The Tesla should not have stopped but there is a cultural problem of driving way too close to the car in front of you. There are other justified scenarios why a car would have had to slam on the brakes at that same spot and the pileup would have still occurred.
Curiously the vehicle behind the Tesla barely hit it whereas the drivers of the cars further back in the pileup seemed to have paid less attention.
There would have been multiple ways for the cars further back to prevent a pileup: Drive slower, keep more distance. If you speed, have both hands on the wheel, look at the road, and be ready to brake quickly.
It does sound like there were no severe injuries, so you could argue that people kept enough distance to avoid serious injuries.
There are lighting conditions that could play a role here since this was when people were entering the tunnel. This might also be something that freaked out the Tesla. But it is hard to know without being in the situation.
Even if all the vehicles following each had the same following distance and reaction time, each of them is going to encounter a vehicle stopping more quickly than the one in front of them did. So even if everyone performs similarly, the cars at the back are going to hit harder.
It’s really hard to tell, but to me it looks like the one behind that one managed to stop in time. But only with the tiniest margin. Action further back was much more dramatic, that’s for sure.
You're right. The first rear-ending may have been unavoidable. However, you can see from the rear surveillance camera that subsequent vehicles could have stopped if everyone agreed to keep larger following distances. As mentioned elsewhere in the thread, there is the game-theoretical issue of other drivers cutting in when large gaps form so it would take some coordination on a technology / policy / cultural level to resolve the issue.
The Tesla changed lanes and instantly became a stopped object. The car that hit it did the same, and so did every subsequent car. The follow distance to avoid a collision at that point is about 10 car lengths, or roughly the average stopping distance.
The Tesla looks like it slowed down 200+ feet over 4 seconds. That's far from slamming on the brakes. It's also, however, very far from expected behavior which could explain why the drivers behind didn't slow down enough in time. They weren't expecting the Tesla to come to a complete stop. I'd guess if the other car was a Tesla it probably would have stopped in time to prevent the collision because it's basing its decisions on the environment not based on prior experience driving in that particular spot.
I'd also argue that the Tesla pulling over into the left lane (merge 1 lane left vs 3 lanes to the right) is the right behavior in a situation that warrants stopping. Try to get to the side of the road so other cars can pass.
that is, in fact, slamming on the brakes. your estimate of 200+ft over 4 seconds, assuming perhaps a low highway speed of 60mph, is faster than many cars are capable of stopping, even assuming zero reaction time.
It's ok, the car may have cut people off and caused an accident, but if you look at it on the basis of the the number of fatalities/mile, cutting off and causing a minor accident is no big deal. Clearly keeping Tesla's stock price up is a bigger deal.
Coming to a full stop in the left lane is really last-resort stuff. That is the speed/passing lane (in most jurisdictions, all lands are speed/passing around the Bay Area and especially on the bridges).
It's one of those things where if you don't close the gap you'll get other cars constantly trying to jump into it. Same with not speeding, you'll get overtaken constantly. Tends to make things more dangerous for you. I'm not sure what the answer is. It'll be hard to generate a consensus without some major policing.
If you are driving in the "slow" lanes this is not the case. I generally cruise in the right or second to right lane with a large gap in front of me and it is certainly not more dangerous. Certainly faster traffic is passing on my left but not in a way that impacts me.
I'm surprised this isn't higher up. I think a lot of people retcon their safe driving behaviors to what they learned in a driving test but not to what they do in practice. Or maybe a lot of people answering haven't driven in major cities? But yeah, drive in LA and see how long you can keep 3+ car lengths in front of you. I hate driving in LA.
> It's one of those things where if you don't close the gap you'll get other cars constantly trying to jump into it.
In general I am ok with people merging in the gap, but some people are assholes about it by merging too close to me that I have to brake. I don't understand why people merge when they know the person behind in the other lane will have to slow down to let them merge.
This is a fine excuse to not maintain safe distance, until the car in front slows or stops all of a sudden and you rear end the car, then it’s your problem.
I haven't read statute, but my understanding has always been that if you run into the car in front of you, it is your fault, period. If you didn't have time to stop, you are "Following too close for conditions", regardless of what those conditions are. I agree that culturally a lot of people in such situations say "It's not my fault, they stopped right in front of me!", but as you correctly point out, there are legitimate reasons for someone to come to a sudden stop, and as a driver it is YOUR responsibility to make sure you are ready for that possibility.
You get situations of people backing into you which looks like you ran into them. I think it's on the absence of other evidence the car with front damage is presumed guilty but if you have additional evidence like a dashcam you can challenge that presumption.
Good luck proving that without evidence. You can see a number of youtube videos of people trying to perpetuate break checks as an insurance scam. Sometimes break checking alone doesn't work so they reverse into you. Relying too hard on simple heuristics creates these problems and additional evidence such as dash cams can help.
This is fair, I should have been more specific. "Following too close" implies "following", doesn't include people unexpectedly pulling into a stream of traffic. In this instance the Tesla changing lanes would mitigate "Following too close" and it would likely be "Unsafe lane change". That's only true for the first vehicle though, not the rest of the pile up.
Well "unsafe lane change" can be totally legit as well. I for once had to do no think abrupt lane change when I saw some largish object appear seemingly out of nowhere and it was flying towards my front window. Still do not know what it was. Abrupt lane change did not cause any problem in this case but in theory it could have.
While you may have had a legit reason, there's no way you can expect a driver in the lane you abruptly changed into to anticipate this behavior. If you hit another car in your abrupt change, you are still at fault even if you have a legit reason.
Of no fault of their own, the car you hit trying to avoid something means you didn't avoid them. The car you hit did nothing wrong, but you did. Insurance will place you at fault at a minimum of partial fault. You will not be blameless in your example if you do hit someone/something
Yeah, I know where you're coming from, but...when you've taken as many defensive driving classes I have from my younger days (we all know why), then some of the things they teach definitely start to stick with you. If you are forced to drive in rush hour traffic every day, there's definitely not much you can do. So, stop driving in rush hour traffic!! After that, you stop driving next to cars. You place yourself in the gaps even if that means slowing down. And that was my biggest Doh! moment was learning to accept that slowing down has so many benefits. One of the take aways form defensive driving was always leaving yourself an out. Being cognizant of not having an out for unexpected need for avoidance maneuvering will go a long way of avoiding the situations you experienced.
FWIW the defensive driving guidance in that scenario is to stay in your lane and brake -- this is most commonly encountered when you see a deer in the roadway, you will almost always do more damage to your passengers, other motorists, and your own car by swerving (and potentially risking a rollover or running off the road) than by maintaining your course and hitting the object at a lower speed
> maintaining your course and hitting the object at a lower speed
Growing up in the Midwest, I remember being taught (not sure if this was official or something I got from my father, older brother, or maybe an uncle) to quickly assess if I believed that I would hit the deer or not.
If I was going to hit the deer, regardless of my actions, I was to accelerate into it.
If I could brake and avoid it, I was to brake.
I've been in that situation a few times. In the case of a collision with a deer, I accelerated into it and only took damage to the front passenger side panel. I do believe that it helped prevent a bad situation from becoming worse.
In the other cases, I was able to avoid a collision by braking.
Later in life, I did swerve once to avoid a large animal that I mistook for a person walking on a very dark, very rural road. My car flipped and slid on it's top ~100 ft. before coming to a stop. Myself and all 3 passengers walked away without incident but I still feel fortunate that we all did these many years later. Swerving is almost always the wrong answer.
I always wondered about that recommendation to accelerate into it. It seems to me it would only end up making things worse by adding more energy into the collision than less.
Sadly it does happen. There are horror stories of live deer going through the windshield, doing a number on the occupants both with their arrival and their attempts to escape.
Animals do not fly towards you. Tires do and they kill people who brake and wait toll it hits you. Becaus all happened very fast I can not bet what the flying object was but it was most likely a tire.
The object flying towards my window was most likely a tire. Sorry but I am not suicidal. Brake and wait till it hits you is worthy of Darwin award. There are countless examples of flying tires killing people on the road.
> That's only true for the first vehicle though, not the rest of the pile up.
I don't think that's right. Because safe following distance can only prevent you from slamming into the car in front of you if it breaks hard. If it hits something and stops way before the breaking distance, there's little you can do.
The stopping distance at 60mph/96kmh (~28m/s) for a typical car seems to be 133ft/40m. So you are right, it is actually significantly smaller than the recommended 3-4 seconds rule with a relatively good margin. It also amounts to ~5 car lengths at that speed, which I rarely see in practice on active highways.
