Tesla autopilot stops a collision [video] | Hacker News Reader

Tesla autopilot stops a collision [video]

youtube.com

275 pointsMattBearman195 comments

Once self driving cars reach a certain penetration level I wonder how long it'll take the remaining human drivers to start taking advantage of that fact by assuming most cars will automatically stop for them when they try stupid maneuvers like this one.

I commute on foot in Boston and I regularly witness pedestrians cross streets in front of traffic when they know they can get away with it, for example, when they're part of a crowd of pedestrians or the traffic is moving slowly. Every time I see that I wonder how much worse it'll be when they don't have to worry about being hit. Our sidewalks are narrow and crowded, will pedestrians just move to the streets?

Based on our geography, drivers and pedestrians I fully expect my city to pioneer the field of taking advantage of self driving car's automated accident avoidance systems.

Isn't one of the ideas of self-driving cars the idea that they'll reduce overall congestion and traffic volume? It might become pretty reasonable to step off into the street, because there won't be much traffic there to inconvenience.

Although it would definitely be a habit that'll end up killing someone.

There's no way self-driving cars will reduce traffic volume. If anything, it will increase since not only you don't need to be able to drive in order to travel in a self-driving car, the car might be completely empty and driving on its own for some sort of reason. And of course people who currently drive a car will still be driving a car (which might be self-driving or not).

There is a way: A lot of traffic these days is people looking for a parking spot. If the car leaves the area right away after dropping you off, it will be less congested there.

I doubt they'll turn major arteries into quiet streets, especially at rush hour.

Unless it becomes cheaper and more convenient to share rides in robotic vans booked on a smartphone and scheduled via a robotic taxi platforms that can match the bids (vans with spare seats going in the right direction) with asks (potential passengers requesting a cheap trip on their smartphone).

They may decrease congestion (backups) due to improved flow, but that does not necessarily mean there will be fewer automobiles, if anything, better flow/throughput likely would encourage new or more motorists to take to the road.

As a pedestrian who has to do this regularly when I have priority at a crossing: I live in hope

"Our sidewalks are narrow and crowded, will pedestrians just move to the streets?"

I hope so. If drvierless cars allow pedestrians to safely reclaim the streets that, as you point out, have marginalized them for so long, then that's fantastic.

It'll do wonders for carjackings, that's for sure.

> Once self driving cars reach a certain penetration level I wonder how long it'll take the remaining human drivers to start taking advantage of that fact by assuming most cars will automatically stop for them when they try stupid maneuvers like this one.

I don't understand this argument. If you do something abusive in front of me ... I will do what a robot will do. I will stop. Because I don't want to crash. I might open the window and yell some profanities at you ... but I can do that anyway, regardless of who triggered the stop, whether it was me or the on-board computer.

If you're arguing that the on-board computer has better reflexes and therefore people will engage in even more demented behavior ... that penetration level would have to be 99%. If it's something like 90%, then 1 out of 10 times you do something that abusive will result in a crash and you'll die.

By the time penetration reaches 99%, you essentially have all cars with automation built in and it's likely that using that automation will be mandatory. Cars that aren't automated will be illegal on public roads exactly because of this problem.

No you don't understand: if 99% of the traffic is automated and resilient to idiots, what prevents me from passing a red light? What prevents me from going the wrong way on the motorway? You'll have overstressed idiots that will get that and start driving like fools to meet their appointments.

It has already happened in some way: the overconfident TMAXes that don't think the laws of traffic apply to them.

TMAXes?

The parent is probably from Europe and refers to TMAX Scooters, http://www.ocregister.com/articles/tmax-672750-bike-motorcyc...

These are very powerful scooters, almost as fast as motorcycles but they don't require manual speed shifting, making. Usually drivers or those are reckless in cities like Paris

what prevents me from passing a red light? What prevents me from going the wrong way on the motorway?

Exactly the same as now: Common sense, self-protection, and the law.

Currently, people do all they can to avoid incidents, so if I do the things you mention, I will be safe to the extent that the driver sees me and can take evasive action. If what I do is detected too late to avoid a collision (such as crossing an active junction on a red light, or driving into the path of an oncoming car on a motorway) then it doesn't matter whether the driver is human or software if the event is detected too late for any response to be effective. Human or software, you're still reliant on brakes, traction, and structural stability. Those things don't miraculously improve when there's someone with faster reflexes at the wheel.

That's not to say it won't be safer when detection and response time is shorter, but it won't be so much safer as to reward stupidity.

