment center level computers -- if google maps crashes, it shouldn't cause a human operated car to crash, and a thoughtfully designed "autopilot" system should mimic that split. No idea if the Tesla autopilot system is thoughtfully designed though.

It seems at least recent versions of the Tesla Model S and X have seperate computers for their infotainment system and autopilot [1]. That wouldn't have helped out the guy from the article, though, I believe he was manually controlling it.

If Teslas have their infotainment (running in React Native?) and motor control/drive by wire on the same computer, I'd be shocked. That's awful.

[1] https://insideevs.com/news/414323/teslas-computers-different...

> What happens is that the Model S and the Model X have two physically distinct computer units: one for infotainment and one for Autopilot. The infotainment computer is called MCU, and it is integrated with the central screen. The Autopilot computer is installed above the glove box.

I wonder what the architecture there is. The author just says "physically distinct," so it is hard to guess much. One one hand the different "computer units" appear to each contain different compute elements -- at least, I think we're seeing the back of the boards and can see multiple CPU cooler retention brackets, right? On the other hand, they all are talking over a network of some sort, so hypothetically they could end up in a situation where the one device ends up dependent on the other (I agree though that it would be shockingly awful and it seems really unlikely).

Even within the Autopilot I guess there must be a mix of safety critical and not so critical elements, right? Autopilot has two tasks -- finding the directions to some address, and not crashing the car. The former is not really safety critical (in the sense that a lost driver can operate the car without crashing it) and is hopefully run on a different computer.

This is literally what Engineering is, designing the components to spec such that when they fail the failure mode is acceptable. This includes the computers.

That said, "throttle stuck at previous speed" is not a horrible failure mode, all things considered. If you were driving safely at a certain speed then you should have some time to react and pull the E-break (push the "Park" button on a Tesla). Slowly losing speed until stopping would be better, but there are much worse failure modes too, such as suddenly braking or accelerating.

As per article, Breaks were working but accelerator was not, Why accelerator needs to be connected to main computer and could not be controlled independently without main computer? Access to accelerator/gas is as important in critical situation as the breaks.

Title is click-bait. Vehicle was not "stuck" at 83mph. Brakes and steering were operational, while car computer rebooted.

That is still a big deal. See also the Toyota accelerator issue: https://en.wikipedia.org/wiki/2009%E2%80%932011_Toyota_vehic...

Apparently you're not very familiar with how that story turned out: https://www.wsj.com/articles/SB10001424052748704422204576131...

tl;dr: we don't really know what caused it but it wasn't electronics. Malcolm Gladwell did a great piece on it: https://www.youtube.com/watch?v=1N3naJ0P0gM

Why does it have to be electronics for my reference to matter?

The idea is just because the brakes and steering still work doesn't mean the car is safe.

> The third and most common problem was drivers hitting the gas when they thought they were hitting the brake, which the NHTSA called "pedal misapplication."

The problem was drivers getting the pedals mixed up. There was nothing wrong with the cars and officials blamed the drivers—hence the title of the article: "U.S. Blames Drivers, Not Toyota"

I thought “we don t really know what caused it.” Now you’re saying we do?

A few months after that article... NASA did some investigating and found tin whiskers in some faulty accelerator pedals. And interestingly enough, these pedals would all test out fine. This particular one the complaint noted that she had to hit the brake to get the car to stop. So she did not end up crashing. So I'm inclined to believe her car really did take off suddenly. This presentation changed up my mind a bit about unintended acceleration in these cars.

https://nepp.nasa.gov/whisker/reference/tech_papers/2011-NAS...

I do believe that it could happen, but also I do believe that in most cases you should be able to slam on the brakes and stop the car.

What about this part?

"the car was stuck going 83 mph"

Seems pretty clear to me.

It seems the brakes were fighting against the motor. Also, try turning at 83 MPH.