I don't know what area you're from but I cannot imagine the DMV advocates for a following distance where you know you wouldn't be able to stop in time. I'd love a citation for that.
> If the car in front of you hits something and comes to an extremely abrupt stop and you don't have the stopping distance to stop, you're too close.
It is physically impossible for a car to come to an instantaneous stop for no reason. It would have to e.g. drive into a concrete wall, but if there's a wall across the freeway you should've seen that from afar.
You only really need to have enough distance to be able to notice the car ahead braking and start braking yourself.
In Australia it's true. It doesn't matter the circumstance, if you hit a vehicle from behind, you are at fault. A friend hit a car with no brake lights and no turn signal that stopped in the driving lane. Doesn't matter.
Personally speaking, in this scenario, with an enormous stream of one way traffic on one of the few roads leading from San Francisco, I hope circumstances are taken into account.
All Teslas could do this at any moment, is my takeaway.
It’s the same in Europe. The car hitting the suddenly stopped car is at fault, every car after that is at fault as well.
You’re to maintain distance and speed that would allow you to stop in time without an accident at all times.
Same goes for potholes, if you hit a pothole and bend a rim or something, you can get a fine for driving without maintaining proper speed. Then you can use that fine to sue your local government for having a pothole there.
In my part of Europe, the lane changing car is in fault, and the cars behind (expect the first of them).
I don't know where it would be allowed to stop in the middle of a highway lane for no apparent reason, especially after a lane change that did not result in a proper distance to the car behind.
Australian here. I was in a multi car pileup years ago. The car in front of me stopped unexpectedly on a freeway's onramp. Everyone was accelerating at the time. I slammed on the brakes. I decided I would stop in time looked in the rear view mirror to see if the guy behind me would ram me. I still remember it vividly - it was a white Commodore. Its brakes were on hard, nose almost grazing the road but it looked like it would be OK so I turned by attention forward again.
The next thing I remember is hearing an almighty crash, followed by the Commodore crashing into us, followed by another crash as we were pushed into the car in front. A moment later a motor cycle flew by the left hand side. It has evidently been unable to stop and and swerved right to a full frontal but was then off balance. The handle bars then creased the left hand panel of every car. The bike was only slightly damaged, and he drove off.
The police arrived. After a few questions, they determined it was the last car in the chain that wasn't paying enough attention and had pushed everyone else the car in front. They were taken to the station. The rest of us heard nothing more of it. The insurance companies paid out without much of a fight after getting the police report.
But that's about the limit of it as far as I can tell. In Australia, if you going though a intersection with a green light and hit someone who has run the red, your up for dangerous driving. It's not "one person made a mistake". As far as the law is concerned you both weren't paying attention.
Only for the first car, though. A lot of the impact of this story isn't the initial fender-bender-scale collision for which the Tesla driver[1] is clearly at fault, it's the horrifying stuff that happened in the resulting "8 car pile up" that surely was due to someone else going way too fast.
If the first crash were the only one, this would never have reached the level of media attention. If the same pileup was caused by a garden variety grandparent, no one would care.
[1] I'd suggest caution in attribution here. The only evidence that FSD was involved is the at-fault driver's statement to the police. I'm not going to say it wasn't operating, but I will say that I've never seen behavior like this from FSD. It's extremely cautious at lane changes, often annoyingly so. And while I've seen the same "phantom braking" everyone has, it's a pulse, not this kind of "bring the vehicle to a stop" behavior. Given the assertiveness of the controls, I'd say it's just as likely that the driver decided to swing into the left lane on their own and then hit the brakes instead of the accelerator by accident (which happens, again, all the time!).
Exactly. If a driver pulls into your lane right in front of you and then slams on their brakes then that driver is at fault. You may need eyewitnesses or dashcam evidence to prove that is what happened, though.
In some states (perhaps even most or all) this is true. It is your legal responsibility to drive a safe enough distance to the car in front so you can avoid collision always if they stop short for any reason.
I had a driver brake randomly and I rear ended them. They ended up admitting it and my insurance went from being mad at me to covering everything. Granted I barely tapped them but still, I think the law is a bit fuzzier than that.
That is exactly how it is handled in Germany. Actually not keeping enough distance on the Autobahn is one of things with serious fines overhere. I know both, the "your fault to hit car in front of you" and the "please pay 200 bucls and hand over your liscense for month as you were to close and we have video" aspects of it.
I couldn't watch the video so can't comment directly on this case but it boggles the mind how close people follow others on the road.
The distance you have between you and the car in front on you is one of the only safety zones you can control on the road. People should cherish it and let it blossom into wide open space.
You often have no control over the left, right and rear. Why would you want to take away that front space as well?
aside, people drive somewhat close in the bay area, but also in my experience it's so much worse in both LA and in Austin, and in TX so many of them are 6000lb Ford F150s.
In LA if you try to have distance someone will fill in the spot. I've experienced this in the Bay too. While I agree with many of these comments about people driving too close, I think it ends up being too simple and naive to real world conditions. This behavior tends to be true in any major city but not so outside.
Yeah I understand the "fill the spot" issue all too well. This is a behavior that our police/legal system is failing to address adequately. The emergent behavior is not the safe, legal, or even optimal behavior for traffic flow/reduction (afaik). But as I've observed in America, it's more about getting 1 car ahead of others than it is about having optimal outcomes across a large proportion of all traffic.
I have a hypothesis that this is another emergent pattern. I drive a lot between WA, OR, and CA. One thing I notice is that in OR I can easily pass people AND that if I'm in the left lane approaching someone there's a fairly high likelihood that they move right. Maybe 1 in 10 people don't move. But once I hit California I end up zig zagging much more, passing people who are in the left lane going below the speed limit (especially in central California). But I don't think the probability changed, but rather the population. Oregon is easier to rive in than Washington which is easier to drive in than California. I'm pretty sure this is just a small number of bad actors who are a constant percentage and their effects scale (probably also providing a feedback loop). I try to slightly combat this by calling the left most lane the "passing lane" rather than the "fast lane." A lot of people treat the left lane like kids treat lightup shoes: they think it makes them go faster.
And don't get me started on elephant racing. I frequently wonder the economic and environmental impact of just this singular type of event in central California.
Can remember there were campaigns in my country in the 90's to increase distance between cars. The rule was to count 1001 1002 1003, so basically you should have a 3 second gap to the car in front. Not sure if it worked, but apparently it made an impression on me, so I constantly count 1001 1002 1003 when driving on roads with a higher speed limit.
You basically can't, realistically, do that in dense city highway driving, which looks like what this is. Like, we still say the driver who hits the back is "at fault" in these kinds of pile-ups but it's entirely a "there, but for the grace of God, go I" situation. They were just unlucky. 99% of drivers on that road drive the same way they were, and the ones who don't are probably driving in a way that makes things less safe.
In several countries in Europe they occasionally paint markings on the road (chevrons or dots or similar) with a sign like "Keep apart 2 chevrons". Some places have electronic speed limit signs to reduce the maximum speed in heavy traffic, which reduces risk and improves flow.
There is very heavy traffic on both sides, so it wouldn't be unusual for the actual speed to be 50mph, 40mph or even less. The second nearest car is braking, and there's a "Queues Likely" sign a little further -- I found the place:
StreetView gives a much better impression of the chevrons; the Alamy photo has an unusual choice of lens. The "M56 B 45.0" [km] milepost is way ahead of the chevron sign.
They'd have a revolt on their hands in short order, around here, and I suspect in most places. You can't handle enough traffic on the parts of our highways that go through the city, while maintaining the kinds of gaps that would be necessary to avoid these sorts of wrecks (well, the later pile-ons, anyway). Commute times would shoot up and areas that don't usually experience congestion would become intolerable around rush hour, if people really started leaving that much space, consistently.
Which is, perhaps, one of many arguments against having so many people on the roads to begin with, but that's what we have.
> You basically can't, realistically, do that in dense city highway driving, which looks like what this is.
Yeah you can.
The maximum carrying capacity of a single lane of highway traffic is around 1900 cars per hour (that is from real-world measurements of what human drivers can actually achieve). At 70 mph that is 194 feet per vehicle. Even assuming something like a 20 foot long F250 or something that leaves 174 feet of spacing between vehicles. Which is nearly 9 vehicle lengths and nearly 2 seconds of room. Up until you hit that point you're not impeding any traffic flow by adding more space. Almost nobody in this video is leaving that much room.
