> According to the NTSB, a witness had come across the stopped CR-V prior to the collision and noted that neither its taillights or hazard lights were illuminated, She was able to change lanes and avoid hitting the Honda, but after she passed it, she saw the Mach-E strike the stopped crossover in her rearview mirror.
This scenario sounds like what I refer to as a "peek-a-boo rear ending". You see it exaggerated in movies with high speed chase scenes, but its real world equivalent is a very common road going scenario and almost certainly leads to a collision if at 45+ mph and the rear car driver isn't maintaining situational awareness of traffic flow several cars ahead...a scenario that aggressive tailgaters simply don't take into account in their public behavioral risk calculus, and clearly BlueCruise isn't capable of handling either.
I had this happen with Teslas FSD about a year ago. A car missed a turn at a light because they didn't get into the turn lane early enough ... so they decided to make the turn from the "going straight" instead.
But the light changed before they made the turn and they became stuck in the middle of the intersection once they lost the protected left ...
So now they're sitting in the "going straight" lane where you wouldn't expect a stopped car. The car in front of me changes lanes without really slowing down, and the FSD prediction model didn't pick up that the stopped car was stopped. I can't really blame it, it took me half a second to figure it out as well and by that time I was hard on the brakes.
Isn’t this a sign that you are following too closely? A driver should typically be able stop a car length per 10mph. So if you are traveling at 40mph you would need about 4 car lengths to react and stop. Ergo if you are following closer than that you may not stop in time. Just because you are using cruise control doesn’t absolve you of this requirement. My car allows me to set a cruise control gap in seconds which translates to reaction time. 2 seconds is the lowest gap mode and usually leaves me with half the distance necessary for a safe stop (although the car has aggressive brakes and can quickly slowdown at the expense of passenger comfort). Due to this I leave the gap mode at 6 seconds which is larger than I would like but would (and has) allow me to slow down in time for a stopped object in the road.
I apologize for any flame wars this creates, since people find driving to be deeply personal and any findings which may criticize their driving habits is generally frowned upon.
* ... should typically be able stop a car length per 10mph.*
This is not correct, stopping distance scales as the square of speed. The distance you'd need to stay from the vehicle in front to be able to stop if it magically came to an instantaneous stop is far, far larger than anyone will reasonably keep.
The stopping distance at 65mph, including reaction time, is at least 300ft. That's ~20 car lengths, not 6.5 that your formula would yield. If conditions are wet, your tires are not in great shape, or you're not paying attention, it's going to be much longer.
The scenario under discussion is when the car in front doesn’t stop but changes lanes, revealing a stopped car in your lane right where a moving car had been. It’s like the car in front of you was suddenly stopped by magic.
The solution is typically to swerve with the other car, but sometimes that’s not possible.
But it is possible, when the cosequence is someone dying, the likelihood do not need to be high to act carefully. Anyway the example here was about a car in front avoiding a stopped car.
This thread is about how closely to follow a moving car. The worry is if the moving car suddenly switches to being stopped. Will you be able to also switch to being stopped before hitting it?
Yea, some people seem to miss that just because the car in front of you can jump over in another lane doesn't mean you can at all. If you have a car beside you, and the car you're tailgating does not, then you're screwed in an emergency situation like this.
If a car has its brakes on, it'll be extremely hard to accelerate
It should be impossible to accelerate. None but maybe the most high-output engines should have enough power to overcome the car’s brakes. Aerodynamic drag is not quite a rounding error, but I don’t imagine it counts for a whole lot in this scenario in comparison.
Following too closely is just how people drive, unfortunately.
The main reason I almost always have adaptive cruise control on is so that it keeps me at a safe braking distance from cars in front of me but I've found that most people do not care.
Just yesterday there was a corolla that was following a school bus so closely i'm sure it could sniff the bus' ass. That driver was one hard brake away from being decapitated by the school bus bumper.
The problem I have with adaptive cruise: if you set your gap too wide, usually the suggested gap, people keep swerving into it, so it makes for a very nauseating drive if you are somewhere with not enough traffic that you are moving slowly, but too much traffic that people don't have to change lanes.
Just yesterday I was driving in the left lane with cruise control set to the maximum allowed velocity, with the maximum gap behind a car driving in front of me. Of course someone overtook me on the right (illegal in Germany) just to swerve in front of me and then drive at the same speed, just 20m further ahead. He ended up being first at the red lights. I’ll never understand it…
People don’t intuitively notice speed, they see following distance as a proxy for speed: people with longer following distances are perceived as “driving slow” and they feel they are going faster if they tailgate. Both untrue, but almost universally assumed without a second thought
If you are traveling 80 in the left lane while the right lane is locked in at 60 (trucks who are limited to power speed limits), it is simply impossible to travel in the right lane. You are constantly overtaking vehicles, so it works out legally, but, yes, if someone wants to go 90, there is a conflict. Ideally, you wouldn’t cruise in the left lane because there would be room to cruise in the right lane, but higher traffic situations don’t really allow for it (well, not so high that you can’t go 80 in the left lane).
This usually happens when you are approaching a city and are going to get a new lane or two pretty soon anyways.
Well, it sort of does. Imagine your doing 100 in the left lane, as does the car in front of you and the three cars in front of them. There’s a truck doing 80 in the middle lane some 500m away. All of you are going to pass it. Now a sixth car starts tailgating behind the last in line and because there’s left a big safety gap he’s overtaking on the right.
I was the last in line, going at the exact same speed as the cars in front of me, just leaving enough space. Why would I switch lanes?
I've found that the adaptive cruise has made me an almost "egoless driver". It does keep a conservative distance, and aggressive drivers do fill in the gap, but I don't care as much when I'm not the one on the pedal.
With the full map and destination time in front me, I don't feel the urge to game the system (or maybe I'm just getting old).
From some motorcycling book I remember the right response to that is to slow down even more and maintain following distance with the new car in front of you. If we let ego go and can be home 2-5 minutes late on average all will be well.
Ya you can do that, but then it feels like dangerous. When it gets like that I usually just shut it off and control distance myself (and "cough" adjust it to dissuade swerving and then back off for safety).
ah. my car doesn't have adaptive cruise control but I've rented cars that do. it took a bit to get used to medium and feel comfortable and not dangerous, but then people aren't cutting in.
I have to imagine the level of discomfort depends on the car. I've never ridden in a vehicle with adaptive cruise control, but if it responds quickly to someone jumping in line, that could go beyond annoying to motion sickness if it keeps happening.
I've used a few adaptive cruise controls that will not maintain a great enough following distance to come to a complete stop in the following distance (eg, in the case that the car in front of you swerves out of the lane revealing an obstacle).
I live in a cramped east coast city. If everyone around here followed the DMV guidelines for distance between cars, morning rush hour would probably be from 4am to 11am. Sure, the fact that everyone's driving in such a tiny little space is the problem, but the infrastructure here doesn't leave us much choice.
Even if the length of the cars end-to-end is often greater than the length of craggy narrow street between tightly packed but unevenly placed intersections with anywhere from 3 to 8 entry points? Even when all of the 5 highways people are trying to enter and exit– all within the same mile– are all completely backed up? I'd have to see that to believe it.
So the point you were trying to make is that single land roads don't face traffic issues from merges and lane changes? That seems to trivially be false in my experience. Having driving all over the USA on roads with a single lane in each direction, merges and lane changes are by far the biggest cause of traffic slow downs.
The situation is similar in the Chicago area. Some roads are almost perpetually bogged down. If there is an accident or stalled car, it gets worse. I see two things that affect flow.
1. As traffic builds and distance between cars decreases, the flow becomes susceptible to "shock waves". (Just search Youtube for "traffic shockwave" for an example. I think it happens when someone taps their brakes and the next car behind them has to brake a little harder ans so on down the line until traffic comes to a complete stop.
2. As a commuter and occasional weekend driver I recognized two traffic patterns. During normal commutes, the same drivers drive the same routes every day and most settle into a pattern of maintaining speed, remaining in lane, moving to merge and so on in harmony with other drivers. There is the occasional outlier, driving slower than everyone else or weaving in and out and trying to get ahead by weaving and passing in whatever lane momentarily has space. On weekends I see drivers who are unaccustomed to the driving situation and simply don't mesh well with other traffic. Even though there is generally less traffic, it does not flow as well. There is more lane changing and stuff like crossing three lanes to get to an exit because they forgot where they needed to exit.
Excess traffic is a problem and unfortunately we seem to be able to overfill the available capacity any time it is increased.
When I worked in downtown Chicago I was very happy to be able to take commuter rail to work rather than to have to drive, but those options seem to be more the exception than a regular option. That's a different discussion.
Yeah I've had my fair share of big city traffic, but this is very different. In many ways driving here is a lot easier than driving in a place like downtown Chicago or Manhattan because it's much much much lower-volume, so there's a lot less action you have to pay attention to, but all of the roads are tiny and twisty. It's like heart circulation vs pinkie tip circulation. And since it's so small, even being slowed down to a crawl from one end of the city to the other as part of a longer drive would affect your drive less than even moderately bad traffic in a metropolis.