"Stuck [unable to go faster than] 83mph" is what they knew you would understand, right? (Sarcasm)

So I agree it is click bait, but it does still sound dangerous to be unable to signal or accelerate on the freeway.

And if the signals were not working, I wonder if the brake lights were? Hopefully they are independently wired and directly switched from the brake pedal and the 12v accessory power. Slowing down on the freeway without working brake lights is also quite dangerous.

Operational brakes vs 'brakes actually capable of overcoming the motor' are two entirely different things. You'll find that for many vehicles, the brakes alone will not be sufficient to stop if the motor is attempting to accelerate; you need to shift to neutral (potentially destructive), or kill the engine. Plenty of people accidentally drive around with their emergency brake engaged and don't realize it.

Pretty much every vehicle except maybe some heavy trucks with really, really, really deep first gears has brakes that can over-power everything else.

This should be obvious when you think about what % of brake force it takes to break traction vs what percent of engine power it takes.

> the brakes alone will not be sufficient to stop if the motor is attempting to accelerate

Malcolm Gladwell went into detail about how this is BS on Revisionist History: https://www.youtube.com/watch?v=1N3naJ0P0gM

It doesn't pass the smell test either: what car accelerates faster than it can stop?

> It doesn't pass the smell test either: what car accelerates faster than it can stop?

A dragster?

But more seriously, modern high-performance electric cars (such as some Tesla models) are capable of both acceleration and braking pretty much at their tires' adhesion limit.

I would still expect the mechanical braking system to be designed to be able to overcome the maximum torque of the motor although I do not know whether or not that is in fact the case.

> You'll find that for many vehicles, the brakes alone will not be sufficient to stop if the motor is attempting to accelerate;

Maybe in 60s cars with 4 wheel drums. But the brake force on even basic modern cars is sufficient to overcome like 800hp. Go ahead and give it a shot in your own car and see. I have a 400+hp car with shitty brakes and it will still stop completely and readily if I hold down the throttle and brakes at the same time in second gear.

Now, if you ride the brakes, they will overheat and lose stopping power. So you're going to want to actually try to stop the car.

>Maybe in 60s cars with 4 wheel drums.

60s 4wd truck or SUV will probably fit the bill. Nice deep 1st gear, ~2:1 low range, and available with manual brakes on the lower trim levels. In the lowest gear it can probably just idle over the force of a small women or scrawny teen pushing the pedal.

Working brakes are mechanically able to stop an out of control engine. The limit is how much force a person can apply to the brake pedal. NHTSA and manufacturers target a 5th percentile person being able to stop the car, assuming power braking is still working.

PDF https://www.autosafety.org/wp-content/uploads/2015/03/The-Br...

> you need to shift to neutral (potentially destructive)

Why is it destructive?

Please specify a vehicle. It would be very rare for a standard passenger vehicle (non-bob tail semi). Even for muscle cars your emergency stopping distance from 60mph is almost always shorter than your 0-60 distance.

This is one of the proof points for people who experience "sudden acceleration" while pushing hard on the "breaks" since the breaks could stop those cars. There are drive by wire elements/booster of many modern cars and EVs do have higher low speed torque. Perhaps if you were going fast and got the breaks nice and hot before they could stop the car? It's interesting to me that the guy admitted he was intentionally exceeding the speed limit.

For the Model S Plaid, however, it may well be the case: https://www.youtube.com/watch?v=rh12lK53hss

If the title had been “Tesla glitched at 83mph” it would have been exactly right and not less scary sounding.

Stuck at 83 until he used the brakes and steering to control the car? Doesn't sound very stuck. This is not a story.

The failure of any component required for functional safety is absolutely a story.

I agree with you, but I also agree with the other commenter because I think the words we use matter. I think the failure could and should have been described a bit more accurately in the title.

> He said all of the buttons and switches - including turn signals and hazard lights - were not working.

That's fairly concerning. If it was in self-drive mode, would the "you're on your own, autopilot disengaged" feature a) trigger and b) make the sound cue? Or would you run straight into the car in front of you?