What causes traffic congestion and causes that 1900 vehicles per hour number to be that small is the people who are tailgating and when someone brakes everyone behind them has insufficient following distance and has to brake harder and harder until you've got vehicles nearly stopped on the freeway and have formed a traffic compression wave (if you haven't just created an accident and snarled traffic even worse).
No, you generally shouldn't drive that anomalously. A driver will pass you and insert themselves in front of you. You will fall back to create distance. Another driver will pass you and insert themselves in front of you. You create constant aggravation and churn around your car.
Source: I spent several months last year helping someone learn to drive in the Bay Area, and they insisted on leaving the recommended gap between cars as they prepared for their driving test.
That doesn't matter. And if you're not in the left lane usually people merging in front of you are heading for the "fast lane" (which is typically even more congested and slower) or heading for the exit lane. Just let them go on their way, into and just as rapidly out of your life.
And if you don't let yourself feel "aggravation and churn" then people will just merge into and later leave your lane and it just becomes normal driving without any ego involvements.
Source: the past 30 years of driving in Seattle traffic that I've done where I've stopped being a tailgating idiot and opened up more space and stopped caring about people "cutting me off".
100 people merging into 170 feet of space in front of you[*] is dramatically safer than 1 person trying to squeeze their way into the 20-30 feet of space you're leaving in front of you when you're tailgating.
They taught this to me in driver's ed 15 years ago, just start counting when you go past a sign or something.
The rule actually works great where I live (a rural area) but if I wander over towards chicagoland I will be punished with constant cutoffs. Also there's too many cloverleafs over there jeeze.
When you leave adequate space (especially) in city driving, another driver will always occupy that space. You back off to create more space, another drive enters that space. Rinse, repeat. Source: Boston driver.
Source: South Bay driver. I was going to say that but was sure I will find a similar comment. I start my day by trying to be a safe driver but as they drive progresses and more and more people jumping in the free space in front of me, I eventually change the mode to not be risky but don't allow space for others.
Re not watching video: This is the most crowded road, after a downhill slope, in a major metropolitan, after a mile of bridge driving, reaching a reduced visibility tunnel, with one way traffic, and 5+ lanes.
There's too much traffic. It backs up into the streets, before this bridge + tunnel regularly. The speed is already restricted.
I think if the Tesla had been avoiding a hazard, there would be a lot more gray area for FSD to hide in. It’s true that sudden braking and swerving is sometimes the least-bad thing for a driver to do, but if a human driver showed a pattern of repeatedly performing that maneuver for no reason, at some point a reasonable person would have to conclude that that driver is at least partly at fault for the inevitable crash that happens.
Did the driver have his hand on the the controls at the time of the accident? FSD = Faulty Self Driving or Fairly Self Driving or Fully Stopped Driving.
That was absolutely not the case in this accident. The Tesla swerved in front of the car in the next lane and then suddenly slammed on the breaks. The two cars behind that one were driving several car lengths back. There was no way for this accident to involve less than 4 cars.
False. There exists a following distance and amount of paying attention that would have involved 0 cars hitting the tesla. If you look really closely at it at least 3 cars (maybe 4) stopped with next to no incident (it's not discernible if they even touched). It's the 5th or 6th car that shoves everyone into everyone else (and quite dramatically gets under the car infront of them and pops them up)
> that while conscious and not in any kind of distress?
I'd be less understanding, yes.
But at this point it's presumable that the autopilot was in "distress" (avoiding a perceived issue). And when such a thing comes up, real or false, the drivers behind need to be ready to react. And several of the cars did so quite well actually. The only reason this is a dramatic incident is because of the like 5th or 6th car that hits everyone else hard.
+1 to this, I ride motorcycle and lane split (in California) and frequently see people doing 80miles an hour looking down at the phone in their lap. This means even their peripheral vision is out of play, at least if they held the phone at 12 oclock over the steering wheel there'd be some non focused vision... smh
Honestly I think Carplay would save more lives if they allowed reading messages via infotainment screen instead of me having to tap my phone (voice messaging doesn't exist in most languages and Apple wouldn't Google to step in, lol)
The design flaw is Tesla's fault, yes. The chain reaction is not. And it's not fair to blame the design flaw for 8 cars. At most 1, but we see that 1 behind it actually more or less stopped.
This argument bears a lot of resemblance with victim-blaming phrases, like:
“She shouldn’t have been raped, but did you see the dress she was wearing on a dark street?”
Its not even remotely close. Women have a right to wear what they want without being raped for it. People don't have to drive dangerously too close and expect to not get in an accident.
Not my understanding of this situation ..they pulled into another occupied lane and hit the brakes...so neither generalization is so simple a case to support.
I agree with that but it seems like the Tesla basically cut off the driver to their left while changing lanes (while signaling) but also coming to a stop. Really weird behavior to see, of course, but this honestly doesn't seem like much of a "safe driving distance" thing to me.
"But" nothing. This is a Technology Failure. Whomever worked on this software should face consequences.
Hitting the brakes in the tunnel for no reason is a "never to this" kind of move.
"But [...] other scenarios" is not the situation here. There is no escaped pet anaconda slithering across the tunnel floor. There's no Godzilla breathing fire inside the tunnel. It's just a car with no obstructions in front of it, none visible to any other drivers, and it's choosing to hit the brakes.
There is no excuse.
It's at the bottom of a downhill slope. There's a limited view of what's up ahead because the slope starts to incline inside that tunnel. Cars start to speed up. There is typically a massive amount of traffic on the bridge.
> shows a Tesla Model S vehicle changing lanes and then abruptly braking in the far-left lane of the San Francisco Bay Bridge
While it is true that there is a culture of following to closely, I don't think it applies here. When a car merges into your lane you should not abruptly change your speed to adjust the distance, but slowly increase it. Otherwise you get huge ripple effects and people 1km back will essentially be in stop and go mode.
The video is literally evidence of the effects of following too closely. From the back angle, there are 4 vehicles starting with the white truck in the leftmost lane that pass through within 1.5 seconds; a few are only a few hundred milliseconds apart! If you switch back to the front view you see that’s exactly the cluster that piles up onto the safety braked vehicles ahead of them, because the white truck they’ve been tailgating has swerved into a moving lane and exposed the stopped traffic.
Everyone beside perhaps the first car that had an incursion into its lane (but arguably failed to reduce speed to avoid an accident with a vehicle braking and signaling their lane) are WAY too close for the speed. Any speed reductions in those lanes would necessitate heavy braking which would cause exactly the ripple effects you’re referring to under normal conditions because there’s no buffer whatsoever to gradually reduce speed/increase distance either.
Drives way too fast all time, even in neighborhoods.
Drives way too aggressive.
Drives way too entitled. I’ve seen some crazy stuff from people driving a Tesla. Driving on the shoulder when traffic backs up. Zipping around people to steal parking spots.
I did not see any emergency stopping lane on this stretch of the road. You would bump onto the narrow pedestrian walkway. So those people were wrong. We absolutely should keep more safety gap.
A great example of how FSD/Autopilot is hugely flawed today. The car just does not give you clear indications of when or _why_ it is slowing down/swerving erratically in the lane when nothing is around. I get that about once a week.
This is 100% the drivers fault though, it is extremely easy to override Autopilot. Tap the brakes will disable the entire system, press the accelerator will keep lane keeping on but override the speed, and moving the steering wheel will disable lane keeping but keep cruise control on.
More often than not, I have my foot on the accelerator and not the brake when I am using AP. I don't drive a lot in general, but over the tens of thousands of miles on AP I had it hit the brakes hard enough to be a problem only once. It's definitely randomly slowed down many times, but with your foot on the accelerator it's very easy to override before it drops more than a couple MPH.
Now, I've also heard people say that the phantom braking issue can be really violent and so it's very possible I've just never experienced true phantom braking, but based on how many of the drivers behind the initial wreck slammed into each other it's clear many folks do not pay enough attention or leave enough following distance to be safe.
If you use it "correctly" (very subjective) it is a great driver assistant. On long road trips I am way less tired and can focus more on making sure my environment is safe (animals entering road) instead of focusing on lane keeping, etc.
Most lane keeping/cruise control setups today from other OEMs would also be sufficient in most of the scenarios I enjoy AP though.