It is not exactly the same as now: there is much greater certainty about what the outcome will be. The physics is the same, but the human is out of the loop.

There's been a subset of people who already behave this way for as long as I can recall, and it probably started soon after the invention of the automobile. Automation will greatly facilitate it.

You won't be able to guarantee there are no cars around you without human drivers. Also, you might get prosecuted for reckless behavior.

I don't get this whole argument. We needn't be so worried about a few others taking small advantages when the overall state will be better for us.

> You won't be able to guarantee there are no cars around you without human drivers. Also, you might get prosecuted for reckless behavior.

That is true now (and obviously more so in the first case), yet people still do it. If it wasn't being done, you could suggest that automation will not change the situation enough for people to start doing it, but as there already are people doing it, then that threshold has already been crossed, and it is implausible that it will not increase in frequency as the risk of doing it is decreased.

Not sure where you drive, but it must be a wonderful place.

Here in Atlanta, the only rule during rush hour seems to be "What gets me to my destination quickest?" (And that's with us having what I consider fairly law-abiding drivers in comparison to some cities)

>You won't be able to guarantee there are no cars around you without human drivers.

By the time we have 99% market penetration of automated cars, we'll have had V2X standards for a decade or more. You probably would have a report of every vehicle nearby, but... At that point cars could just interleave at intersections, and the lights can just switch on when a human driven car shows up.

The robotic cars have faster reflexes but they aren't magic. It is possible that the car will simply not physically be able to stop in time, imagine blind intersections with a large highway.

Also you should recognize that right now most of the time people run red lights or stuff like that they don't crash. If they did we wouldn't need red light cameras.

I wonder if there has ever been studies into automated vehicles that only move considering the amount of energy that can be dissipated at any time.

> Exactly the same as now: .... and the law.

One might assume that if the majority of car do not speed, do not make illegal maneuvers, do not drunk drive, and do not get into accidents then law enforcement will severely reduce enforcement.

> One might assume that if the majority of car do not speed, do not make illegal maneuvers, do not drunk drive, and do not get into accidents then law enforcement will severely reduce enforcement.

HA! Who will pay their salaries? They'll likely double-down on enforcement against "old school" drivers.

I figure it will happen like this. Just like red-light cameras catch people running red lights now, in the future all the cars are now moving cameras. You do something illegal like in this video, you have their license plate, and proof that they are doing something dangerous, so they should get a ticket.

You still don't get it. It's the pedestrian doing the wrong, not the driver.

He's saying pedestrians will take advantage of cars auto-braking for them.

Pedestrians are subject to traffic laws. And perhaps some self-driving cars will be equipped with paintball dispensers.

Pedestrians are often equipped with more effective methods of retaliation than paintballs.

This seems to imply a giant sense of entitlement on the part of automobile passengers. If you want to go fast get in an airplane. Urban areas are shared with others. If a car has to stop every ten yards to accommodate pedestrians, then that implies a certain maximum possible speed, which is fine. "Jaywalking" is as fake a legal concept as "intellectual property".

"Jaywalking" is not a fake concept. Streets are public property, maintained by the government, so the rules for who gets to use them and how are very legitimately in the government's purview.

If you're talking about the justification for governments making those rules in the first place, I could say the same about a rule saying pedestrians can cross in the middle of the street and cars have to yield - both are arbitrary rules, and should be judged based on their public policy outcomes.

Jaywalking law, including its ridiculous name , was pushed through by the auto lobby in a power grab for auto manufacturers.

Does that mean it's a bad public policy? For comparison, lots of public transit development is pushed through by developers who have an interest in their property values going up, but that doesn't make the results a bad thing.

The separation of traffic of different speeds has popped up in different countries and for different modes of transportation repeatedly, because the logic of the situation makes it make sense. In this case, a fast car loses a lot if its utility if it's limited by slower traffic that it's sharing the road with. The existence of car-only roads added value, and from the perspective of a society just introducing the technology, if that reduced the space available to pedestrians then so be it.

I think there's a good argument to be made that too much land has been allocated exclusively to motor transport, especially in city centers, but the existence of such dedicated thoroughfares makes total sense.

Just because areas are shared doesnt mean there are no rules.

Otherwise why pay attention to traffic lights and street lanes? Just drive wherever you want - in fact why stop for pedestrians at all? Don't pedestrians have to accommodate cars too?

They haven't always had to. Presently, they do. Eventually, in a better world, they won't. Walking hasn't changed for millennia, and it won't in the foreseeable future. Riding in automobiles, however, is about to change completely.