The traffic is not frequently caused by accidents. It's a small city so it's pretty clear when that happens. We have some of the lowest accident stats in the country. It's probably because any time there are a significant number of cars on the road, the entire region slows to a crawl.
I'm pretty much out of the habit of using cruise control even with newer adaptive cruise control because it really encourages maintaining a constant speed even on roads that are really too busy to allow it.
In this city, even at 3:30 in the morning on Sunday, cruise control is just useless. It's a spaghetti pile of twisty roads and a billion intersections. Even on the major thoroughfares, you're not going to drive more than 20 seconds without hitting a intersection. To make matters worse, it's all way too disorganized to efficiently coordinate-- almost nothing is perpendicular-- so when traffic is light they just have very short light cycles and you're nearly guaranteed to hit a red light in a few blocks, max. Most residential streets are only wide enough to accommodate one car, but in many cases there's no logistical way to make them one way, so you just have to get good at constantly negotiating who should yield and who should go with oncoming traffic. That, specifically, causes shockingly few problems! But it makes for complex traffic patterns that require a lot of cognitive load to handle. If there was good public transit here, it would be great.
I often try to follow the car-length-per-10mph rule. It's often times not possible. Cars with just duck in front of you and people behind you will pass you. One of many reasons why I hate daily driving.
> I apologize for any flame wars this creates, since people find driving to be deeply personal and any findings which may criticize their driving habits is generally frowned upon.
I appreciate your attitude here. Mine is opposite. Dangerous driving injures, maims, and murders both perpetrators and innocents alike. Criticise bad driving loudly and often. You might save lives. I had to tell my mother in law that her daughter, my wife, lost her life in a road traffic accident. There are few aspects to serious road traffic accidents that are not horrific, and speaking up may save lives. Every year, 1.19 million people die as a result of a road traffic accident. Between 20 to 50 million people suffer non-lethal injuries, many of those leading to disability.[1]
In the U.K. some years ago the police changed the term in common use from "Road Traffic Accident" (RTA) to "Road Traffic Collision" (RTC) in order to underscore that these were not accidental but mostly due to poor driving of one sort or another.
"Road Traffic Collision" is encoded in the applicable laws.
Deepest condolences on your loss, and thank you for your candor—I can only speak for myself, but I’m definitely turning up my driving conservatism after reading your post.
I’m very sorry for your loss. I agree with your stance on being vocal about poor driving.
People speeding through the 30 mph limit in my village boils my blood like nothing else and I’m one of those parents who shouts at traffic on the school run, even driven by people I know. This has likely cost me a few invitations to the pub but I don’t care.
Can you lobby your gov't rep to install automated camera? It seems reasonable. If you are brushed aside, email/write/call a local newspaper or TV station. They would probably like to report it.
1 car in front of your vehicle, between you and the stopped car.
Assuming the car in front of you switches lanes around the stopped car with 1 car length of space between it and the stopped car.
So 2 car lengths + your personal gap for you to slow down.
So you could be going no faster than 20mph worst case (0 personal gap).
The issue is that the car in front of you doesn't need to stop, so can potentially "suck you in" at an unsafe speed, before you realize you'll need to come to a complete stop (assuming your lane change is blocked).
It's a fair point, and at least I think of myself as a fairly conservative driver.
I think it is the only time though that I've been faced with a stationary car suddenly appearing where I wouldn't expect one, and that's very different to the car in front needing to slam on the brakes with whom you can share some stopping distance.
fwiw 6 seconds at 45mph is 120 meters. That we can comfortably go that far in 6 seconds is a modern marvel, but it's also 20 car lengths.
I understand 6 seconds is a lot of space. My cruise control has a delay before reacting so 6 seconds allows for the car to react and if it’s not to my liking then leaves me with the necessary space to stop. On the highway I usually control my distance by switching between 4 and 6 seconds.
What car is that? Of all the cars I've driven I've never encountered one that allows for more than ~2 seconds on the maximum setting. Thus leaving me a bit nervous.
I'm almost hesitant to believe that any car on the market has good enough sensors to do 6 seconds. That is a very long distance at highway speeds.
Quote: "For example, if you drive at 56 mph (90 km/h), the distance is maintained as follows: Distance 4 - approximately 172 feet"
172 feet at 56 mph is 2.1 seconds.
Not saying that you are wrong (maybe a recent model has better cruise control), but I think most people vastly overestimate the adaptive cruise controls in their cars and ~4 seconds would, for reference, vastly outperform 2023 Tesla Y in that aspect.
Huh, yeah I have a 2021 and your math seems correct.
I always drive setting four and it keeps a very comfortable amount of distance. I think something is incorrect here because I definitely get at _least_ three seconds, or maybe I'm counting too fast.
That rule explicitly states that it's not for being able to stop, though, it's just to have time to react to the driver ahead stopping. Two seconds isn't nearly enough to stop in front of a stationary object. At highway speeds it's even barely enough to not rear-end the vehicle ahead if it has to panic stop.
>Isn’t this a sign that you are following too closely?
If you open enough space ahead of you for a car, someone will cut in. Driving too closely isn't something only you can do something about.
In fact, because of the aforementioned behaviour, it's safer to drive too close because it's far more dangerous for a car to cut in in front of you because he perceives "enough" space to do so.
> Isn’t this a sign that you are following too closely?
But they said it's FSD who's following too closely.
I'd expect automated system should react even better, because it pays attention constantly. That's the whole point, especially when it's called "Full".
There's a lot of technical measures that prevent people doing harmful actions. Even in spaceships, where pilots are highly trained, focused, selected professionals.
For example, most cars prevent fiddling too much with its entertainment system while driving, even though drivers are expected to not do that while driving.
> I can't really blame it, it took me half a second to figure it out as well and by that time I was hard on the brakes.
To me, it seems like this is where a computer should be a million times better. It might take me or you a half second to figure out whats happening, but software should be able to immediately detect that a car is stopped in front of you.
This shouldnt be a problem at all with a LiDAR based system, but I can see how a camera system might have trouble with it (possibly indicating that LiDAR should still be used for at least the near future)
Tesla FSD (but not autopilot) deliberately tracks the car in front of it, as well as the lane and tries to follow that car (and takes it as a safety signal and slows when the car above swerves out of the lane). It's interesting behavior, and usually annoying, but this is why it's there.
Most radar cruise control is set low on the bumper which allows visibility into the vehicle in front. My 12 year old car with adaptive cruise control will sometimes begin slowing when the car two cars ahead starts slowing. Sometimes because it requires a truck or suv or otherwise a vehicle with a higher ride height for radar to pass under. When it’s a Miata in front of me, I will only track that Miata because it’s so low. At the same time.. I can see over these lower cars. The ford flex however is a car that stands out as a low sitting vehicle that is rather tall for its ride height.
That said none of this matters if the object or vehicle in front of the following car is stopped. The radar system will ignore it as part of the background.
I would've thought a stopped vehicle without any visible lights would be much easier to detect with a lidar/sensor based system vs human reaction.
Do these platforms like BlueDrive and FSD not constantly survey the distance of objects in front of and around you? Or do they still rely on visuals like tail lights?
A system with lidar would have definitely detected the obstacle and braked. For a lidar, which is an active sensor, it doesn't matter if it's daytime or complete darkness since it makes its own light.
FSD is inferring distance of objects as it has no direct measurements. It's a real challenge in darkness since camera is a passive sensor and it has to rely on visual cues. I believe BlueCruise is using a radar, but I'm not sure how robust it is.
I believe BlueCruise uses radar, gps, and mobile eye which is dual camera inputs. At least my 2019 ford uses radar+camera combination for lane centering, ymmv.
Edit: looked at a Mach e and it uses dual cameras behind the rear view mirror so likely intel’s mobile eye platform
The problem is the stopped car was hidden by the moving car that avoided it successfully. Only when that car swerved to avoid the stopped one did the BlueCruise car have line-of-sight to the stopped one. There was no visual at all, and no lidar, but possibly some indirect radar.
And radar generally ignores all stopped objects because there are too many false returns. It’s really only useful for objects traveling at some speed. If the CRV had been going even 5 miles an hour it maybe it would’ve been detected.
Every radar system has this issue. It’s basically the same thing that caused that Tesla to run into the side of a semi so many years ago. The semi was stationary (relative to the road direction) so if the radar saw it probably ignored it.
I’ve never heard of a system that uses the cameras. Except Tesla since they removed the radar I guess. No one else seems to.
Sorry. I meant cameras only. As far as I’m aware every system on the market uses cameras for lanes and radar for obstacles ahead except Tesla since they removed their radar.
I’m guessing (but didn’t check!) Subaru also has the radar too.
Need to understand the modular nature of cars. The radar is one system and camera is another. Camera is only used for lane detection the only output from the camera is if road lines are detected and how close to either line. Based on this the car can “ping pong” since it detects it getting too close to one line, or if it has two cameras it can determine what the center of the lane is by keeping the distance from each line equal.
Radar typically applies to the ACC system while camera applies to LKAS or lane centering. Rarely are the two connected. FSD and comma.ai are probably the only systems which combine the two.
Subaru EyeSight cameras (there are 2x front cameras and 1x rear camera) are also used to detect moving and stationary objects, and it will engage the emergency braking if, e.g., you are driving or reversing into a wall. Lane detection can be disabled on Subaru cars, yet the EyeSight will continue to scan surroundings and be ready to start breaking.