You really have to be vigilant. In the summer of last year, my wife and I went to the Bay Area and I rented a Tesla with FSD. While driving from Santa Cruz to Redwood City, I decided to put on the FSD. The car drove fine for a bit then randomly decided to switch to the left lane despite there being no cars in front of us. While trying to figure out what was going on I start to hear a honking as there was a car in my blind spot that was trying to overtake me at the moment the car tried to switch lanes. I grabbed the wheel and luckily nothing happened. Never trusting FSD after that.
I've had autopilot try to execute a lane change at the same time a driver two lanes over was trying to do the same thing into the same spot. It was one of those unavoidable situations, but Autopilot corrected quickly back to its original lane without me needing to intervene.
This is 100% the driver's fault for deciding to enable Tesla's "Autopilot" on a public road in late 2022. No one should be surprised when "Autopilot" shits the bed.
Why? Cruise is driving around the streets of SF with nobody in the drivers seat. Why shouldnt someone be surprised that their autopilot works?
What we should be more surprised about is the total failure of our government agencies to be with the times and create proper regulation and specifications of operation and testing.
Cruise and Tesla use completely different technologies for the self driving solutions. Cruise self driving is built on solid technological foundations, whereas Tesla self driving is built on hype and hope, surrounded by a cult of personality. I believe that's why the parent comment included the date - to indicate that we should all know by now in 2022 that Musk can't deliver on their claims. That doesn't mean others won't.
> This is 100% the drivers fault though, it is extremely easy to override Autopilot
That's speculation. Do we know if the car had suddenly lost the ability to continue driving? Was the car pulling to the closest side of the road? If you lose power in a gas powered vehicle, have a flat, or some other failure when you're in the far left lane, what should be done?
From the video, it looks like a lot of people are following way too closely. Far less than the 3-second rule would allow for.
The first vehicle behind that crashed got caught by the sudden lane change.
The second vehicle avoided rear ending them. You can see the weight and suspension settle just before it gets hit from behind.
Three was slowing and trying to avoid the crash (it shifts lanes and makes it past), while the fourth was in a rush to get past three (you can barely see it, that's how close they were following).
While the quick pull to the left side of the road probably initiated things and caused the first vehicle to rear end the Tesla, the pile up is the fault of the fourth car (which can barely be seen behind the pickup that managed to avoid it), and all the others that were clearly following way too closely.
Safely merge right and stop in the shoulder if possible, but in heavy traffic like this put the hazard lights on and slowly stop without changing lanes. Call the police -- do not get out of the car while there are still lanes with moving traffic.
In wealthier-Europe the police would activate electronic "60km/h", "lane closed" and similar signs. It's a long time since I drove in the USA, I can't remember if you have these.
> have a flat
If you must stop quicker pulse the brakes if possible (the flashing lights draw more attention), but again, don't change lanes. Changing lanes caused the problem, which was compounded by the too-close following vehicles. Some of those vehicles also changed lanes dangerously, when they should have stopped in their current lane.
For what it's worth, a vehicle two ahead of me crashed into the central barrier in my final driving lesson.
It's an anti-pattern, this product. And, I would not say that it's 100% the driver's fault.
The feature is named autopilot or full self-driving, it's misnamed. All the marketing and hype puts prospective drivers into the nod and wink mindset that they don't have to drive when they have a self-driving vehicle.
It gives the driver the wrong impression.
Only after experience would you have accumulated the wisdom to treat it a lot more conservatively.
In my opinion if I were to own one of these things I would be Embarrassed and offended at the notion that I spent tens of thousands of dollars on a piece of software in a special vehicle only to have it not do what I expected it to do.
Vehicles unfortunately have to come to a stop all the time on a highway for various reasons. This is clearly due the cars going too fast and too close from behind.
> This is clearly due the cars going too fast and too close from behind.
This is due to both - buggy software being tested on public roads and people going too fast and driving close behind. The first two collisions couldn't have been avoided because of how abrupt Tesla changed the lane and stopped, thats on Tesla.
Is this what happens when the driver ignores the steering wheel nag warning? I've never let mine get to that point but I think the car pulls over and stops? (Not saying what the Tesla did here was ok just curious if that's what it was doing)
In just autopilot, I don't think it actually pulls over. I tried it once, and it put on the hazard lights and began to coast to a stop. I was able to just turn off the hazards and use the pedal to resume driving as normal, so you can override at any time. It does penalize you by preventing you from turning AP back on during that drive.
ah interesting. I guess the lack of hazards here might indicate this wasn't a case of the driver ignoring the warnings. They should have been ready to hit the gas regardless.
Venturing off-topic here but I can't imagine owning a vehicle that sees fit to "penalize" me as if it's my parent or credit card issuer rather than an inanimate object I own to use as I see fit.
Driver was obviously not paying attention. How does someone notice their car braking significantly and not just press the accelerator to counter it? AP will disengage the brake at that point.
I think it’s made extremely clear that you need to be paying attention during FSD/AP. Car may have partial blame for doing something unexpected but driver didn’t do anything to get out of that situation.
>Teslas accounted for almost 70 percent of 329 crashes in which advanced driver assistance systems were involved, as well as a majority of fatalities and serious injuries associated with them, the data shows.
This is pretty disingenuous given that I believe Teslas have probably covered by far the most miles with an "advanced driver assistance system".
The Tesla definitely screwed up here, and I'm not sure you could have prevented the initial crash, but the pile up behind it is entirely the other drivers' fault.
...I'm not sure you could have prevented the initial crash, but the pile up behind it is entirely the other drivers' fault.
If other drivers should have seen and avoided the pileup, then it seems reasonable that the Tesla driver should have seen the unexpected braking and overridden it.
In both cases, humans failed to react fast enough to something that required active focus and fast, split-second reaction.
> This is pretty disingenuous given that I believe Teslas have probably covered by far the most miles with an "advanced driver assistance system".
On top of that most other manufacturers don't have telemetry about it either and instead rely on dealerships to collect that information.
> The Tesla definitely screwed up here, and I'm not sure you could have prevented the initial crash, but the pile up behind it is entirely the other drivers' fault.
I'm still confused about the whole situation.
Police report mentions FSD v11, which is only for employees so far.
FSD v11 is what makes the freeway "FSD", otherwise it's their old "we go straight in lane, maybe take exit or change lane" code...which wouldn't be FSD.
So either an employee crashed with internal only versions, or the police and owner are just incompetent about what they actually have.
I'm not entirely convinced Teslas have had the most miles traveled with ADAS systems. ADAS systems have been around longer than Tesla has even existed. Ford had ADAS systems on the Taurus before Tesla released the Model S and had ADAS functionality available on the F-150 the year after Telsa enabled Autopilot.
It might be true that Tesla does have the most miles, but I wouldn't say its a sure bet. I'd love to see some actual data on it to know instead of making the assumption.
It sounds like ADAS is a term of art from a particular regulation. Do you know if it would only cover things that include at least autosteer or does even a basic radar based cruise control count?
This dataset for crashes in question is about the NHTSA ADAS crash data. For a while the reporting criteria was Level 1 and 2, these days its just Level 2 and above. Level 2 is adaptive cruise control + lane assist. Every Honda Accord sold today meets this standard. Tesla sold ~405,000 cars last year, far more than any other year in their history. Honda sold ~157k Accords in 2022, all with SAE Level 2 standard. Honda alone moved almost 39% of the same volume of Level 2 cars with a single one of their products and have been selling these features for quite a while. Add in all the other auto makers selling cars with Level 2 and there's more likely way more non-Teslas with SAE Level 2 autonomy on the roads than there are Teslas.
I actually figured it would have been a bit bigger of a lead for Tesla, but showing that figure about Accords really made me doubt that Tesla leads in all ADAS SAE Level 2 miles traveled.
Also, FWIW "advanced driver assistance system" (ADAS) is definitely a term of art which applies to a ton of technologies. Those technologies include radar-based adaptive cruise control, collision avoidance systems, advanced traction control systems, traffic sign recognition, and other technologies. The first Taurus with adaptive cruise control would be considered an SAE Level 1 autonomous car with its advanced driver assistance system. So even a car with just emergency braking technologies but otherwise an SAE Level 0 would be considered as a car with ADAS on board, even though it might only has the ultra basic cruise control.
> This is pretty disingenuous given that I believe Teslas have probably covered by far the most miles with an "advanced driver assistance system".
I'm not sure that this is true given almost every Honda sold since 2019-ish (and a big proportion for years prior to that) has had ADAS features included and enabled by default, most notably Automated Emergency Braking.