Quick side question... so he just admitted to speeding?

I imagine this story was worth orders of magnitude more to the news network than the ticket would be worth to the man... I'm sure they could make it worth his while :)

After having been behind a Tesla emitting smoke on Sand Hill Road, I'm not keen to purchase one anytime soon.

When such an event happens, do you think Tesla replaces the battery pack with a fresh, less defective one? Somehow, I doubt it.

Stories about Tesla support don't inspire much confidence. Once they have your money, the goodwill and standing by their product is mysteriously absent.

With ICE cars being comparatively easy to fix, I'm content to stick with them until the e-kinks get worked out. It's still the early days (easy to forget when Teslas are so ubiquitous in the Bay Area).

Because an ICE vehicle has never caught fire.

Electric cars are a bit harder to extinguish.

https://www.motor1.com/news/315476/bmw-i8-fire-reponse-nethe...

ICE vehicle fires are a lot less scary than lithium fires.

That's true, but they are plenty scary and they can happen really fast as well. From 'everything fine' to 'inferno' in less than 10 seconds.

Good point (I'm personally on your side), but what is needed from both EV-cautionary and ICE-cautionary commenters is some odds ratio or data on both, rather than impressionistic and qualitative quips about either.

For all those people calling it a clickbait headline... You're not wrong the headline is a deliberate scare tactic.

At the same time... While Tesla is marketing it's "Full Self Driving" terminology, I'm actually glad that someone else is also using misleading language to grab attention, and remind the public that the state of the art is still quite limited.

I thought I remembered reading something similar a while ago (except for the speed being stuck), a quick google yields quite a few similar cases:

https://www.google.com/search?q=tesla+computer+shutdown+whil...

I've had my screen freeze once before while on the highway. I did a soft reset to be able to restore it but it took about a minute. I was still able to drive, speed up, and slow down while the screen wasn't working but I could only guess how fast I was driving without the screen displaying it.

Non-deterministic computation has no business being inside a safety-critical system. How is this legal?

Only short 5MPH. GREAT SCOTT!

Just 5 mph too slow :/

Or alternatively, a safe 33mph margin to keep from exploding!

Well Tesla does share the abdominal quality that the DeLorean had.

That gut-wrenching feeling.

Indeed, and it is so hard to stomach.

Too bad it wasn't a model X with gull-wing doors.

Sigh, how do news sites get away with blatantly wrong titles? The driver was obviously able to brake, so it wasn't stuck at 83 MPH - although the display is frozen and showing that speed.

Imagine if it's a headline like "Putin declares war over" and actually it's him saying something like "War will be over when Ukraine has a pro-Russian leadership".

You joke but I wouldn’t be surprised to see a headline like your example at some point. I’ve seen similar clickbait, albeit for lower stakes.

We have an example right here: https://news.ycombinator.com/item?id=31015652 - the title implies but doesn't say, and the reality is that sanction busting was involved. Titles have become less than useless.

> Sigh, how do news sites get away with blatantly wrong titles?

I don't mean this snarkily:

Notice how your reaction was to complain about it, but (AFAIK) take no concrete action against the news site? I suspect most/all of the public did the same, and thus the new sites get away with it.

I saw a link to the "chief editor's" Twitter, but fuck Twitter. I just noticed "Report a correction or typo" which has the words "As a local news organization, we are committed to upholding our journalistic standards online, including accuracy." so I just told them now how they violated that standard.

Well fuck, I've become a "Letters to the Editor" writter.

This really could be a bigger problem.

fun fact, on the Model S (2016), before tesla removed solitaire while driving, you could reliably crash the infotainment computer if you dragged the cards around fast enough. Actually, just tapping very fast could do it too. The car would still drive (headlights on, accelerator and breaks worked) but no turn signals. fun prank as the passenger :)

Reply on news.ycombinator.com