As of this year, the EyeSight is available even on Subaru BRZ 2nd gen with the manual transmisson.
It is just emergency braking (electronic brake distribution across all wheels according to sensor readings), plus ABS in a hard braking scenario. The clutch is not affected.
You don’t have to, the car will stall out. Another scenario: you are in 5th gear and the cruise control slows down to 15mph. The car stalls. My manual did that 25 years ago
I got in a situation like this in my model x a couple years ago. I was going 125 km/h using autopilot on the highway towards Sweden in quite dense traffic and a car two cars ahead of me crashed.
When I hit the brake, it was already max engaged by the car, probably .5 seconds ahead of me.
I managed to stop, but the car behind me had to steer into the side to avoid hitting me in the rear.
I figured it that day that a proper emergency brake system in the car is worth the money!
Edit: I always drive with the max distance to the next car when on autopilot.
> I always drive with the max distance to the next car when on autopilot.
This. Also the Tesla Model 3 max-distance setting is really not big enough for my taste, especially for winter driving. I have a feeling that it does not even satisfy the legal requirement of 1,5 seconds we have here.
I have had exactly this on a 70mph 2 lane road in Britain known as a dual-carriageway. I will phrase this carefully because we drive on the left, so the overtaking lane is on the right.
I was following a 'Luton van' (Which is like a small truck body on a normal 3.5tonne class long wheel base van chassis, you might call it a small box truck in the US). Incidentally Luton Vans are not allowed to go 70mph on a dual carriageway, they are limited to 60. I couldn't see past at all due to his wide body, so I was following at a safe distance and matching his speed at 70mph. Suddenly he swerved into the overtaking lane and I was met by a queue of stationary traffic! I knew this road well and stationary traffic would not normally be backed up this far. I braked very hard and felt the ABS rattle. I had just enough time to realise that there was a narrow gravel shoulder on this section I could turn into if needed, but the car grabbed just in time to not need that option. It was one of the scariest things to happen to me on a road. If I had been tailgating the truck I would not have had nearly enough time to stop and it would have caused a really nasty accident as I would have hit a stationary car into a line of stationary cars. I don't see how lidar, radar or cameras could have helped much, except to have reduced the reaction time slightly perhaps? It probably would not have been able to judge that the shoulder was a good alternative option.
Emergency Brake Assist would help if the van was equipped with it.
Swerving at the last moment and without warning would not be possible because EBA would activate brakes earlier and you would be able to see van's stop lights.
I have had a couple of scares driving at night, even doing my best imitation of a tree full of owls. Dark cars w/o taillights (and that happens way more often than I'd like) on a street with poor lighting (ditto) is the perfect recipe for a disaster. I don't know why people drive without lights, but they do, and it's a damn nightmare when you come across one.
Sometimes, you do everything right and the Universe tries to destroy you anyway.
Why doesn’t every country mandate daytime running lights like Canada has for decades?
Modern cars also always detect low light and crank up the dashboard brightness. There’s no longer a good clue to the driver that they forgot to turn the headlights on.
I would argue that daytime running lights is the cause of many cars driving without taillights. In the US, DRL are only forward-facing. Nothing rear-facing is illuminated.
Drivers that have switched their headlights to the off position, the DRLs are "bright enough" to drive and not notice your headlights are off.
I see them very often, and for some reason a large portion of them are Hyundai. Maybe because the dash light, button backlights are independent of the headlights, so there's no real indication inside that your lights are not on, except the switch position.
I pretty much always keep my lights on Auto. But I've had them turned off when I've had the car in for service and, probably because previous vehicles haven't had DRLs, I haven't noticed on a couple of occasions.
My 2017 Golf R would intentionally turn off the dashboard backlights at night if the lights weren't on, providing me excellent feedback that they weren't on. Either head lights should always be on auto by default, or more cars need proper feedback to the driver when they aren't.
What we don’t have is mandated automatic headlights. I remember there was a discussion recently about requiring automatic headlights or maybe not automatically increasing the brightness of the dash when it’s dark out. But I don’t remember what happened with it.
> Why doesn’t every country mandate daytime running lights like Canada has for decades?
Not everyone drives an up-to-date vehicle and TBH, some vehicle owners do not care that they are a hazard, regardless of what the law requires. Here in Austin, when the police get in a snit, they just stop enforcing the law and even when they're not in a snit about something, there are parts of town where traffic laws are traffic suggestions. All of this makes for a somewhat hazardous driving environment at times.
>Why doesn’t every country mandate daytime running lights like Canada has for decades?
Because for most countries in the world that have sun bright sun most of the year it is totally ridiculous.
Spain, Argentina, Morocco, Nigeria. Greece, Italy, Thailand, Filipines, Mexico...
Those countries have ridiculous daylight saving time changes because countries in the North believed this hack was a good idea.
The darkness Canadians experience in Winter and low sun in Summer are conditions in Norther countries other countries do not have.
> Spain, Argentina, Morocco, Nigeria. Greece, Italy, Thailand, Filipines, Mexico. Those countries have ridiculous daylight saving time changes because countries in the North believed this hack was a good idea.
I'm curious to know where this list came from. Argentina, Morocco, Nigeria, Thailand and the Philippines do not use DST.
Spain, Greece and Italy do use DST, as members of the EU I believe they are obliged to do so.
There are very few good reasons to have your headlights off while your vehicle is running. Having a switch to manually turn them off if you're i.e. boarding a ferry and don't want to blind workers is fine but the default should be full on all the time - daytime, night time, adverse weather, whatever.
This is a very very very common way of dying, and is why German highway kits require reflective vests and warning triangles (just in case your emergency lights are broken or even as extra redundancy). It might be a good idea to carry road flares along as well. This reminds me I should update my kit even though I'm in the US.
This situation nearly happened to me just a couple hours ago - thankfully I was maintaining a proper distance from the semi truck in front of me as it swerved left out of the lane - otherwise I would have had no chance to notice the car going <15mph with no lights on, in the middle of the night, on a 70MPH interstate.
...it's an example of a common human driving behavior whose setup conditions lead to a high probability of collision if a peek-a-boo scenario were to manifest, namely by:
- obstructing the rear driver's ability to assess forward traffic conditions, especially if the leading vehicle is larger and/or has dark tinted glass; and
- posturing for insufficient clearance/time to safely react to an abrupt change in forward obstacle speed upon surprise discovery.
The tailgater pejorative is what drivers are most likely familiar with; implied risks are much easier to internalize when framing the scenario from this oft encountered context. It's just one possible path towards the conditions and outcomes being discussed, which is what really matters.
Sure, this might naively beg the question: Was the BlueCruise vehicle tailgating as a typical driver might characterize?
If yes, then the outcome should be unsurprising, and it makes me wonder how such an operating mode is even permitted.
If no, then it begs the even harder question of whether the automation tech is operationally mature enough to safely handle such a common known unknown.
My guess would be not adhering to the "keep right except to pass" signs on the highway. I don't tail gate, but I do find it rather annoying when people block traffic by driving right beside another vehicle rather than passing them and clearing the lane.
So that we can all not adhere to the "maximum" speed signs on the highway to the degree that we prefer :)
I was a passenger while a friend was driving once and he said "I'm already going fast enough. Nobody should be going faster than me anyway." So there's one mentality.
I see a number of negative stories, so sharing my positive story with self driving cars.
Two weeks ago driving home, Tesla Model Y with FSD, FSD on, nighttime, suburban streets. Car phantom brakes and I can’t see a reason why. Glance down and I see the UI thinks a human is running into the street.
I look back up, canceling FSD and taking over… and a damn invisible guy in dark brown jacket and black pants races past my headlights.
He had stopped before crossing my lane, but when he saw FSD slow down he booked it across. Other cars slowed after he cut straight into my headlights as they noticed him for the first time. No one else saw him either. I was super impressed with FSD.
The promised perfect future isn’t here - you can’t take a nap and the car drives you - but the current state is a value add.
The "funny" thing, though, is if you hadn't been using FSD at all, you probably never would have seen him at all, and he wouldn't have crossed until after you passed. FSD slowing down is what caused him to think it was safe to cross, because he thought "you" saw him. But you -- very reasonably! -- thought the car was slowing for no reason, and you might have hit the guy had you manually accelerated sooner. Not saying this was your fault in the least. Just thought it was interesting that FSD -- in a very strange way -- could have actually caused you to hit a pedestrian, even though it saw the person and did the right thing.
This sounds like terrible practice and you are gambling with you life. And probably the lives of those around you.
Fine, you might have right of way. Fine, the car driver might legally be in the wrong. But you'd still be dead and nothing changes that. You are not invincible, and the law will change fuck-all about you being in a wheelchair breathing through a tube for the rest of your life. I doubt youll even change the behaviour of the at-fault driver that much.
Lobby your council for better infrastructure for pedestrians, and steeper fines for infractions. Well lit and marked zebra crossings, etc. In Ireland _every_ zebra crossing outside of Dublin has bright flashing yellow lights on both sides that constantly flash day and night. If you are half asleep at the wheel they sure wake you up as you approach the crossing!