Projecting our typical pissed-off aggressive auto-commuter attitudes into the future is a mistake, and will come to be seen as a sort of mental illness. Well-adjusted robocar passengers will pay no attention to traffic, just as normal people today don't constantly bother air stewardesses about how long their flight is taking.

Has nothing to do with what he said.

> You still don't get it. It's the pedestrian doing the wrong, not the driver.

With cops freed up on traffic duty, they can start making examples of jaywalkers disrupting automated traffic.

You still don't get his point. You're trying to think up solutions but that has nothing to do with what he said.

Let's imagine a world in which every car has active collision avoidance systems. These do not even need to be self-driving cars; they just need to be able to reliably detect imminent collisions and override driver input to prevent them.

In that world, could one safely run a red light? Sometimes yes, but there are limits even to the highest-performing sensors and servos. An intersection with large buildings on every corner would still be dangerous--radar and lasers can't see through walls or around corners, and even with perfect reaction times and brake control, a car doing 30mph has as a minimum stopping distance that would take it into the intersection.

So, I don't know that people would think that they are invincible. And traffic signals would still be enforced by law.

Maybe a non sequitur, but maybe a future step would be putting the radar at every intersection and linking it all up to the cloud and then to cars. It would be like Google Maps traffic except super granular. If every car had data about what cars were coming up the street and generally at what speed, perhaps not only could accidents around corners be prevented, but automated cars could just go anytime it's safe without regard to traffic signs.

What a surveillance-state dystopia that sounds like. Do not want!

I would guess the 99% will report you in and road law and the courts will do the rest. Perhaps the 99% will gang up on you ;-)

make law-breaking maneuvers auto-detected by cars (even on manual mode) and auto-detection systems mandatory on ALL cars and that's it..

You'll be doing it in front of the other 99% of traffic who all have constantly in use cameras, which could be sent to relevant law enforcement with a voice command.

Enough people getting made at stupid people and getting the feature of auto-reporting people to the police :)

Making the cars more aware and having better reflexes will not step the from ploughing straight through you if you're dumb enough to jump in front of them. They would try to avoid and mitigate, just as drivers these days try to do, but they still have to obey the laws of physics.

I think the argument is: While a sane/rational person will stop, you are still risking your health. You're doing something unexpected (Like the idiot in the YT video).

The premise is that self driving cars will and can react defensively. While someone driving home after a long day in the office might not.

So - yes, you can abuse human drivers, today. But the GP seems to think that it's far more risky now and will be ~trivial~ given good enough self driving cars.

Agreed... it will be similar to putting your hand into a closing elevator door knowing it will open for you.

I'd imagine new laws will be made to restrict pedestrians from egregiously abusing the self driving car.

Would love to see how self driving cars perform on crazy Boston roads and with Boston drivers especially near hvd square, route 2,3 ;)

I wonder how the next generation of automated cars will respond to this generation? Will they dodge around an auto-car in front of them that's emergency braking? Will that cause them to emergency brake for oncoming cars in that lane? Will those cars have to dodge back into this lane? What loops and bugs and combinations of emergency responses will occur, at the speed of electronic control? Who will law consider 'to blame' in all this?

Everything will be recorded.

That would be awesome. The streets can return to belonging to pedestrians instead of to cars. We could get rid of crosswalks and crosswalk lights and just let the automated cars deal with it.

Or, in the dystopia of my nightmares, bicycles and pedestrians are forbidden from using all roads, not just highways. For their own protection, of course.

The various threats and exploitation strategies that will be used against self-driving cars are going to be very interesting (and likely ranging from inconvenient to lethal). There needs to be significant public discourse on the subject.

Sadly, I expect that the fact that general purpose computers are a major component of self-driving cars will mean further legislative restriction on the use of general purpose computers. The idyllic vision of everyone riding in self-driving cars with "tamper-proof" and "trusted" firmware / processors is seductive (and arguably easier to build) than developing systems that are resilient in the face of malicious actors. Cars will end up being computers you ride in, running code that you legally prevented from examining or modifying. I hate the thought of it, but I can't see another scenario actually playing out.

At least part of one alternate scenario is playing out, the EFF won the petition to inspect and modify software that controls cars https://supporters.eff.org/civicrm/mailing/view?reset=1&id=1...

For the next three years, and then the exemption could be revoked. When industry lobby successfully stirs up other groups with "safety" concerns I don't think it will be so easy to win another exemption.

Security is definitively going to be a concern. Tesla currently deals with it by some strong sandboxing the user interface. I've heard of breaches into the controls with other manufacturers but no internet bourne breach afaik.