You might be on to something there. I wonder if it was a vision-only version.
Anecdotally, my wife and I used FSD out in Oregon while on vacation a year ago and had these these "phantom brakes" occur two times on the same stretch of road. It was very sunny day through the gorge to Hood River from Portland and I'm thinking there definitely would have been sharp contrasting shadows from the trees and sunlight.
Is it possible that this is the result of the driver ignoring the attention "nags"? Seems more plausible than FSD having a brain fart (unless many other FSD Teslas are making that same decision in that same spot). But even so, is Tesla's fail-mode really that unsophisticated? Pulling off into the left lane, even if it's the nearest off-lane, seems problematic, especially if it's in front of another vehicle who is at full speed b/c it had (until that moment) a clear road.
I don't expect FSD to be able to find its way to a proper rest stop or off-ramp, but seems like a less dramatic shutdown mode should be possible, e.g. driving a couple of miles at moderate speed until there's a safer pull-off opportunity.
It is strange that if it is the case of driver-inattention-disengagement, that the Tesla wouldn't at least put on hazard lights. Maybe it wouldn't be enough in this situation, but in most cases, it would be a more obvious signal that the left lane change is an emergency maneuver.
I haven’t looked at my tail lights, but in the US most cars have used the same bulb for hazard and brake. So the bulbs are already fully illuminated.
Cars meant for the rest of the world have separate yellow lights in back and you could flash those. I think they are adding a rapid burst flash for alert with those.
It’s the visibility problem at the entrance to a tunnel. While you’re in the bright light, it’s harder to see into the shadow. I know that theoretically you should slow down, but practically nobody does, so it’s unsafe to suddenly break during a brief loss of visibility, as demonstrated.
I'm sorry, but this is an 8 vehicle pile up from people following too close (or paying too little attention). At least where I'm from if you run into the back of a car it's your fault. It means you followed too close, failed to react to a situation in front of you.
Should the car have stopped there, no?
But is it that car's fault that others ran into the back of it? No.
The Tesla changed lanes with very little clearance to the car behind it and then immediately braked. The first car that rear-ended the Tesla could not have avoided the accident short of mind-reading.
And the car behind it braked with minimal impact (hard to tell, but you dont see the tesla jump in any major way like the subsequent cars do)... So it didn't completely cut off the car behind it.
> short of mind-reading.
This is called defensive driving and it involves running "what if scenarios" and being prepared for even those. "What if the tesla has a tire blow out and comes in my lane" Is something that can be thought of any brand of car.
It absolutely is the Tesla's fault for coming to an illegal stop in response to no event ahead of them during a lane change which is clearly demonstrated in video. The rest of your comment is approaching victim blaming.
Correct, but if you read and watched the video you would see the Tesla changes lanes and slows to a stop in response to *zero* events ahead of them.
It is a neat fact to bring up but that doesn't apply here. The Tesla is still at fault and you would be hard pressed to find any adjuster or courts agreeing with you thanks to this excellent video evidence.
My point is if someone were to respond "I thought I saw a deer/child/car swalloing pothole" they wouldnt be at fault. It's their right to avoid perceived issues. And its not like the tesla sufficiently cut off the stopping distance for the car behind it (one lane to the left) that car was able to brake substantially/sufficiently (hard to tell exactly in the video).
Luckily it's not up the adjuster, it's up the courts.
Fortunately, the law follows a "reasonable person" standard, not a "the AI tried its best" standard. No reasonable person would have perceived any dangerous situation in these circumstances. Any person who suddenly slowed to a crawl and moved into the far left lane on a fast-moving highway for no discernable reason would be found to be substantially at fault here.
I think it is both. It seems to me that initiation of this mess was Tesla's fault. At the same time this pile up was fault of later cars not paying attention.
Seems like a very strange move. The Tesla seems to signal a lane change and abruptly slow down at the same time on a highway with no lights, intersections, turns, or obstacles.
This was precisely the type of driving I figured would be easy to automate - rather than deal with urban settings, pedestrians, bicycles, etc.
Tesla's systems sometimes see phantom obstacles (e.g. shadows or reflections). Trying to change lanes away from them and then braking when they cannot isn't unusual at all. One of the reasons it is surprising regulators let this "beta" test be conducted on public roads with other drivers who weren't asked nor agreed to the disclaimers.
The problem of driving a robot straight through a corridor while avoiding obstacles is Robotics 101 stuff. With the right sensors. This scenario calls for long rang distance detection, which is the natural wheelhouse of LIDAR.
Unfortunately, Tesla seems to have something to prove when it comes to LIDAR, and stubbornly refuses to adopt the technology, despite crashes and even deaths that have occurred, and general industry understanding that it's the enabling technology for driverless cars.
Is it legal to create a two-ton machine with admittedly beta-quality software, and a sensor stack that is known to be insufficient; and allow it to autonomous roam around town, crashing into things and killing their drivers? Apparently that is fine in the eyes of the law. Is it unethical? Society seems to be deliberating this one. I hear a lot of "the ends justify the means" style thinking in this area.
Either way, this should embarrass the hell out of Tesla engineers. I used to feel sorry for them, but it's been long enough that they should know better.
Comments mention driving too close to the car in front, or too fast, but I'd like to bring attention to distracted drivers. You can see how late the 4th-8th car apply the brakes, and quite reluctantly too. The first couple of cars behind the Tesla did crash but at a pretty low speed, I'd say not even at airbag-deploying speed. But the 6-7th end up hitting the pileup at a good 40-50mph.
The paramount issue here is that the software isn't operating as intended. Reckless driving and distracted driving are compounding issues to that major concern. Rather than dismissing the accident as possible with a driver fully in control, distracted and reckless driving are great reasons why we shouldn't entrust auto-pilot software to public consumption yet.
Agreed that FSD is crap and shouldn't be available to the public. Maybe it's okay to be tested out in the public by professional drivers, paid to do that, and monitored for attention/alertness, much like Waymo did for oh-so-long (and Uber didn't before the infamous 2018 accident).
BUT... that Tesla could've been a student driver mistaking the brakes for the gas pedal, or a car running out of gas, or somebody passing out, or any number of other plausible circumstances where a car just stops in the left lane for no apparent reason.
Enough is enough. Stop using the public as testing grounds for software development and marketing defunct wares. If tech is meant to bring us the utopia we all need but don't know we want, the least that these companies who exploit public space can do is to get some ethics before someone does die. As if it isn't enough that a multi-car pile up won't ruin some people's days and waste so much time dealing with insurance, why are we expected to put up with this in the name of "progress?"
I also don't accept the argument that human are imperfect drivers, so it's OK if Tesla's shit software is placed at the hands of experienced drivers. People are not good at driving, and that's exactly why we don't need to enable our stupidity in a high-speed death box. Even if someone is a good driver, the tendency to relax when you are told that you don't need to be totally poised for action is a great way to delay response time, so I'd expect this to happen 99/100 with the best drivers in the world. In a situation with a relaxed driver, there's no humanly possible response time that's going to prevent this crash from happening.
Tesla ought not proffer software that is anything less than perfect and expect the public to fill in the holes. It's clear that drivers today are less patient, more distracted, and less competent all while embattled with environmental distractions. And as if it isn't hard enough to monitor 360 degrees, now I need to focus on which kind of car is in my area? In some places, it's not even possible to avoid this kind of collision. I can only imagine what would happen on I-40 in Raleigh or I-95. That's truly frightening.
I drive a Tesla and actually really enjoy the car, but I still get phantom braking issues, albeit the car does not fully brake and change lanes. Thank goodness I turned all of those features off. And that leads me to think that Tesla knows they have kinks to iron out. With all the data that they collect, they must have known accidents like this could happen and still put their product out. I want to see criminal charges, fines, and a Congressional investigation leading to legislation.
What's the price of a life, Mr. Musk? Would you put your kids in one of your cars, turn on auto-pilot, and expect them to make it from Point A to Point B each and every time?
0 mistakes seems like a pretty extremist and impossible standard. If that standard had been applied to automobiles before they could be driven on roads, we'd still have horse and buggy everywhere and we'd be dealing with catastrophic amounts of horse shit on all the city streets
> 0 mistakes seems like a pretty extremist and impossible standard.
Perhaps, have an upper limit for the speed at which these "beta" features are illegal. I don't know why Tesla doesn't support "self driving" feature only at lower speed, where the risk of fatal collisions is lower.