In Seattle, if a pedestrian wants to cross at any no-stop-light intersection (it doesn't matter if it is across a busy stroad!), cars are supposed to stop and let them cross safely, even at night. Yes, it is a poorly thought out law (unsafe for both pedestrians and cars, without improving the intersections to make them safe for pedestrians by not adding crosswalks, not removing pedestrian obstructing parked vehicles and trees, etc...), but in that sense even if the guy were in full black camo on a moon-free night without street lights around, cars are still expected to see and stop for them.
So in that case, if the pedestrian was at any kind of intersection, FSD would be doing the correct thing.
I've always been taught that pedestrians have the right of way, even when they don't. But I don't think that applies to invisible pedestrians.
Of course, Seattle is home to lots of terrible traffic designs and practices. I have relatives in Ravenna; in their neighborhood people park on either side of the street and residential intersections signal a 4-way stop by absence of any stop signs.
I teach my son that he doesn't have the right away, no matter what the laws really say, because even if the car is at fault, he is still dead (and that scares me). Cross walks only for now, and definitely no going to try and cross 15th NW (a four lane that carries lots of traffic) like that.
It does apply to invisible pedestrians, but cops are more likely to rule it a drive-at-fault accident than something more serious like negligent homicide. I live in Ballard, people park on either side here also, and...there isn't much buffer even at designated cross walks to see people waiting (or just going) to cross.
I love your comment because it highlights another complexifier of FSD debates - what was the actual right action?
Yep, pausing signaled to the guy he could run across. However who’s to say he was going to be reasonable and wait? What if he wanted to get hit, or didn’t care? How much influence will those edge cases have on our (human or FSD) norms?
As a society we haven’t answered those questions; it further muddles the FSD debate.
As the tech improves it will be interesting to see how we humans define the rules of engagement.
It's very clear to me, let FSD stop the car, that puts all further blame on the guy if he dies or something happens. We should do everything in our power to reasonably and safely prevent it, and letting FSD stop does so, the rest is on everyone else.
Yep. The key words are: to you.
When I have convos about this, a spectrum of views come out.
Other views I’ve heard:
- pausing encourages ad hoc crossing, increasing odds of hurting someone by volume even as odds decrease.
- if you stop you enable carjacking (walk into street, car stops, steal car / hurt the human, profit). I don’t want my car to set my loved ones up to be attacked.
- I don’t want the car to stop unless it gets to my destination, there’s a red light, or I tell it to stop. I don’t want behavior I might not understand. I’ll decide if I stop for a human or puppy or cardboard box.
Everyone says human safety first but in practice there’s a spectrum of opinions in practice.
My hope is FSD will force a common viewpoint and setup a future generation with a more unified approach to road safety.
Same thing has happened to me. FSD sounded the alarm and hard braked on a dark two lane highway. I couldn’t see why until I fully passed someone dressed in dark clothing who had just crossed the road. I was paying full attention and never saw them. They were probably a couple hundred feet from the front of my car.
Over 10 years ago. Colleague comes into the office and says: "I nearly died yesterday." Tells a story how he (as a passenger) was driving with other colleague to an exhibition. The driver thought it was a good idea to explain the car radio to the others...while driving over 200km/h on the German Autobahn. The car, a VW Passart, suddenly engaged full braking while slowly steering to the right (onto the breakdown lane). They came to a full stop besides a standing car.
Without the auto braking they would have driven into the end of a traffic jam with very high speed.
There is no moral. It just happend. The End. People have to travel sometimes and can't always be the driver/pilot/operator of the vehicle. And if they are, they can't always be at their 100%.
BMW still has it. Mine will display a large warning triangle with a pedestrian in it projected on the windscreen, and then you can look at the nav display to view the night camera view from the grill with the person highlighted in yellow (the rest is black and white for night vision)
It responds to bicycles and pedestrians either on the road or moving laterally across your path. Ignores the ones that are on the sidewalk and going the same direction as the car.
"Snake oil is a term used to describe deceptive marketing, health care fraud, or a scam. Similarly, snake oil salesman is a common label used to describe someone who sells, promotes, or is a general proponent of some valueless or fraudulent cure, remedy, or solution"
Selling something as "full self driving" or "autopilot" when it is nothing of the sort is in fact deceptive marketing. Claiming something will be "ready in 2 years" for 10+ years in a row is deceptive marketing and a scam.
Fine print saying the system has limited capabilities does not make it not a scam. Using a name consumers associate with 100% computer control is flat out deceptive and indicative of snake oil.
It's not a value-add given it causes chain-reaction crashes from phantom braking and quite regularly slams into the back of emergency vehicles on the side of the road (likely because the cameras are blinded by the emergency lights.)
Teslas had radar assist. It was too expensive for Musk. And to keep new car owners from being annoyed by older Telsas having more features than theirs, he ordered that the older Teslas have their radar units be disabled.
Actually, it appears that getting rid of "sensor fusion" has resulted in almost all of the phantom braking issues. My car still has radar assist, and only uses visual models now. And in that time, Phantom breaking has become a thing of the past.
The hitting a emergency vehicle thing - which is incredibly common as a kind of car crash - see why police always set their wheels to face into traffic when they do a traffic stop - doesn't seem to be occurring.
Having worked with these kinds of systems, it can be difficult to tell from the driver's seat the actual effects of that change. It may have reduced false positives and true positives. It may have simply moved the false positives into scenarios you don't encounter. It may have not affected the true positive rate at all, but might have made positive identification slower and thus violated a real-time constraint elsewhere in the system leading to accidents. I've actually seen this last one root caused before.
The only people who have the data to know whether removing radar was safety net positive or not work(ed) at Tesla and can't say. Even then, they probably wouldn't know whether the system could have given better results with different changes.
Ehh. I had the same streach of road that resulted in three phantom breaking situations (one of them not in a tesla, but in a toyota). Steep descent on a road with a rail road crossing with exposed rails. No issues at all once the vision only upgrade hit both my M3 and MY (which has HW3). It's possible that they put in some other fix at the same time, but they stated at the time that sensor fusion was a big reason for phantom breaking, so I take them at their word that this was fix.
That's not quite their point though. Completely deactivating the emergency breaking system would also solve the phantom breaking problem, and since the average person does not end up in situations which ought to trigger the feature that often it's unlikely any individual driver would ever notice that is was broken.
I'm not doubting that it went away for you. You know what you've experienced a lot better than I do. Everything I listed is consistent with that observation though and it's impossible to differentiate between them without internal data.
I’ve owned 2 cars with radar. I’ve been in at least 2 others owned by family with radar.
None Teslas. None had phantom braking.
Tesla may have solved their issue (no 1st hand experience) but having radar shouldn’t have caused it if implemented properly. It’s clearly possible, seeming every other brand has managed.
Really? I saw phantom breaking all the time from both Toyota and Hondas. Typically on a slope, with something like rail crossings where the signals got a little squirrely. having the same glitch across multiple vendors with different implementation makes me think it's more a basic physics problem.
Regardless, the Tesla vision system - love it or hate it, is doing far far more safety systems then radar only based systems.
In the 6 years I had it I got a handful of false warnings from my Honda, but never a false brake activation.
In the three years I’ve had my Mach E has had fewer false alerts than the Honda.
The one exception is it has a reverse brake assist the Honda didn’t, designed to stop you from hitting things when backing up. One friend’s driveway is so steep where it joins the street that if you’re not going very slow it will trigger the reverse brake assist. That was a surprise the first time.
Anecdotal I know, but I bought a model Y last year without radar. We’ve used FSD substantially and haven’t seen phantom braking issues even once. I think it’s been mostly resolved.
I've never driven one but I follow a youtube channel that documents beta FSD, posting raw video feeds and analyzing/critiquing them, and I've noticed they've been gushing about it more and more this past year. Far more than in the past he's always talking about how its mastered very complex stuff it would have disengaged or awkwardly completed only half a year earlier. I think the platform has improved quite a bit in recent years.
Of course opinions on the internet will have been established from headlines in prior years so who knows how much impact it will have on discourse if it turns out to be the case.
A car stopped on the highway with no warning lights or anything in center lane, at night... I'm not sure I'd be able to spot and react in time, this is what I would expect radar and other sensors on car to pick up and break on... those also failed...
Especially on Texas highways where speed limits exceed 80mph, you'd have to be very alert and likely have to swerve over into the next lane (like the other driver did).
At 80mph, the stopping distance of an average car is around 300 feet once the brake is fully engaged. That exceeds most car radar systems. And with no warning lights the camera won't do much good either.
This seems insane, are you telling me that these cars will travel at a speed whose distance is longer than the radar can detect? That is _guaranteeing_ crashes.
High beams are really for rural roads with very little traffic. If there are at least three lanes of traffic in one direction[1] and enough traffic that a car in front was able to see the collision in their rear view mirror, high beams are not advised.
Using them in this situation will cause major glare from mirrors of the drivers in front, and possibly for the drivers in the opposing direction.
[1] as indicated by the stopped vehicle being reported in the center lane
I mean yeah if you're on a pitch black highway and you can't see anything outside your headlight beams, you probably shouldn't be going 80 even if that's the speed limit.