I think the key is common sense good security engineering with strong verification of security models. We have had a few networked systems that can be target by malicious actors for a while now, like factory control systems; thankfully there's no big motivated enemy out there, but the risk will surely increase in the face of possibility of mass modification (overriding the manufacturer's firmware en masse), even accidental modification by the manufacturers.

Of course, if the user purposely modifies his own control systems he's going to be liable for any damage it causes.

I think you're wearing rose-colored glasses, and you're underestimating the number of attackers that will be targeting autonomous vehicle systems.

Your faith in "common sense good security engineering" saving the day flies in the face of years of IT security reality. Computer security has been discussed, both formally and informally, since there have been networked computers (have a look at the "Ware Report", prepared by the RAND corporation for the DoD in 1967), yet the general state of IT security is horrendous.

The Toyota uncontrolled acceleration bug and the subsequent software quality issues were reported by Phil Koopman make me think the auto industry isn't particularly qualified to take on security, reliability, or formal verification of their software. (Toyota might be a single example, but I find it hard to believe that software quality is generally any better across the industry. VW's software-based emissions defeat device makes me that much more suspicious of the industry as a whole.)

Attackers are going to be everyone from nation-states who want to surveil their citizenry to individuals who "game" the algorithms in other drivers' autonomous cars to allow themselves to move more quickly through traffic. I'm sure that staged accident insurance fraud, carjacking, hit-and-run, and other vehicle-related crimes will evolve in the face of the rise of autonomous vehicles.

I expect an outcry from law enforcement and "grass roots" groups to demand draconian measures to "lock down" autonomous vehicles. I expect that copyright maximalists and others who would benefit from restrictions on the use of general purpose computers will jump on the bandwagon too. I'm envisioning a future like Stallman's "Right to Read", but applied to our cars and any computers that interface with them.

I'm pretty sure that's a given that people drive more aggressively when their vehicles have safety systems ABS being the big one.

I don't know about about your city, but I would expect drivers to stop for pedestrians even if they are crossing in the middle of the road. The reason I never cross in front of a moving car is a matter of courtesy and not wanting to break the law, not out of fear of being run over.

> I commute on foot in Boston and I regularly witness pedestrians cross streets in front of traffic when they know they can get away with it, for example, when they're part of a crowd of pedestrians or the traffic is moving slowly. Every time I see that I wonder how much worse it'll be when they don't have to worry about being hit. Our sidewalks are narrow and crowded, will pedestrians just move to the streets?

Once upon a time, streets were for people. And then the automotive industry coined the term jaywalking...

Adam does ruin everything.

For anyone interested in this, 99% Invisible did a great episode on this[1].

[1] http://99percentinvisible.org/episode/episode-76-the-modern-...

Once upon a time, people shit in bowls and threw it out their windows.

Pedestrians taking roads from motor vehicles is a modern example of the tragedy of the commons.

Once upon a time I never heard this silly idea that the idea that you shouldn't blunder out into the street is a conspiracy by automakers. Then it was an article on HN and now I hear it all the time.

Separating pedestrians from vehicles just makes sense.

Streets before 1900 still had sections for horses and sections for people. And you didn't need to tell people not to stroll down the horsey sections, they were covered in horseshit.

Really? https://www.youtube.com/watch?v=9RF9wV-57dE

Doesn't look like Market Street in SF did (admittedly, 1906, not before 1900. But pretty close).

I'm not sure what you're looking at, but there's clearly kerbs with pedestrians and trams/horses/cars in the middle.

Skip to 1:46 and you can see it easily on the right hand side of the frame (too much darn traffic most of the time).

> I commute on foot in Boston

Boston is special. You know how sometimes pedestrians will cross on red and when they see a car and make eye contact with the driver they will turn their head the other way to make sure that the driver understands that they aren't looking in their direction and is forced to slow down? Yeah, Boston is the only place on the planet where I've seen actual drivers do the same thing. Unbelievable.

In Philadelphia everyone maintains eye contact with the driver. Less bassy, but also less passive aggressive I think.

> when they try stupid maneuvers like this one.

It's called natural selection. Eventually they will meet each other.

> I wonder how much w̶o̶r̶s̶e̶ better it'll be when they don't have to worry about being hit

There. Fixed it for you. (I commute by T and foot in Cambridge and Boston, by the way; and, yes, sounds flippant, but how awesome would it be for all streets to be "complete streets"?)