Fine, but if we are to lower the threshold from perfect to some mistakes, then this category of error needs to he empty. Phantom braking is in a less severe category, as it is manageable. In this case, I think the driver or following car could’ve escaped an accident.
> Zero mistakes is something a machine can be built to fulfill.
Even the lunar lander and other NASA software that received mountains of testing/review has had errors. I don't think it's possible (for a human at least) to design perfect error-free software. And that's before considering hardware faults/failures that do happen in the real world.
> Re the hyperbole of horse and buggy: What if your pacemaker decides to go to 200 beats per minute late at night.
I would hate that of course, but I'd rather have the option to get a pacemaker knowing that in rare cases it might kill me. Without the pacemaker option a lot of people would be dead anyway.
> The cult of ship fast and fix it based on feedback is taking us further from ethical behavior.
Too late for that. Tesla autopilot is responsible for decapitating not 1 but 2 people in separate accidents. Tesla had 3 years to fix the problem after the first decapitation but decided to double down on their flawed (and now fatal) autonomy strategy. Imagine if bridges were built this way.
NATFB (not a tesla fan boy) but isn’t there an inherent bias towards reporting and exploiting FSD’s faulty behaviors vs reporting the times where it saved butts?
Elevators undoubtedly save lives over time, but it's a lot harder to count how many heart attacks didn't happen vs. how many people were suddenly crushed to death. But if an elevator ran perfectly for years at a time but also could, at any moment, instantly and inexplicably kill its occupants — it's still every important to understand why there's a flaw in the automation.
In any case, how would you measure when FSD/AP "saved butts"? Every driver is expected to operate the systems with the same ability and awareness to manually intervene (i.e. save themselves) as if the automation wasn't in place.
> In any case, how would you measure when FSD/AP "saved butts"?
I would not know how to measure it, but Karpathy provided a clear example during a talk at CVPR 2021 where the FSD (rightfully) slammed the break as a driver was driving his car literally into a lake.
3. Come to a complete halt in a tunnel with busy high speed traffic for no apparent reason.
And the excuse that the driver should have been ready to take over within fractions of a second
of the Tesla making a mistake while in FSD is absurd expectations on human reflexes and attention.
It's non-deterministic, sure but it also is software so one should expect that Teslas with the same software and same hardware in the same situation will cause the same problem.
Do your earbuds Bluetooth and wifi experience signal loss with the connection breaking? This occurs at sensitive locations of national security interests to prevent oh, let's say remote bomb detonation. At key infrastructure it ain't gonna happen.
This means all data and responses are tied to Tesla Inc through nodes tracking tfo of everyone and is unsolvable for military installations, critical infrastructure are embedded throughout America for Liberty. Elon Musk is not and never will be American in the pursuit of Liberty.
Autonomous is a lie.
Autonomous means the vehicle itself needs all within it.
This is RC autobot controlled vehicles with humans inside.
I always have thought Musk a person of low moral/flawed character. But to do this type of wicked makes me despise him. He has never read a science fiction book in his life. If you love hacker news you Capiche
.
This an excellent article btw but most comments compare humans lacking situational awareness. This is a computer in a vehicle that requires ALL CAPABILITIES ON BOARD ACTIVE AT ALL TIMES. Ok, maybe as you drive along like 250 min then get a localized updat for all points 250 mile like a triptick from AAA, is a compromise. This is coding hell Heinlein warned us about.
204 comments
[ 3.7 ms ] story [ 272 ms ] threadThere would have been multiple ways for the cars further back to prevent a pileup: Drive slower, keep more distance. If you speed, have both hands on the wheel, look at the road, and be ready to brake quickly.
It does sound like there were no severe injuries, so you could argue that people kept enough distance to avoid serious injuries.
if they do, so what? even if 10 people cut in front of you, with a 30 minute drive, how much time are you really losing? 50 seconds?
I'd also argue that the Tesla pulling over into the left lane (merge 1 lane left vs 3 lanes to the right) is the right behavior in a situation that warrants stopping. Try to get to the side of the road so other cars can pass.
Coming to a full stop in the left lane is really last-resort stuff. That is the speed/passing lane (in most jurisdictions, all lands are speed/passing around the Bay Area and especially on the bridges).
In general I am ok with people merging in the gap, but some people are assholes about it by merging too close to me that I have to brake. I don't understand why people merge when they know the person behind in the other lane will have to slow down to let them merge.
Added bonus is that if you witness an incident, the police is very happy when you can send them the video file to prove how it all happened.
Well "unsafe lane change" can be totally legit as well. I for once had to do no think abrupt lane change when I saw some largish object appear seemingly out of nowhere and it was flying towards my front window. Still do not know what it was. Abrupt lane change did not cause any problem in this case but in theory it could have.
Of no fault of their own, the car you hit trying to avoid something means you didn't avoid them. The car you hit did nothing wrong, but you did. Insurance will place you at fault at a minimum of partial fault. You will not be blameless in your example if you do hit someone/something
I am not expecting anything. Just trying to save my life.
I have better solution. Why don't we stop manufacturing cars. No more car crash victims.
Growing up in the Midwest, I remember being taught (not sure if this was official or something I got from my father, older brother, or maybe an uncle) to quickly assess if I believed that I would hit the deer or not.
If I was going to hit the deer, regardless of my actions, I was to accelerate into it.
If I could brake and avoid it, I was to brake.
I've been in that situation a few times. In the case of a collision with a deer, I accelerated into it and only took damage to the front passenger side panel. I do believe that it helped prevent a bad situation from becoming worse.
In the other cases, I was able to avoid a collision by braking.
Later in life, I did swerve once to avoid a large animal that I mistook for a person walking on a very dark, very rural road. My car flipped and slid on it's top ~100 ft. before coming to a stop. Myself and all 3 passengers walked away without incident but I still feel fortunate that we all did these many years later. Swerving is almost always the wrong answer.
Reason given was that braking causes the front end to drop causing the deer to be “pushed” up onto the hood and into your windshield.
No sure if this is true or not.
I don't think that's right. Because safe following distance can only prevent you from slamming into the car in front of you if it breaks hard. If it hits something and stops way before the breaking distance, there's little you can do.
There was something you could have done: leave more space.
That is your opinion, and my local DMV does not specify a distance that long.
It is physically impossible for a car to come to an instantaneous stop for no reason. It would have to e.g. drive into a concrete wall, but if there's a wall across the freeway you should've seen that from afar.
You only really need to have enough distance to be able to notice the car ahead braking and start braking yourself.
Personally speaking, in this scenario, with an enormous stream of one way traffic on one of the few roads leading from San Francisco, I hope circumstances are taken into account.
All Teslas could do this at any moment, is my takeaway.
You’re to maintain distance and speed that would allow you to stop in time without an accident at all times.
Same goes for potholes, if you hit a pothole and bend a rim or something, you can get a fine for driving without maintaining proper speed. Then you can use that fine to sue your local government for having a pothole there.
I don't know where it would be allowed to stop in the middle of a highway lane for no apparent reason, especially after a lane change that did not result in a proper distance to the car behind.
The next thing I remember is hearing an almighty crash, followed by the Commodore crashing into us, followed by another crash as we were pushed into the car in front. A moment later a motor cycle flew by the left hand side. It has evidently been unable to stop and and swerved right to a full frontal but was then off balance. The handle bars then creased the left hand panel of every car. The bike was only slightly damaged, and he drove off.
The police arrived. After a few questions, they determined it was the last car in the chain that wasn't paying enough attention and had pushed everyone else the car in front. They were taken to the station. The rest of us heard nothing more of it. The insurance companies paid out without much of a fight after getting the police report.
But that's about the limit of it as far as I can tell. In Australia, if you going though a intersection with a green light and hit someone who has run the red, your up for dangerous driving. It's not "one person made a mistake". As far as the law is concerned you both weren't paying attention.
If the first crash were the only one, this would never have reached the level of media attention. If the same pileup was caused by a garden variety grandparent, no one would care.
[1] I'd suggest caution in attribution here. The only evidence that FSD was involved is the at-fault driver's statement to the police. I'm not going to say it wasn't operating, but I will say that I've never seen behavior like this from FSD. It's extremely cautious at lane changes, often annoyingly so. And while I've seen the same "phantom braking" everyone has, it's a pulse, not this kind of "bring the vehicle to a stop" behavior. Given the assertiveness of the controls, I'd say it's just as likely that the driver decided to swing into the left lane on their own and then hit the brakes instead of the accelerator by accident (which happens, again, all the time!).