No street lights doesn't imply pitch black. It may imply poorly illuminated. I can't find a date, only a time, so I can't guess at lunar illumination or cloud cover.
But we don't know how many other vehicles were on the road and what light they were casting. The bystander who reported witnessing the crash was within visual distance, so couldn't have been too far ahead of the MachE. It wasn't reported if there were any other vehicles. If you have enough vehicles on the road (with their lights on) lack of overhead illumination isn't too significant.
Some vehicles (like my Subaru) offer automatic emergency steering that can save you when there's not enough distance to brake. The Mach-e doesn't have that per my quick Google (they have some assisted thing but it requires manual steering)
> At 80mph, the stopping distance of an average car is around 300 feet once the brake is fully engaged. That exceeds most car radar systems. And with no warning lights the camera won't do much good either.
I think this is a pretty important point for any well-built self-driving software. Does anyone know if radar advancements could help increase this?
The big problem is that a stopped vehicle looks exactly the same to the radar as any other stationary object.
So if it sees something stationary it could be a stopped vehicle or it could be a reflection from a road sign. Or a lip in the pavement. Or something else like that.
Basically from what I’ve heard if it were to try and stop whenever it saw something stopped ahead it would constantly slam on the brakes unnecessarily. So stationary things have to be ignored, and it can’t save you from them.
They all suffer from this. The only solution I’ve heard of is LIDAR but obviously that’s a huge step up.
This is why cars are required to have red retroreflectors on the back, so even if their lights are off you can still see red lights as long as your own headlights are on. I wonder why this failed.
Retroreflectors are good but at a long distance you need a significantly brighter light source/reflection to be visible (intensity falling off with distance squared). At highway speeds that distance for safe stopping is probably in the hundred+ feet range.
It seems like in this case the retroreflectors may have worked. The problem is the car that saw and avoided them blocked the BlueCruise car from seeing them until it was too late.
Exactly. It's easy to avoid a stationary object on the road in normal visibility out if your headlights work a damn.
Maybe I am just too accustomed to driving on rural third world roads where cows wander at night.
From experience, coming up to traffic jams on highways in the past, you can get deceived by cars ahead either stopped or slowing down - you dont expect that, especially after long monotonous driving, hence need for warning lights.
Indeed. I have never driven more than 4-5 hours at a time.
I find driving very taxing for me. I tend to be very focused on the road and drive like everyone and everything is trying to kill me. For this reason I don't use driving aids either not even standard cruise control.
I find adaptive cruise control definitely helps with the mental fatigue as well, because I can concentrate on looking for hazards far ahead instead of having to pay attention to keeping distance to the car ahead. But yeah, I never drive more than a few hours without taking a break anyway.
Especially if they don’t slow down/swerve until the last second. If they had done it earlier you would’ve had more time to start noticing something was going on.
This probably needs to be added to a safety test suite. Either for the cars or for drivers.
Most adapative cruise control systems explicitly ignore stationary objects while at highway speeds. A vehicle immediately ahead that stops will be detected and responded to. A vehicle that did not move while in sensor range will be ignored.
Radar struggles to identify stationary vehicles. Their reflected signals hide in noise from road signs, roadside vegetation, and the road surface itself.
It's worse than that, I think. The typical radar beam is so wide that it can't discriminate between in the road and near the road, so the Doppler selection is the only reason it even works at all. This is very noticeable on our car when you observe how far away laterally a car you've been following has to turn, e.g. when it takes an offramp, before the ACC ignores it.
What brand and model year is that? The first (15 years ago) system from BMW did that too, but the newer ones (2019 model, don't know about the years in between) know which lane the other car is in. There is a setting on European models on whether to overtake on the right or not, so it knows the traffic ahead is in your lane or in the next lane over.
I came across this a few weeks ago on a snowy highway.
Car way in front of me had its hazards on in the right lane. The center lane was clear, so I decided to make a lane change. I assumed this car was just moving slow in bad conditions, but it was actually completely stopped in the right most lane.
I didn’t realize that until I passed it. Given the conditions, it would have been extremely hard to avoid this car if I had not proactively changed lanes.
There was just absolutely no reason this car should have been stopped in a lane when it could have gotten to th shoulder.
> According to the NTSB, a witness had come across the stopped CR-V prior to the collision and noted that neither its taillights or hazard lights were illuminated
I’m shocked by how many people drive at night with daylight running lights in US, they seem to be completely oblivious of that, even when I stop next to them at a traffic light and try to tell them - they fail to understand that daylight lights don’t have a rear light
I sometimes wonder if the daytime running lights do more harm than good, my making it harder for people to realize their lights are off.
That said, I don’t know why automatic headlights aren’t standard at this point. My base model Toyota from 25 years ago had them. It doesn’t seem like any car made in the last 20 years should require drivers turn on their headlights when it’s hard.
Possibly because early auto headlights were annoying and then there was backlash? I've got a 2003 s-10 with auto headlights, and it's reasonable, until you want to turn the headlights off (which is required when driving onto a ferry, very important at drive-in movies, and maybe there's some other reasons). There's no user visible switch, you just have to toggle the dome switch 10 times and the truck beeps to acknowledge. If this was my first vehicle with auto lights, I would have avoided it in future vehicles; but I got the cars below before this one (used light truck, because I needed a light truck and nobody makes light trucks anymore / I'm not putting enough miles to justify a new purchase anyway, unless it replaces an existing csr)
I had a 2007 Ford and a 2011 VW without auto lights at all. My current 2014 Ford and 2017 Chrysler do have auto lights, but it's the last position on the light dial: Off, parking lights, head lights, auto. So you can easily turn them off when you need to. Might be nice to have a chime and warning text for 'lights not in auto' on start / at intervals, too. Because driving off a ferry during the day, you don't always remember to set it back to auto until the next night driving.
The Mach E has auto headlights with the same setup you describe. It’s a 4 way dial. Off, auto, parking, on.
Every time you start the car it’s in auto.
I’ve found it to be a perfect system. On the other hand a family member’s car has it but acts more old-school. It stays at whatever you last set. Last set to off? Stays off. Makes it easier to forget.
Sounds good, although I tried to find a picture by searching and instead found many reports and a special service bulletin about needing a switch replacement because the switch would turn on by itself. Whoops.
My old Toyota didn’t have an off either. I had to turn the car off once just to turn them off in a certain situation. That seems like an easy fix, like you mention. No reason to abandon them completely.
I had a 2016 VW with them and every time I took it to the dealer for service they’d turn them off and I’d have to set it back to auto. I’m sure it was done as part of their testing, but it seems like a horrible default state to set them back to.
I agree that some kind of warning when they are off, maybe a light on the dash, would be good.
> I had a 2016 VW with them and every time I took it to the dealer for service they’d turn them off and I’d have to set it back to auto. I’m sure it was done as part of their testing, but it seems like a horrible default state to set them back to.
Probably engaged in the shop and they didn't like headlights in the eye.
Daytime running lights (which are now mandated in many countries) and headlights are distinct, separate and independent from each other – the former are not as intensely bright (and are oftentimes LED contour light or a separate, dimmer globe – at least on cheaper vehicles) and the latter are activated by a sensor when the ambient luminosity drops below a certain threshold «on top» (so to speak) of the daytime running lights.
IME it's not people with their lights on auto that are the problem, it's people who just haven't turned their lights on. Around dusk it may be that it was light when they started driving, and they haven't been paying attention.
When I'm driving I can occasionally communicate the problem successfully by flashing my lights, but I'm usually on a bike, and it's almost impossible to get it across visually or audibly, and so they stay dangerous.
No, he means what he said. You shouldn't drive at night with only daytime running lights. You need to turn the headlights on because they are brighter but more importantly because that turns on the taillights too.
Now we just need people to treat car crashes like plane crashes even when humans are driving.
Unfortunately, given that that would result in dozens to hundreds of fatality accident investigations a day, this is likely impossible.
It's like self-driving cars suddenly sidestepped a hundred years of normalized deviance and everyone is finally, suddenly realizing that cars are fundamentally, spectacularly, and unrelentingly lethal.
The CRV was stopped with hazards and lights off, at night, in the center lane of a 10 lane freeway. Certainly a human driver could have easily made this mistake.
I've never really expected self driving cars to lower the rate of accidents simply because cars are dangerous whether or not you have the reaction time of a computer.
Can it save the human in a moment of inattention? Absolutely, my prius has shouted at me to brake when I was messing with the radio, but it doesnt matter how good your radar is when you're flying around a blind turn or coming up behind an invisible, stopped object.
(%50 reduction would be good, drunks taking robo taxis home is great, but IMO driving is easy and robots wont be better at it)
> Moravec wrote in 1988, "it is comparatively easy to make computers exhibit adult level performance on intelligence tests or playing checkers, and difficult or impossible to give them the skills of a one-year-old when it comes to perception and mobility".
> Steven Pinker wrote in 1994 that "the main lesson of thirty-five years of AI research is that the hard problems are easy and the easy problems are hard."
Stuff like checking the rear view mirror when slamming the brakes upon some surprise uncertainty, true. Humans will always be bad at this, but for a computer it's trivial to include the tail situation in the brake decision.