A self-driving car will have better reactions, but it's still governed by the laws of physics. Beyond a certain point, all the self-driving car can do is minimize the collision speed. After that, you can be sure the cameras on the self-driving car have recorded exactly what stupid maneuver someone pulled.

The automated cars will all be taking video of the incident, recording speed and direction info from radars and lidars, and transmitting all the data to some cloud server. At the cloud server, license plates will be read, driver faces identified, and the relevant info forwarded to law enforcement, insurance companies, and, of course, Youtube.

There was an article awhile back, I think linked from HN, about this. Not about expecting self-driving cars to stop, but about the fact you can take advantaee of their "drives like grandma" behavior -- cutting them off, tailgating etc

automagic cars by design have tons of cameras, at some point politicians will legislate mandatory reporting of recorded infractions .. for the children, you know

It's very likely that once self-driving cars reach a certain penetration, we'll see continuous-flow intersections where no human driver would dare try and cross.

https://www.youtube.com/watch?v=4pbAI40dK0A

The thing that strikes me in that video is that the Tesla is driving really fast considering the right lane is completely stopped.

A human driver would never be driving that fast. Taking into account familiarity with typical human behaviour, if you're in a lane that is moving well, you will expect at least some drivers from a slow lane to try to squeeze their way into the faster moving lane. Not to mention watching out for facial expressions on drivers in the opposite moving lane, hoping to take advantage of slow traffic to make left-turns into oncoming traffic. A focused human driver will be taking all these things into consideration when operating the vehicle.

I can easily imagine a scenario where a car turns into Tesla's lane and the Tesla not having enough reaction time to avoid collision.

Bottom line is, what I'm seeing in this video is a Tesla autopilot not being smart enough to regulate its speed according to conditions on the road and having to resort to some emergency braking maneuver to avoid a potentially catastrophic collision.

Well, I guess the devil's advocate argument is that what you say may only be true for a human driver.

After all, this video demonstrates the vehicle has the ability to rapidly react to a situation like that. So maybe the vehicle is driving at a safe speed for the conditions given what a computer is capable of.

You make an interesting point but I have to disagree. You can see at the very beginning of the video the traffic is still moving in the right lane and the buildup is just beginning. As far as we know the Tesla could have already begun decelerating to account for the change in condition in the right lane. Even if it's not, going at that speed in the left lane is not what I'd consider abnormal.

Additionally, what's not safe for a human driver might be safe for an automated one which knows it can react very quickly to a dangerous situation.

> A human driver would never be driving that fast.

I don't think it's clear at all that the car is driving faster than a typical human driver in this condition.

Agreed, video is very deceptive about the speeds and speed differentials of vehicles.

Have you not seen human driving? A lot of people will go that fast compared to.stopped traffic.

More seriously theres no details, but i suspect the guy just chooses that speed is right, you're supposed to keep your hands on the wheel and change speed/etc as needed by the conditions

Agree the speed differential was too much. It's dark, wet, with stationary vehicles and there's a junction (?)

It's cool to see how quickly the car stopped, but I think I'd be on edge about more neck jarring autopilot incidents in the future if I had been in the car

Better a neck jarring autopilot incident than a neck jarring collision!

True, although if the Tesla had been going at a less inappropriate velocity there would have been less need to brake so hard.

The old adage 'if you can't brake in time to prevent a collision then you are traveling too fast'.

That said, we could see an interesting era of robocars driving in ways which seem reckless but are still safe due to the superhuman reaction time.

I notice in the new Mercedes ad the line 'vehicle will not self-brake in all circumstances' which wouldn't fill me with confidence. It was often said that accidents actually increased after the introduction of ABS due to the false sense of security it gave drivers.

> A human driver would never be driving that fast.

They totally would. I see it... not infrequently. Basically any scenario where people in the right lane are there for a reason (e.g. taking an overburdened exit on a highway, turning right waiting for crossing pedestrians off of a main drag, HOV traffic vs non-HOV traffic etc.) I'll see traffic going by completely stopped traffic as fast as 60mph in the adjacent lane, and be the odd one out going by "slowly" at 40mph or so. Which looks to be about the same ballpark as in the video (which the description suggests was <45mph) and many surface streets (where you're passing parked cars.)

Not only would humans totally do it, I see it all the damn time. Faster than the Tesla is going, too. It always amazes me, their faith that no one will pull out in front of them...

According to the testimony of the driver, he was in fact focused on the right lane exactly in case someone tried to pull out.

He had no chance to see the car pulling a U-turn from the left.