The distance you have between you and the car in front on you is one of the only safety zones you can control on the road. People should cherish it and let it blossom into wide open space.
You often have no control over the left, right and rear. Why would you want to take away that front space as well?
And don't get me started on elephant racing. I frequently wonder the economic and environmental impact of just this singular type of event in central California.
https://www.alamy.com/stock-photo-keep-apart-2-chevrons-moto...
(Obviously we can't tell the speed of the cars in the photo, but if it's the standard speed many drivers are still ignoring it.)
https://www.google.com/maps/@53.3015568,-2.722796,3a,75y,235...
StreetView gives a much better impression of the chevrons; the Alamy photo has an unusual choice of lens. The "M56 B 45.0" [km] milepost is way ahead of the chevron sign.
Oh the money that could earn.
Which is, perhaps, one of many arguments against having so many people on the roads to begin with, but that's what we have.
Yeah you can.
The maximum carrying capacity of a single lane of highway traffic is around 1900 cars per hour (that is from real-world measurements of what human drivers can actually achieve). At 70 mph that is 194 feet per vehicle. Even assuming something like a 20 foot long F250 or something that leaves 174 feet of spacing between vehicles. Which is nearly 9 vehicle lengths and nearly 2 seconds of room. Up until you hit that point you're not impeding any traffic flow by adding more space. Almost nobody in this video is leaving that much room.
What causes traffic congestion and causes that 1900 vehicles per hour number to be that small is the people who are tailgating and when someone brakes everyone behind them has insufficient following distance and has to brake harder and harder until you've got vehicles nearly stopped on the freeway and have formed a traffic compression wave (if you haven't just created an accident and snarled traffic even worse).
Source: I spent several months last year helping someone learn to drive in the Bay Area, and they insisted on leaving the recommended gap between cars as they prepared for their driving test.
And if you don't let yourself feel "aggravation and churn" then people will just merge into and later leave your lane and it just becomes normal driving without any ego involvements.
Source: the past 30 years of driving in Seattle traffic that I've done where I've stopped being a tailgating idiot and opened up more space and stopped caring about people "cutting me off".
[*] not all at the same time...
The rule actually works great where I live (a rural area) but if I wander over towards chicagoland I will be punished with constant cutoffs. Also there's too many cloverleafs over there jeeze.
There's too much traffic. It backs up into the streets, before this bridge + tunnel regularly. The speed is already restricted.
Would you make excuses for a human that did that while conscious and not in any kind of distress?
I'd be less understanding, yes.
But at this point it's presumable that the autopilot was in "distress" (avoiding a perceived issue). And when such a thing comes up, real or false, the drivers behind need to be ready to react. And several of the cars did so quite well actually. The only reason this is a dramatic incident is because of the like 5th or 6th car that hits everyone else hard.
On the other side I do not understand why this 'Autopilot' is not banned from the public roads.
Would the headline be "Michelin causes 8 car crash just after rubber contract with Brazil" ?
The title is completely unfair to Tesla.
It's fair to assume "full self driving" will continue driving, rather than change lanes into an occupied lane, and stop abruptly.
It should have slowed gradually.
It's Tesla's fault.
I’ve been in the situation when Tesla auto breaking saved the car from crush, which was similar situation to this.
He shipped a car which is Not designed for Today.
Hitting the brakes in the tunnel for no reason is a "never to this" kind of move.
"But [...] other scenarios" is not the situation here. There is no escaped pet anaconda slithering across the tunnel floor. There's no Godzilla breathing fire inside the tunnel. It's just a car with no obstructions in front of it, none visible to any other drivers, and it's choosing to hit the brakes.
There is no excuse.
It's at the bottom of a downhill slope. There's a limited view of what's up ahead because the slope starts to incline inside that tunnel. Cars start to speed up. There is typically a massive amount of traffic on the bridge.
Blame Tesla, full throatedly.
While it is true that there is a culture of following to closely, I don't think it applies here. When a car merges into your lane you should not abruptly change your speed to adjust the distance, but slowly increase it. Otherwise you get huge ripple effects and people 1km back will essentially be in stop and go mode.
Everyone beside perhaps the first car that had an incursion into its lane (but arguably failed to reduce speed to avoid an accident with a vehicle braking and signaling their lane) are WAY too close for the speed. Any speed reductions in those lanes would necessitate heavy braking which would cause exactly the ripple effects you’re referring to under normal conditions because there’s no buffer whatsoever to gradually reduce speed/increase distance either.
Drives way too close to other cars (tailgating).
Drives way too fast all time, even in neighborhoods.
Drives way too aggressive.
Drives way too entitled. I’ve seen some crazy stuff from people driving a Tesla. Driving on the shoulder when traffic backs up. Zipping around people to steal parking spots.
In Germany we are taught to keep a lot more gap because the time it takes to break or react will eat up a lot of distance.
The last time I pointed this out I was told: well, I can still steer to the side if that happens.
This is 100% the drivers fault though, it is extremely easy to override Autopilot. Tap the brakes will disable the entire system, press the accelerator will keep lane keeping on but override the speed, and moving the steering wheel will disable lane keeping but keep cruise control on.
Now, I've also heard people say that the phantom braking issue can be really violent and so it's very possible I've just never experienced true phantom braking, but based on how many of the drivers behind the initial wreck slammed into each other it's clear many folks do not pay enough attention or leave enough following distance to be safe.
In this case it looks like the car not only breaks ... but it turned into an occupied lane. Then it stopped. It didn't stop gradually either.
So I don't see how this is correct behavior on part of autopilot.
Most lane keeping/cruise control setups today from other OEMs would also be sufficient in most of the scenarios I enjoy AP though.
What we should be more surprised about is the total failure of our government agencies to be with the times and create proper regulation and specifications of operation and testing.
Autopilots on aircraft don't suddenly nosedive the airplane and try to land on whatever arbitrary terrain is underneath the aircraft.
This is Tesla's fault for designing a product that is not appropriate for current roadways.
That's speculation. Do we know if the car had suddenly lost the ability to continue driving? Was the car pulling to the closest side of the road? If you lose power in a gas powered vehicle, have a flat, or some other failure when you're in the far left lane, what should be done?
From the video, it looks like a lot of people are following way too closely. Far less than the 3-second rule would allow for.
The first vehicle behind that crashed got caught by the sudden lane change.
The second vehicle avoided rear ending them. You can see the weight and suspension settle just before it gets hit from behind.
Three was slowing and trying to avoid the crash (it shifts lanes and makes it past), while the fourth was in a rush to get past three (you can barely see it, that's how close they were following).
While the quick pull to the left side of the road probably initiated things and caused the first vehicle to rear end the Tesla, the pile up is the fault of the fourth car (which can barely be seen behind the pickup that managed to avoid it), and all the others that were clearly following way too closely.
Safely merge right and stop in the shoulder if possible, but in heavy traffic like this put the hazard lights on and slowly stop without changing lanes. Call the police -- do not get out of the car while there are still lanes with moving traffic.
In wealthier-Europe the police would activate electronic "60km/h", "lane closed" and similar signs. It's a long time since I drove in the USA, I can't remember if you have these.
> have a flat
If you must stop quicker pulse the brakes if possible (the flashing lights draw more attention), but again, don't change lanes. Changing lanes caused the problem, which was compounded by the too-close following vehicles. Some of those vehicles also changed lanes dangerously, when they should have stopped in their current lane.
For what it's worth, a vehicle two ahead of me crashed into the central barrier in my final driving lesson.
We don't know if it is 100% the driver's fault.
But we do know with reasonable certainty that AP did not "lock out" a driver or anything. Autopilot is universally override-able.
The feature is named autopilot or full self-driving, it's misnamed. All the marketing and hype puts prospective drivers into the nod and wink mindset that they don't have to drive when they have a self-driving vehicle.
It gives the driver the wrong impression.
Only after experience would you have accumulated the wisdom to treat it a lot more conservatively.
In my opinion if I were to own one of these things I would be Embarrassed and offended at the notion that I spent tens of thousands of dollars on a piece of software in a special vehicle only to have it not do what I expected it to do.
It's an anti-pattern.
This is due to both - buggy software being tested on public roads and people going too fast and driving close behind. The first two collisions couldn't have been avoided because of how abrupt Tesla changed the lane and stopped, thats on Tesla.
But that said, it’s fine. People who feel that way should not buy a Tesla or, if they do, don’t use AP. Free market and all that.