I think the issue is that basic ACC doesn’t understand intent and the larger picture. If someone merges in and I have control, I can usually just let off the gas a little bit, as they will likely speed up, or move out of the lane shortly. Actual braking is almost never required. But the ACC just sees some slow moving object inches from the front of the car and flips out to try and get back to a good state.
But this paragraph from the report clearly states that at least one human did not. A human driver avoided the crash and a computer control system crashed.
It’s not unavoidable. But it’s a black car stopped without lights in the middle of a freeway at night. It’s just a matter of time if no one intervenes (e.g. cops with flashing lights drawing attention).
Ehh. I think it's actually 100% avoidable with a robust self driving solution. One with a smart enough margin of safety would never be in a situation where it has so little information that it initiates a crash on the highway
"According to the NTSB, a witness had come across the stopped CR-V prior to the collision and noted that neither its taillights or hazard lights were illuminated, She was able to change lanes and avoid hitting the Honda, but after she passed it, she saw the Mach-E strike the stopped crossover in her rearview mirror."
That means the witness car likely obstructed the view of the stopped vehicle, then changed lanes at the last minute, revealing the stopped vehicle after it was too late for the Ford to stop.
> Certainly a human driver could have easily made this mistake.
except a human driver in front of the Mach-E did see the car and avoided it successfully.
that's why cars with driver assist features are supposed to be using radar to detect obstacles.
im not familiar with the Mach-E's feature here that's referred to as "handsfree". Usually these features are supposed to make sure the driver is still in control of steering at all times, like the Ioniq 5. I'm probably going to be a teeny bit more wary of Mach-Es on the road if they are using a not totally reliable radar detection system.
except a human driver in front of the Mach-E did see the car and avoided it successfully.
That makes it harder for the car behind. The Mach-E is following a moving car, which is blocking the view of the stopped car. She swerves out of the way, and the situation goes from everything normal to disaster in half a second. A difficult situation for humans or autonomous system.
The Mach E just shouldn't have been following the way it was. Autonomous systems need to maximize their road information, not follow social trends (i.e. not follow closely on the highway at the sacrifice of visibility)
That can be true, while also being true that a moving car in front of you can better see an obstacle while obscuring it from the car behind. I’m just saying that the car in front that avoided it was in a different visibility situation to respond than the following car.
BlueCruise hands free is truly hands free. You do not need to keep hands on the wheel at all. It can steer on its own and handle the throttle. It will not change lanes without the driver signaling. It just keeps going forward in its lane at the set speed subject to the car in front.
It uses the radar, GPS, laser highway maps, and the car’s camera system to do the driving.
There is a camera directly behind the wheel monitoring the driver at all times. You don’t need your hands on the wheel but if you look away for more than a few seconds it will disengage.
Also notice I mentioned laser highway mapping. It is only available on premapped divided highways that have no crossings, only on ramps and off ramps. Some areas will not be supported when they have otherwise would be due to a sharp bend or the number of cars entering and leaving the road there, or if the sensor data doesn’t match the maps (construction/etc).
I’d say it’s a very well designed system from a safety perspective. It does its best to ensure it only runs when it’s capable in areas pre-determined to be within its capabilities.
But it’s still level 2. The human has to be engaged.
If the car in front of it was driving like normal and swerved away from the stopped CRV late, the Ford wouldn’t have even known that CRV was there until the other car swerved out of the way. I’m not aware of any ADAS system that can interpret what a sudden swerve might mean. And no system has radar that can safely detect a stopped vehicle.
This is a situation it was never designed to handle and why the driver is required to pay attention the whole time.
If it was on at the time then either the human was paying attention and didn’t do anything or they had very recently looked away. Unfortunately at TX highway speeds a few seconds is a lot of ground.
Automotive radars often ignore stationary objects because they have trouble differentiating between a car parked in a lane vs. a car parked on the side of the road vs. an overhead metal road sign. If they didn't ignore stationary objects they would be slamming on the brakes in the middle of the freeway all the time.
Perhaps true, though I suspect the public will demand higher safety than humans to feel comfortable with these on the roads, even if such fears could be irrational.
Not at all. I don’t care too much one way or the other for driving jobs, but I do care quite a bit about self driving car safety. They are being trained in an imperfect real world, and I don’t particularly want to find out if they ignore unmarked crosswalk laws (or other safety related stuff) as thoroughly as human drivers do.
I’m not alone. I’ve talked to plenty of people that recognize that car infrastructure is inherently unsafe in many places, the last thing I want is an incentive to have more vehicles driving around because now you don’t need humans inside the car.
Of course people care about safety. That doesn't translate to either public action nor public action requiring self-driving cars be safer than human-piloted ones. Look at every regulatory proceeding around self-driving cars, and the safety arguments are either drowned out by the jobs arguments or made principally by people motivated by job-loss concerns
My gut feeling is we're perfectly happy with self-driving cars that are even substantially worse than human-piloted ones, so long as they crash the way humans do. Rear ending in the dark, okay. Barreling full speed into a highway divider [1] or dragging pedestrians to get out of traffic [2], not.
> My gut feeling is we're perfectly happy with self-driving cars that are even substantially worse than human-piloted ones, so long as they crash the way humans do. Rear ending in the dark, okay. Barreling full speed into a highway divider [1] or dragging pedestrians to get out of traffic [2], not.
No. I’m very not ok with this. Many human drivers are terrible. Most collisions are entirely avoidable, and I certainly don’t want more of the same from computers. If I keep getting my life put at risk more often by incompetent driverless cars, I’m going to be pissed! Maybe people that spend their time in cars agree with you, but as a cyclist and pedestrian the common accident modes are the ones that pose the greatest risk to me. Pinball off the jersey barricades all you want, but stop passing bikes too close, and plowing through crosswalks just because you can get away with it.
I’ve been hit twice on my bike, both times 100% the driver’s fault. The stakes are life and limb for me, so I absolutely do not accept a higher accident rate.
The owners manual clearly says you are still responsible and it can’t detect stopped vehicles. The system does its best to ensure you’re watching the road with cameras to track your gaze.
Laws always tend to put the driver as the last stop in the chain responsible too.
This is very unfortunate. There’s a good chance that CRV was going to be hit if the Mach E avoided it.
But the person in the drivers seat is still the one responsible.
I would love to know the stories of both the Ford and Honda drivers. Doubt the public will ever know.
Though such rules and precedent may be clear, they don't put any blame on road design or manufacturers of cars. Someone driving 150mph--it is the drivers fault even though the manufacturer knows that there is no road that this is possible.
> Certainly a human driver could have easily made this mistake.
Certainly a heavily inebriated driver might have made this mistake if the car were difficult to see at a distance, but otherwise... no, I don't see this happening at all.
You lack imagination and/or driving experience; see https://en.wikipedia.org/wiki/Highway_hypnosis. that said, driving tired is not uncommon and just as bad if not worse than driving drunk.
I've experienced this many times as I used to drive a lot at night. To me it was like half of my mind falling asleep while the other half followed the white line for hours. I used to drive close to 70k miles at night every year, all along the same routes, with no stimulation, just the white line. I know I wasn't asleep or tired, but it was like driving in a dream sometimes.
It is probably much, much more common than we know, and I think this state would completely fool the attention-tracking systems in self-driving cars.
What is worse, many people drive the same routes commercially every night, like sleeper coach drivers. I think if they even did so much as spoke about this, it could significantly harm their careers. So it's unlikely that this demographic that could be most affected will talk about it openly. But there is almost nothing you can do to prevent it. After a certain number of thousands of miles along the same route, seeing the same white line, for hours, in the dark, you zone out.
Another thing that's making it worse is that I could only ever catch myself in this state when I already did something irresponsible. Like I'd go past a speed limit sign without reacting, and in a couple of seconds my brain would wake up to say "hey, something is happening, let's pay attention again". Otherwise, it's almost impossible to catch your mind drifting gently away.
I'm happy to finally put a label to the experience.
I easily see it happening especially knowing the Ford was following another vehicle that could have easily completely occluded the stopped car until only a few of seconds before impact.
I own a model 3 with FSD. One of the key realizations I had after a few months with it is that self-driving at the moment is like supervising a newbie teenager driving. You have to stay alert and watchful, but also don’t have the direct control. When you are teaching a new driver it’s worth it.
Honestly, it’s a lot less work to just do it yourself.
This is one of those engineering situations where the 0.0001% edge cases matter, and could lead to fatalities. I don’t know if any of the implementations are up to the mark. FSD isn’t.
>> I don’t know if any of the implementations are up to the mark. FSD isn’t.
Having talked to engineers in the industry over the years, I feel confident to say they are not.
It's a problem where the base functionality - following a well marked lane - is actually f-ing easy, but every nonideality beyond that gets harder and harder. So people look at the base case and get all impressed. It really takes AGI to do properly. Management looks at each failure as if "just one more fix" is all they need.
I went to a conference focused on self driving cars back in 2019 - even back then, the best talk was from some professor in Florid, who's main point was that you probably don't have an alligator crossing the road in your training data. The point being, you really, really, actually do need nearly an AGI to drive safely. The entire world is out there!