Maybe I am a bad driver then, but that is exactly the speed I would have been traveling in those conditions. And that is how the average driver in my area would also be driving. The right lane is not completely stopped where the Tesla is, and it has plenty of room in front of it. I would have been aware of that car attempting to turn, but definitely not expecting it to. I have doubts as to whether I, or an average driver, would have reacted fast enough to them actually turning so suddenly. The Tesla however did react in time and not have a collision so it did what was expected of it.

Yes they would - the Tesla is not traveling at a very different speed from the car in front of it, which I assume is human-driven.

In an interview, the driver stated that he didn't see the turning car because his attention was on cars that might switch out of the right lane.

>if you're in a lane that is moving well, you will expect at least some drivers from a slow lane to try to squeeze their way into the faster moving lane.

Wait what? The car that squeezed in was from the the opposing lanes of traffic. In other words you are suggesting that the driver that almost caused the accident basically figured "oh no my lane is too slow, I guess I better go back in the opposite direction"? That makes no sense whatsoever.

This was a human driver, Tesla's don't have a real autopilot system that can literally drive the car for you. The "autopilot" feature is just to avoid collisions, so in this case it just applied the breaks when it detected the car moving in front of it.

In addition, the right lane probably had so much traffic because it was an exit. In this case no one in that lane is going to suddenly hop into the left lane (they would obviously miss their exit if they did). In that situation it's really not that ridiculous to be driving that fast.

He was driving, the autopilot stepped in to stop the crash only.

This not the effect of Autopilot, Tesla calls it 'Automatic Emergency Braking'.

Most modern cars come with similar collision avoidance systems. Volvo introduced it first(?) in 2008.

Most? Maybe on luxury brands.

Not sure many people would consider Volvo, Audi, Fiat, Ford, Honda, Hyundai, Mazda, Mitsubishi, Nissan, Peugeot, Renault, Skoda, Subaru, Toyota and Volkswagen to be 'luxury' brands.

Most of the vehicles they sell don't offer CAS, only the top end trims in certain models. Many offer only CWS (no auto braking, just bleeping).

The issue here is their use of the term "most." "Most" vehicles in 2015 is category untrue. Out of the top 10 most popular cars in the US, zero ship with CAS as a base feature, and half don't offer it at all.

In the luxury market, it is more common/popular/standard.

The luxury market is usually where most features now considered standard, originate. CAS is becoming significantly more common within the everyday market. Nissan's new Pulsar for example comes with CAS as standard.

My wife's 2015 Mazda 3 has it. It's kind of liberating to know it's there for situations like in the video

I think it is an over-statement to claim that "Most modern cars come with similar collision avoidance systems."

Many have collision WARNING systems (i.e. no auto-braking), and even then only on the highest trims or with an expensive options package ($3K+).

However collision avoidance systems do exist, and manufacturers like Subaru have made them so inexpensive that there's little excuse to not buy it (e.g. Subaru Legacy Premium W/Eyesight is $25-26K, similar to a mid Camry or Accord, and those lack CAS at that price point ($30K+ Limited trim to get CWS in other cars, let alone CAS)).

Aside from the luxury market (which Tesla is in) a lot of cars lack CAS or make it so expensive that they may have well. Subaru is the only manufacturer that makes it within the realm of your average joe (e.g. sub-35K).

That's the thing that kills me - or almost did last month... [ 1 ]

People are delirious about self-driving cars, but even the simplest automatable task of stopping when something is in front of you is not only an option, but an expensive one.

If CAS is a pricey option, how much more will full driverless cost? How prevalent will it be? How likely is it that only the top 1% will have it? How long until it trickles down into middle-class? How much longer will it take to trickle down than CAS?

[ 1 ] Another driver and myself were rear-ended hard in a senseless accident that would have been avoided if "car 3" had had CAS (or was driverless or if the driver had been paying attention, not speeding, not following too close). My last 6 weeks have been paperwork, doctor's appointments, painkillers, ice packs, phone calls and soon physical therapy appointments. All cutting into my productivity and good mood.

Well in Subaru's case the cost of CAS has dropped by 1/4th. From a 3K "Eyesight only" package on the top trim few years ago, to a 1.3K package on the mid trim this year that also gives you other features (e.g. winter driving assists, like heated mirrors, and so on).

From rumours Subaru wants to make Eyesight (CAS) standard by the 2018 model year (mid-2017 release).

This is always how it goes; everyone knows it will be mandatory eventually so manufacturers milk it in premium packages for as long as possible.

NHTSB just made backup cameras mandatory and those have been around for a long time.