I think it’s made extremely clear that you need to be paying attention during FSD/AP. Car may have partial blame for doing something unexpected but driver didn’t do anything to get out of that situation.
I imagine shock and panic would be a reason.
This is an unexpected type of unexpected; how do you develop the proper reaction?
This is pretty disingenuous given that I believe Teslas have probably covered by far the most miles with an "advanced driver assistance system".
The Tesla definitely screwed up here, and I'm not sure you could have prevented the initial crash, but the pile up behind it is entirely the other drivers' fault.
In both cases, humans failed to react fast enough to something that required active focus and fast, split-second reaction.
On top of that most other manufacturers don't have telemetry about it either and instead rely on dealerships to collect that information.
> The Tesla definitely screwed up here, and I'm not sure you could have prevented the initial crash, but the pile up behind it is entirely the other drivers' fault.
I'm still confused about the whole situation.
Police report mentions FSD v11, which is only for employees so far.
FSD v11 is what makes the freeway "FSD", otherwise it's their old "we go straight in lane, maybe take exit or change lane" code...which wouldn't be FSD.
So either an employee crashed with internal only versions, or the police and owner are just incompetent about what they actually have.
It might be true that Tesla does have the most miles, but I wouldn't say its a sure bet. I'd love to see some actual data on it to know instead of making the assumption.
I actually figured it would have been a bit bigger of a lead for Tesla, but showing that figure about Accords really made me doubt that Tesla leads in all ADAS SAE Level 2 miles traveled.
Also, FWIW "advanced driver assistance system" (ADAS) is definitely a term of art which applies to a ton of technologies. Those technologies include radar-based adaptive cruise control, collision avoidance systems, advanced traction control systems, traffic sign recognition, and other technologies. The first Taurus with adaptive cruise control would be considered an SAE Level 1 autonomous car with its advanced driver assistance system. So even a car with just emergency braking technologies but otherwise an SAE Level 0 would be considered as a car with ADAS on board, even though it might only has the ultra basic cruise control.
I'm not sure that this is true given almost every Honda sold since 2019-ish (and a big proportion for years prior to that) has had ADAS features included and enabled by default, most notably Automated Emergency Braking.
Anecdotally, my wife and I used FSD out in Oregon while on vacation a year ago and had these these "phantom brakes" occur two times on the same stretch of road. It was very sunny day through the gorge to Hood River from Portland and I'm thinking there definitely would have been sharp contrasting shadows from the trees and sunlight.
I don't expect FSD to be able to find its way to a proper rest stop or off-ramp, but seems like a less dramatic shutdown mode should be possible, e.g. driving a couple of miles at moderate speed until there's a safer pull-off opportunity.
I'm fairly certain this isn't a legal option in the United States. That would be unattended self driving.
It's probably a better solution, it just isn't available.
Cars meant for the rest of the world have separate yellow lights in back and you could flash those. I think they are adding a rapid burst flash for alert with those.
> resulting in an eight-vehicle crash
I'm sorry, but this is an 8 vehicle pile up from people following too close (or paying too little attention). At least where I'm from if you run into the back of a car it's your fault. It means you followed too close, failed to react to a situation in front of you.
Should the car have stopped there, no?
But is it that car's fault that others ran into the back of it? No.
> short of mind-reading.
This is called defensive driving and it involves running "what if scenarios" and being prepared for even those. "What if the tesla has a tire blow out and comes in my lane" Is something that can be thought of any brand of car.
The Tesla is 100% at fault here.
It is a neat fact to bring up but that doesn't apply here. The Tesla is still at fault and you would be hard pressed to find any adjuster or courts agreeing with you thanks to this excellent video evidence.
Luckily it's not up the adjuster, it's up the courts.
This was precisely the type of driving I figured would be easy to automate - rather than deal with urban settings, pedestrians, bicycles, etc.
Shame they don’t have LiDAR or some other form of actual distance/range sensing.
Unfortunately, Tesla seems to have something to prove when it comes to LIDAR, and stubbornly refuses to adopt the technology, despite crashes and even deaths that have occurred, and general industry understanding that it's the enabling technology for driverless cars.
Is it legal to create a two-ton machine with admittedly beta-quality software, and a sensor stack that is known to be insufficient; and allow it to autonomous roam around town, crashing into things and killing their drivers? Apparently that is fine in the eyes of the law. Is it unethical? Society seems to be deliberating this one. I hear a lot of "the ends justify the means" style thinking in this area.
Either way, this should embarrass the hell out of Tesla engineers. I used to feel sorry for them, but it's been long enough that they should know better.
BUT... that Tesla could've been a student driver mistaking the brakes for the gas pedal, or a car running out of gas, or somebody passing out, or any number of other plausible circumstances where a car just stops in the left lane for no apparent reason.
I also don't accept the argument that human are imperfect drivers, so it's OK if Tesla's shit software is placed at the hands of experienced drivers. People are not good at driving, and that's exactly why we don't need to enable our stupidity in a high-speed death box. Even if someone is a good driver, the tendency to relax when you are told that you don't need to be totally poised for action is a great way to delay response time, so I'd expect this to happen 99/100 with the best drivers in the world. In a situation with a relaxed driver, there's no humanly possible response time that's going to prevent this crash from happening.
Tesla ought not proffer software that is anything less than perfect and expect the public to fill in the holes. It's clear that drivers today are less patient, more distracted, and less competent all while embattled with environmental distractions. And as if it isn't hard enough to monitor 360 degrees, now I need to focus on which kind of car is in my area? In some places, it's not even possible to avoid this kind of collision. I can only imagine what would happen on I-40 in Raleigh or I-95. That's truly frightening.
I drive a Tesla and actually really enjoy the car, but I still get phantom braking issues, albeit the car does not fully brake and change lanes. Thank goodness I turned all of those features off. And that leads me to think that Tesla knows they have kinks to iron out. With all the data that they collect, they must have known accidents like this could happen and still put their product out. I want to see criminal charges, fines, and a Congressional investigation leading to legislation.
What's the price of a life, Mr. Musk? Would you put your kids in one of your cars, turn on auto-pilot, and expect them to make it from Point A to Point B each and every time?
Perhaps, have an upper limit for the speed at which these "beta" features are illegal. I don't know why Tesla doesn't support "self driving" feature only at lower speed, where the risk of fatal collisions is lower.
Re the hyperbole of horse and buggy: What if your pacemaker decides to go to 200 beats per minute late at night.
The cult of ship fast and fix it based on feedback is taking us further from ethical behavior.
Even the lunar lander and other NASA software that received mountains of testing/review has had errors. I don't think it's possible (for a human at least) to design perfect error-free software. And that's before considering hardware faults/failures that do happen in the real world.
> Re the hyperbole of horse and buggy: What if your pacemaker decides to go to 200 beats per minute late at night.
I would hate that of course, but I'd rather have the option to get a pacemaker knowing that in rare cases it might kill me. Without the pacemaker option a lot of people would be dead anyway.
> The cult of ship fast and fix it based on feedback is taking us further from ethical behavior.
I couldn't agree with you more on that one.
I feel they shipped a product suitable for a Different reality of roadways. Different requirements. Not what we have today.
Too late for that. Tesla autopilot is responsible for decapitating not 1 but 2 people in separate accidents. Tesla had 3 years to fix the problem after the first decapitation but decided to double down on their flawed (and now fatal) autonomy strategy. Imagine if bridges were built this way.
https://cdllife.com/2019/feds-say-autopilot-was-engaged-in-f...
In any case, how would you measure when FSD/AP "saved butts"? Every driver is expected to operate the systems with the same ability and awareness to manually intervene (i.e. save themselves) as if the automation wasn't in place.
I would not know how to measure it, but Karpathy provided a clear example during a talk at CVPR 2021 where the FSD (rightfully) slammed the break as a driver was driving his car literally into a lake.
1. drive straight into a concrete barrier https://youtu.be/Tr-oF7J0cBw?t=13
2. drive straight under a tractor trailer https://www.youtube.com/watch?v=9BgV-YnHZeE
In this instance
3. Come to a complete halt in a tunnel with busy high speed traffic for no apparent reason.
And the excuse that the driver should have been ready to take over within fractions of a second of the Tesla making a mistake while in FSD is absurd expectations on human reflexes and attention.
273 FSD crashes in 1 year. Over how many miles driven? And how does that rate compare with non FSD vehicles?
Unless ... magic.