That’s not how I would put it. I don’t feel unsafe with it. It doesn’t feel like a new driver at all, I trust it to do its job. I haven’t used FSD but from what I’ve heard it sounds less trustworthy to me (mostly because it tries to do more in more situations, BC is very tightly controlled).
But its job is to stay in the lane and keep a good speed even in traffic.
When using BC your job just like less advanced lane centering + cruise control: maintain situational awareness for what’s comping up + what other cars are doing. Additionally I keep my hands on the wheel anyway to take over when needed even though BC is hands free.
I do find it a benefit. But it’s “just” a lane keep + radar CC system that’s a bit more advanced. Just like every single other system on the market you can’t take your eyes off the road. Ford enforces it. If it was still active they couldn’t have been looking away for too long.
It annoyingly hugs the right side of the lane for some reason, especially when you're activating BC. Once it goes far enough right it works well. There's a few spots on my 20 minute commute that it likes to disengage because they repainted lines and did a shitty job but that isn't strictly the car's fault.
Newer versions, 1.2 and up, are supposed to be far better. Even moving over a bit in the lane when next to a big truck. That may be 1.3.
Everyone will be getting upgraded to 1.4. We’re just waiting for the OTA update. It’s overdue, but anyone who had followed BC so far knows that’s just par for the course.
History for those unfamiliar: the Mach E was shipped before BlueCruise was ready. It took 9 months or so before there was an update available at the dealer for it.
Eventually newer cars cane with 1.2. Older cars were promised 1.2 but for various reasons it got delayed and a recall issue took priority.
Then they released 1.3 on new deliveries, and the promise became everyone gets 1.3.
That got delayed a bit and they were ahead on 1.4 so Ford decided to just move everyone to 1.4 as it was done early. But they hit problems on the rollout beta test for existing vehicles. Between that and seeming to give priority to the recent updates to work with Tesla superchargers, 1.4 missed the promised date of Q1.
It’ll come. And it’s supposed to be a very big improvement over 1.0 from those who have tested it. But Mach E owners who follow the forums know not reliable Ford software update dates are.
I never drove one of these, but I always wonder if we're already past the "way too old person behind the wheel"-point where I rather see those people in a car driven by a computer than having them in control.
It becomes quite a discussion if a really horrible ancient happens here in Germany, but since this population group loves to go voting and Germany is a car nation, it dies down pretty fast when the topic of mandatory tests for elderly people come up.
It's the same problem with those who just got the license, of course, too.
The latest version 12 FSD beta seems significantly better, to the extent that experience with previous versions is less relevant. There's no real reason to expect these systems to get worse, they'll only improve. Today's videos versus those of even just a year or two ago shows huge progress.
Not better than doing yourself today in general, but it will be in more and more situations. https://www.youtube.com/watch?v=fpoXr_z_6a4 is illustrative of the current level of ability. A good example where the self-driving compares favorably with poor human driving is at 10:15 in the video.
There are some pretty major gaps (for example, limited ability to reverse and read signs), but I wouldn't bet on them not being resolved in the next five years.
I've tried FSD a handful of times during this month's beta and, while it's way better than when I tried it for a month a while back, it's at the point where I can very much see why the NTSB/etc are concerned about people trusting it too much.
It's generally really impressive and then suddenly very wrong. As in, it navigated a four-way stop with a in the crosswalk flawlessly. And then, later, it was lining up for a left turn into a parking lot and decided that the oncoming truck wasn't relevant, so it spun the steering wheel very quickly and lurched forward until I mashed the brake.
I'm very hopeful that we can replace the mistake-prone meatbags that drive cars now, but I'm incredibly concerned with people on other forums saying that they trust it completely.
I also tried it out during this month's trial, and while it is convenient in certain scenarios, I wouldn't consider it production-ready.
If you treat it as pre-release and truly understand the risks, then it functions adequately. As a side note, I am pleased with the progress being made in this space, when I worked in that specific industry years back, it was nowhere near the functionality you have now.
It's clearly not at the point of reliability where it should be used without close supervision, but the progress is undeniable.
See 29:14 in the video I linked for an example of it behaving potentially dangerously, in what the video author describes as a regression in its behaviour and a similar situation to what you describe.
Will they only improve? Or will they get better at most things while old previously handled edge cases sneak back in without warning? There are so many situations that it needs to handle 100% correctly.
Iirc Tesla has a giant internal validation sets of edge-case-related disengagements they test against using a simulator whenever they’re gonna deploy a new version.
Even with normal cruise control and adaptive cruise control, when there is a lot of traffic, it’s easier to turn it off and take control. All these features seem great when no one else is on the road, to give the driver a bit of a rolling break, but that’s about it.
That's funny, I only ever use adaptive cruise when I'm in a lot of traffic on the highway. At slow speeds it's not going to do anything and I'm pretty much just going to follow the car in front at low speed. I use lane keeping the rest of the time and it really reduces fatigue.
My issue in traffic is that if someone gets between me and the car in front of me, the car hits the brakes to make some room. I’ve had it slow down from 50mph, from 75. It’s dangerous, if the person behind me isn’t paying attention, they could easily hit me. If I ever have it on in traffic my finger is on the switch to turn it off incase someone merges in front of me.
Of course, I use the max follow distance, so I have time to react if something happens and the car doesn’t do its job. I know some people who use the closest distance, which I guess would solve the merge issue, but would stress me out and if something happened, I’d be screwed.
Not sure what kind of car you have, but that’s the way my wife’s car works too — aggressive braking to keep the gap. My Tesla is the only car I’ve used adaptive cruise control where it doesn’t freak out and brake when you get cut off. Instead, if the car that pulled in ahead of you is traveling at a higher speed, it just lets the gap grow naturally.
Now, of course I’ve still had it brake at shadows on the road, so it’s not perfect, but sometimes it works like I’d expect.
A VW GTI was where I had it happen a lot. I don’t have it anymore. I assume my new car works the same way, but after the VW I haven’t given it the chance.
If someone merges while accelerating, the cruise control generally handled it ok. However, most of the time people merge, then speed up. During that first part, the car freaks out and slams on the brakes, and keeps on that course of action even as the car ahead speeds up. With the lead car speeding up, and mine going 50 when the new gap is correct, it ends up growing even bigger, so I then need to catch up. It makes me look like I can’t drive, lol.
And at slow speed in traffic jams, it plays accordion. When the car in front accelerates, it has a delay; when it decelerates, it has a delay… then it slams the breaks. If there’s one thing an automated system should be good at, it’s keeping a constant distance with the car in front, “constant” with the twist that it should be proportional to the speed.
I have only found one practical application for adaptive cruise control so far: when driving in tunnels where speed cameras strictly and vigorously enforce speed limits, hence 99% of the drivers behave and adaptive cruise control is safe to use.
Outside the tunnels… other drivers' erratic behaviour makes adaptive cruise not very useful or too monotonous on long stretches of state and interstate highways to the point that it lulls [me] into a sleep.
Yeah, there is a lot of variation in interpretation here. I'm farther along the spectrum still: I'm on FSD essentially 100% of the time in highway environments now because I absolutely trust the car's attentiveness and precision more than my own.
Think back on how many times you've started a lane change and aborted because there was a car there, how many times you've looked down to navigate and hit the rumble strip, how many times you've had to rely on ABS to avoid hitting the car you didn't notice stopping in front of you, etc... Not once in almost three years and 40k miles in my Model Y has it done something like this. That's not to say it's never going to, but 100%, for sure, it's better than I am.
Think back on how many times you've started a lane change and aborted because there was a car there, how many times you've looked down to navigate and hit the rumble strip, how many times you've had to rely on ABS to avoid hitting the car you didn't notice stopping in front of you, etc.
I don't want to criticize your driving but these are not common occurences for me (ie. the answer to most is 'never'), and I would be quite concerned if they were. Most are solved by maintaining awareness around you (including checking blind spots) and follow distance ahead of you.
This is BS. It takes a couple seconds to check a blindspot. Same with most other things you look at, navigation, radio, whatever. The rest of traffic doesn’t wait around while you are focused on that one thing.
I think his point was that the time to check the lane over is before you initiate a lane change, not after. About the only reason to initiate a lane change without pre-checking where you're going is if there's an emergency (like the incident we're discussing) and then it's questionable whether it's better to do the lane change anyway...
Let's just say I've been astoundingly impressed at the quality of drivers on this forum. Come on, everyone stomps on their brakes for unexpected stuff, or has to jump back into the lane when it wasn't as clear as you thought[1]. You just code it, like almost all drivers do, as "someone else's fault", so it doesn't get filed away in your brain as something automation can fix. But it is, and I'm telling you straight up that it does.
[1] Seriously, you see this happen around you hourly (or frankly more often) in urban traffic. You really think none of those drivers count?
Opinion: most of the replies to this comment are missing the bigger picture: FSD is trustworthy enough for highway WITH supervision.
I also use FSD on highways, and often on streets too. I find it’s very good when you adjust expectations. You still need to watch and be aware; but it’s absolutely less draining than driving yourself.
Excluding turns, I almost never interrupt FSD.
Turns, it’s not confident enough to go. Hoping 12 helps but I expect a few more years to fully solve. Even without turns it’s 100% worth it - I’m more attentive over the span of the drive with the car handling basics.