Only recently Europe has made mandatory to have an emergency communication system. I barely know the details but I guess it's something like airbag worthy event -> phone whatever safety institution. It's not even high tech, and it's also not cheap for the customer. This should have been implemented long ago and as cheap as possible (I mean profit wise) .. society is so damn slow.

just look at the s-class, anything that shows up in a s-class will likely be in an accord in about 10 years

radar guided cruise control in s-class around 05, accord gets "adaptive cruise control" in 2014 and standard on 15.

Sorry, I wrote that only with experience of the European market and none of the US.

The Euro NCAP car safety assessment mandates this kind of collision avoidance for a 5-star rating. According to their website, half of the cars tested this year achieved such a rating.

(However, I think NCAP always test with all available options fitted. On the other hand, Skoda only charge £275 for this option on a £10k vehicle.)

A used Volvo is pretty affordable compared to a new Subaru 'round where I live.

Mazda has a ~$1500 tech package option on the Mazda 3 with CAS.

I'm sure Subaru makes a decent car (my brother has a generation old Legacy, which seems OK), but saying they're the only manufacturer that makes it within the realm of your average joe is a big stretch I think.

> Mazda has a ~$1500 tech package option on the Mazda 3 with CAS.

I cannot figure out how you'd order that from their site. Even grand touring (S or I) in the Mazda3 doesn't appear to offer a tech package at all.

> but saying they're the only manufacturer that makes it within the realm of your average joe is a big stretch I think.

I concede that. It is an over-statement. For around $25K there are slim pickings, but if you have up to $35K (which is common for many people buying cars), then there are several systems to consider.

I think you are still stretching. These systems available to most buyers. Considering that "Americans paid an average price of $33,560 for new cars and trucks in April 2015"[1], and that all manufacturers offer it. For instance, contrary to your statement above Accord has it available for all trims, including on LX at $25K.

[1] http://www.usatoday.com/story/money/cars/2015/05/04/new-car-...

My memory failed me. It's the $2,600 package (I had a Mazda3 S-GT earlier this year, nice car, but the sloping roof line makes the 5-door hatch frustratingly impractical for the things you want a hatchback for).

Their site sucks so I can't figure out how to copy/paste it, but from another site:

"To go all the way, you may then add the $2600 Technology package, which brings a capacitor-based brake-energy recapture system, radar cruise control, low-speed collision mitigation, automatic high beams, lane-departure warning, and forward-obstruction warning."

I bought the car for about $23K with every option but the tech-package. A quick search says there's vehicles with the option in my area for under $28K. But I think under $25K at least is probably a more realistic price.

Anyways, yeah. $25K to $35K there are a good number of choices if you look around.

From the source on TMC the user in question that posted this said he was using AP at the time. While the final braking was AEB I'd still say the title is fair.

I'm not convinced that this is an instance of autopilot "saving the day", as the title of the driver's video would suggest.

The driver states in his video description that he was "watching stopped traffic to my right". Think how you would react were you in full control of a car in that instance. You have a virtually stopped lane of traffic to your right, you see cars braking about 100 ft ahead of you. Wouldn't you start slowing down? I would, as clearly travelling as fast as the driver is where all surrounding traffic is slowing or stopped doesn't make sense, however the driver doesn't appear to slow down at all, by his own admission focused on the cars to his right. Was he legitimately distracted by cars to his right, or was he depending on the car's situational awareness? I'm leaning towards the latter, what I believe is an affordance offered by autopilot systems.

I think we'll start hearing and seeing many such "success" stories of close calls involving autopilot, but I'm not sure how many of these will be incidents of autopilot truly saving an otherwise attentive driver by reacting faster/better than the attentive driver. It seems to me that autopilot is going to be both cause and saviour in far too many cases.

Google has posted some interesting situations where their cars have predicted other driver behavior and avoided accidents in situations where it is quite possible an average human driver would not have been able to make the same prediction and avoided the accident. I agree that not every situation like this would have resulted in an accident for certain, I also strongly suspect the computers are much better at avoiding these types of accidents more consistently than humans.

The advantage of the autopilot is that its reproducible, 100% of the time. When something could have been done better given a situation, the code is fixed and the autopilot will handle the situation better every single future time.

Humans arent as reliable. And then of course theres perfect collision course distance calculations and millisecond reflexes.

I've always thought drivers of automatic transmissions seem less aware of what is happening around them, because they don't have to pay as much attention.

Since joining the 3 pedal master race, going back to an auto definitely doesn't engage me as much as manual.

I think regardless of the car, you can't count on humans to always be paying attention.

This is true.

