Hacker News Reader

Comment by remmargorp64 | Hacker News Reader

On the flip side, how cool is it that cars are now smart enough to diagnose themselves and prevent you from driving before what could potentially be an unsafe situation has a chance to occur?

View on HN

The fraction of serious accidents that are caused by mechanical failure is just 12%. Of mechanically induced accidents, two-thirds were caused by the failure/degradation of the tires, wheels, or brakes.

https://crashstats.nhtsa.dot.gov/Api/Public/ViewPublication/...

So for diagnostics other than tires, wheels, and brakes, we are talking about a system that can prevent at most 4% of accidents, and probably much, much less. Given the huge number of false positives, I'd much rather cars notify the driver in detail (not just a vague "check engine" warning) and let the driver decide how to proceed, rather than irrevocably disabling themselves.

Consider: we could also probably eliminate 20% of accidents by prohibiting drivers from driving in wet conditions, but instead we reasonably instruct drivers to drive more cautiously and accept that some wetness induced accidents will occur nevertheless.

In what world is preventing 4% of serious accidents trivial? That’s likely hundreds or thousands of crashes per year given the numbers involved.

See "probably much, much less" (since diagnostic systems have many false negatives, and only work on some subsystems) and my argument about wet conditions (which causes greater than 4% of accidents). Furthermore, the absolute number of accidents is not a sensible number to look at unless you're also totaling up the absolute number of drivers inconvenience. Both the absolute number of accidents and the absolute number of downsides (wasted time, driver frustration, unnecessary repairs) will be proportional to the number of cars, so you should divide out by the number of cars.

Probably much, much less is meaningless since it means nothing.

I’m pretty sure if there are 20,000 serious accidents per year and some lines of code alone prevent 100 (0.5%) that’s noteworthy.

First, you are not engaging with any of the points in my reply. (Do you disagree that the fraction of accidents that diagnostic systems can plausibly stop is <<4%? We can't tell.)

Second, you keep suggesting that I made claims about this being "trivial" or "noteworthy". I did not. I made a claim about what would be a good system, not the notability of the absolute number of accidents stopped.

>> prohibiting drivers from driving in wet conditions

I will not consider this. Because it sounds ridiculous. Prohibiting whom and how exactly, from driving how/what?, and how wet? And how do you measure that? And at the cost of what?

"Now"? This has been implemented for years - most people just call it "Check engine light" ;)

The check engine light doesn't send a notification directly to your phone with a diagnosis of what's wrong.

No, it doesn't send your car location to the cloud and tells it to everyone involved, true.

The diagnosis however is available for retrieval via industry standard connector - OBD2.

With the right OBDII reader it does.

It also doesn't leave you stranded. Some cars have been driven for tens of thousands of miles with this light on.

I put 60,000 miles on a Jeep with a check engine light. Also spent thousands on trying to fix it at various shops which would replace all sorts of things, to no effect, including about 5 attempts to replace a costly computer board.

Eventually had a redneck guy working out of the back of his pickup truck have a go at it. After hearing about the history he spent hours unplugging, cleaning, and replugging every single wire connection and socket. And the light stayed off.

> And the light stayed off.

Are you sure he didn't just leave that particular light unplugged?

It's a good idea, but such a ruse would be easy to catch. A properly-wired check engine light illuminates whenever accessory power is on and the engine isn't running: i.e., right before you start the car, every time you start the car.

It also doesn’t stop you from driving the car to a mechanic or service station.

They still don't have a diagnosis of what's wrong with the car. They've been told about a whole bunch of things going wrong, including battery-related stuff (which is probably why the car was prevented from driving further!). At least it does seem to be an exceptional case, I'm pretty sure we would know if it wasn't.

So, yes, car manufacturers do exactly this right now. But if you want exact details, or if you have an older MY car you can plug in your own odb-ii reader and download one of the many apps to get those details.

My car just threw a CEL and it was because I put low octane gas in. I knew this but also found this by using the handy Forscan app

The check engine light doesn't prevent you from driving.

True, though some cars fall into “limp mode” making it challenging to drive very far. That typically limits the max RPM if a potential catastrophic drivetrain issue is detected while not also stranding you.

If a car is in limp mode though you probably shouldn't be moving it at all unless you have already committed to writing it off.

I don’t think that’s true, though it may depend on the make and model - the related service manual should be consulted. Limp mode is intentional and is mostly meant to limit function while sensors may be disabled, compromised or uncertain. There’s a sensitivity to drivetrain timing that narrows as RPMs increase (e.g. approach redline), limp mode seems to mostly use sane defaults for everything to avoid catastrophic failure in that type of scenario.

Which people drive around with for weeks without care because it usually feels like a light that indicates the dealer would love to extract some exorbitant service fee from you light.

The check engine light is only for emissions related failures which is why in most cases the car is still driveable. In serious cases like this most cars will flash the CEL but AFAIK that is not legally required. The whole point of the CEL is that when the emission system is compromised its not immediately obvious, in fact the car may even perform better depending on the fault.

So just a nitpick, but the CEL is not exclusively for emissions related codes. There are a wide range of things that can fail that do not affect the emissions output of the car, but will still trigger a code and light. All of them can fail you in an emissions test, but not all of them are emissions related.

Edit: All of them will fail you -> All of them can fail you. Not all codes will trigger an immediate fail in CA smog testing anyways, depending on if they are manufacturer specific, or what they effect. I passed my most recent smog with a few codes.

I have the exact opposite reaction. There are very few things that can go wrong with a car that are dangerous enough for the car to automatically refuse to be driven. For what is the equivalent of the check engine light coming on, the car refuses to even be driven to the mechanic. I've had a history of older jalopies, but I've driven for probably a decade with a check engine light on. The odds of having your Tesla deciding to strand you in the middle of nowhere seems a lot higher than it saving itself from bursting into flames.

What about this situation said it was unsafe to drive? Not able to drive and not safe to are different things.

I’m reminded that many gas engines can drop into a degraded mode and still run when there is a problem like a sensor failing. Even a small oil leak can still be driven with. Degraded but not stranded.

That was my thought too. And the article even says that the first evidence they had of a potential problem was when it was charging slowly earlier, so this is less "suffered a failure while parked" and more "suffered a failure while driving and diagnosed itself while parked".

Depends what the false positive rate is, and it's no excuse for proper engineering and quality control.

I don’t think it’s cool at all that this happened on a car with 5200 miles. That’s a bug, not a feature.

If they can do all this real-time surveillance, why not have a conditional somewhere to send a loaner for the case when you sold someone a lemon?

Reply on news.ycombinator.com