There are two kinds of drivers: drivers who know they're fallible, and fallible drivers who don't. Maybe you're an outlier who doesn't make these mistakes, and if so I applaud your skill. But most likely you're just like the rest of us, and have the added handicap of self-censoring your mistakes as "someone else's fault". Data filtering is a hell of a drug, as it were.
I’m someone that checks my side mirror when I make a turn just to collect data on whether my tire touches the line.
So I agree with OP. If you consider mistakes like “not checking your blind spot and aborting a lane change” as a part of normal driving, I would have to say that you are not a good driver.
You don’t get to a point where you become proud of your work without keeping track of your mistakes and improving on them.
Well said. I try to be 100% attentive (spoiler: am not perfect) and rather than losing interest in “boring” driving situations I play a game of trying to guess what other drivers will do, and being very conscious of whether I am right.
It’s mostly a way to maintain alertness, but it’s also helped discover patterns. Like, on city streets, if someone pulls out in front of you unreasonably and makes you get on the brakes, odds are good they will are only going a short distance and will turn, likely necessitating braking.
But yeah, just throwing up hands and declaring that everyone probably makes the mistake is no way to improve, at anything.
I’ve had ABS activate like twice in my entire life. Agree with the person you’re responding to: if ABS activates regularly when you’re driving then you may not be a safe driver.
Yes, I do agree. The hope is that full-self-driving systems will help people who should really not be on the road, negligent drivers, the elderly, those with impaired capacity from drugs and alcohol, etc.
I'm not a _good_ driver - the rumble strip one will happen to me about once a year - but I haven't had an "ABS prevented a crash" in well over a decade.
I love my adaptive cruise during traffic; frankly I think that's where it provides the most value. Stop-and-go is the most annoying thing. Having the car do it for me is fantastic.
This is why I consider adaptive cruise to be close enough to FSD. Steadying the wheel is the easy part, I just want the speed taken care of as that’s the stressful part for me.
That said I’ve only had rentals. I drive a 2005 Tacona that doesn’t even have cruise control.
Yeah I don't understand it either. They're features that seemed aimed directly at one customer: the guy who is already watching YouTube and texting while going down the interstate and just keeping a vague idea of where the lines and cars are.
With an LLM in there (if that even) one wonders if each software version deployed to a driveless / level [can't remember] car should literally have to pass a driving test.
I say this because in line with what you say about teenagers is that the driving test is about being safe-enough to not kill someone, not being a good driver.
> I don’t know if any of the implementations are up to the mark. FSD isn’t.
It is very clearly called "Beta" software.
You can agree or disagree if Tesla should be allowed to put that out there, but I don't think it should come as any surprise that software labelled as beta is not 100% flawless.
I am. I don’t think ADAS system should just be able to be blindly unleashed on the roads without some sort of standard testing for minimum safety/non dangerousness.
But above I was just arguing slapping a word on the end doesn’t release you from responsibility.
Traffic never gets better just by small infrastructure improvements. You are just lowering the price in time for a trip from A to B for people with cars.
When I'm using BlueCruise, I can't look at a sign beside the road without beeping warning me to keep my eyes on the road. There are few roads where BlueCruise even remains active for long without reverting to "hands-free" mode, so that if my hands aren't positioned properly on the wheel, or if my hands even rest too passively on the wheel during turns, there's an alarm.
I honestly don't know how people lose track of the road using BlueCruise. Wearing glasses with eyes painted on them, with rubber hands hooked onto the wheel?
I’ve got the same question because of the eye monitoring. The best I can think of is that they looked away at just the wrong time and it hadn’t quite timed out yet. Which means it was incredibly inopportune.
Or, they actually were looking ahead, but lost in thought or something and not paying attention.
You can’t do the “fall asleep in the front seat” thing that you could in Teslas when Autopilot first came out.
The eye tracking with blue cruise is really strict. It made me realize how much I took my eyes off the road before I had it. Even my 'glances' at the instrument panel often exceeded the warning threshold.
I am so glad that I am not involved with self driving systems (except as a squishy object on the same road).
The systems are so incredibly complicated and convoluted that I just cannot see how they could ever be assumed safe. I've worked on (relatively) large legacy hairball embedded systems where you don't have confidence in the behaviour determinism and it was deeply discomfiting even when the highest price of being wrong was broken hardware or ruining a production run and having the customer give a support engineer an earful.
I can't imagine the stress of having to deal with a much larger, more complex system that ends up in an open, public, uncontrolled environment where the price of a mistake is smearing a once-living human into paste.
I'm also intrigued by how these systems are tested. Presumably, as well as unit testing everything to death, you use some kind of integration test where you replay scenarios and check outcomes, and/or some stochastic generated scenarios. This must take a lot of time and resources.
As a shower thought, wouldn't it be interesting if you could get a virtual car model of a public-road self driving car and be able to play scenarios at it at will to allow external testing.
BlueCruise or not, the fact that our society accepts the number of highway fatalities that we do while underinvesting in obvious rail corridor and public transit opportunities is saddening.
The amount of yearly daily car crash fatalities would be like if 20 fully loaded commercial airliners crash every day. We would never just sit back and accept that in the world of commercial aviation.
Philadelphia and San Antonio are two very large cities where many daily car trips could be eliminated (this does not mean I'm telling people to sell their cars and walk everywhere) with even a moderateley conservative amount of public transit investment.
I suppose Ford could just do what Tesla does and have the car disable BlueCruise right before the crash, so the car isn't _technically_ 'self driving' when the crash happened.
So, neither that article nor you substantiate the "_technically_" part of the argument. It's obviously true that autopilot will routinely disengage in circumstances where collisions are imminent, either because the car isn't responding to the controls (e.g. in a spin) or because the driver grabbed the wheel or hit the brake, which overrides and causes a disengagement.
But you and Motor Trend are going farther, and implying that this is some kind of trick invented to reduce the number of accidents reported. Which (1) clearly isn't working, since the NTSB got reports about the accidents, (2) isn't substantiated by any reporting about the company behavior, and (3) is directly contradicted by the one place where Tesla does make public numbers about AP safety. They explicitly say it includes all accidents where AP was enable within 5 seconds of impact.
To wit: this is a conspiracy theory, and you need to stop repeating it.
I was driving home on the freeway late at night and there was a car on its roof that had been in an accident in the lane I was in. It was at the end of a long sweeping curve and I just barely saw it in time.
I was really glad I was in the car I was in and paying enough attention. I was in my fun car going too fast but was also more alert because of that.
Great, lets investigate every single fatal crash and make actual changes based on those. Let's not limit it to crashes involving systems like Bluecruise or Autopilot. If human drivers caused a fatal crash, let's make changes so that doesn't happen again as well.
I believe cars with autodrive etc often interpret surroundings by looking at objects moving relative to its surroundings. If an object is first seen by the car when stationary it is much harder for the software to filter it out from the “backdrop”
Would the pre-collision avoidance systems that are standard in a lot of vehicles now been able to prevent an accident with this kind of "peek a boo" scenario?
Anyone who knows anything about AI would only use it to drive their car as a last resort. The driving public's ignorance about AI's capabilities is astonishing.
What I meant is that I believe it’s just a rules engine. A set of predefined if-then actions humans created. I don’t think it’s a trained neural network making its own decisions.
That’s sort of my personal demarcation. Ever since the term got hot people have applied AI to everything from things for simpler things for far more complex. It’s not quite as easy to talk about as it used to be.
I know that seems like a lot of sensors, but in my mind it’s quite simple. GPS knows where it is. Using that it can look up the laser scan map that was previously created. From there it knows what the highway SHOULD look like. The cameras can see where the lanes are, same as systems have been doing for more than 10 years. It can check that those match the map and get some fine-tuned guidance that way. And radar lights at see how close it is to the car in front and if it is getting closer or not.
Just to be clear, since I’m not sure what I wrote was, the laser scans are not being done by the vehicle. They were done by Ford. For the vehicle it’s just a data source to consult that’s more accurate than a standard GPS map.
As expected, half of these comments are victim blaming and trying to validate that this isn’t BlueCruise’s fault…and any human would have done the same. At the same time, everyone keeps ignoring the fact that a car, moments before, drove around the vehicle safely.
So many Tesla haters here with vindictive mentalities
I think the Tesla hate is well deserved given that it turns out Teslas apparently disable autopilot right before a crash so the car isn't "technically" in autopilot when the crash happens.
Depending on how close the Ford was following they may not have had enough time (if driving manually) even though the driver ahead managed to avoid it. That would still make it their fault.
This is just sad for everyone involved.
I find it an intersection discussion though. I would qualify as a Tesla hater. I also own a Mach E and like BlueCruise. It’s just not that black and white.
This scenario sounds like what I refer to as a "peek-a-boo rear ending". You see it exaggerated in movies with high speed chase scenes, but its real world equivalent is a very common road going scenario and almost certainly leads to a collision if at 45+ mph and the rear car driver isn't maintaining situational awareness of traffic flow several cars ahead...a scenario that aggressive tailgaters simply don't take into account in their public behavioral risk calculus, and clearly BlueCruise isn't capable of handling either.