I guess it's a trade off - the more the car does for you the less you have to think about what you're doing, but then again I see people driving in the dark without their lights on all the time.

I remember seeing this a while back: http://www.columnfivemedia.com/work-items/infographic-stick-...

It's a MINI ad but still interesting.

I have a gut feeling that people who drive manual cars are less likely to get in an accident simply because you kinda have to pay more attention, its harder to text and drive, etc. Also, in the US it seems like generally the only people that drive manual these days are people that enjoy driving and take it more seriously so I would expect less accidents due to that as well. I don't have any data to back any of that up so take it as you will.

Still, even the best driver isn't gonna react as fast.

I watched the video for the first time full-screen, as if I was driving, and tried to stomp my foot when I would have applied the brakes. I would have been in an accident.

The NHTSA is talking about requiring at least radar-controlled braking on new cars in a few years.[1] It's been working well in high-end cars, and it's not that expensive. It can't prevent all collisions, but it will almost totally prevent rear-ending the car in front in traffic. Most of the major automakers are on board with this.[2]

Earlier thinking was only to apply the brakes automatically as a collision-mitigation strategy, so that there would still be a collision, but a less severe one, after which the air bags would probably fire. But the technology got better, and now it looks reasonable to go to automatic braking sooner and prevent many collisions entirely.

This will increase freeway capacity. The biggest cause of freeway delay is accidents, most of which are rear-end collisions.

[1] http://www.detroitnews.com/story/business/autos/2015/06/08/n... [2] http://money.cnn.com/2015/09/11/autos/automatic-braking-nhts...

Here is what I want.

I want the car to avoid the collision, record the whole incident, and submit it as evidence to penalize the other driver.

The penalties could take the form of a lawsuit. Or could take the form of tagging the incident, and submitting the video where the other person's insurance company can automatically pick it up, examine the evidence, and then change rates.

That way we not only prevent accidents, but we start to address the minority of reckless drivers that put all of us at risk.

Or even better. Get people out of the car driving business altogether. We suck at it.

When will insurance companies realize the risk reductions of self driving cars and begin lowering their rates and raising the rates of manually driven cars. I think that this will be the biggest market driver for this technology. Eventually insurance rates will push people into self driving smart cars.

You may well be right, but I think it's possible that as self driving cars proliferate, insurance rates across the board may drop. It makes sense that rates would be lower for self driving vehicles than human driven, but just as each individual (no matter their vehicle) benefits from a lower incidence of accidents, they might also share the lower cost.

This is where autopilot (and self-driving car technology in general) really shines: faster-than-human reaction time. The driver didn't have enough time to even honk, let alone stop the car.

Maybe the driver should focus more on braking than honking?

He honked only after the car braked, so at that point there was no reason to focus on braking.

Actually, I've found that my first reaction in any traffic situation is to get out of it, either by maneuvering or braking. I never seem to find time to honk (or even think about it) and I'm puzzled as to how others reach for it the first thing.

I do both simultaneously. Both have saved me. There have been times where I immediately hit the brakes and horn when I saw someone doing something stupid (about to pull out into traffic, about to merge into me from another lane, etc) In some of these situations, the horn causes the other person to stop what they're doing and a collision that could not have been avoided solely by me braking, is avoided.

I have seen this with a lot of other drivers. My very first reaction is to brake. However, in all of my close calls, I have never noticed the other vehicle even attempt to slow down- they just honk their horn and continue on their way.

You're right, but if you watch videos of accidents, or talk to people that have been in, or seen a lot of traffic accidents, you'll find most people first honk, then think about braking.

I personally think it's because people have the "get out of my way" mentality, and don't want to be inconvenienced or slowed down when they know they're in the right.

My car was totaled in a near identical situation last month. And this was avoided with only automatic emergency braking... I long for the day when fully self-driving cars are the norm on roadways.

What kind of Uber driver owns a Tesla? Is the guy just doing it for fun/to meet people or is it actually feasible to make payments on a Tesla with an Uber driver's salary?

"I drive Uber in the mornings before work"

https://www.reddit.com/r/teslamotors/comments/3qkc7h/tesla_a...

This person is apparently from Washington, I've gotten the same Lyft driver that drives his Tesla. It'd be funny if it was the same guy.

From my sample size of one, college kids that invested their entire savings into Tesla very early then cashed out some to buy a Tesla are the kind of Uber/Lyft driver to own a Tesla.

It's been done: http://www.marketwatch.com/story/this-uber-driver-picks-you-...

Basic gist seems to be that the novelty factor of the Tesla helps, but making money can still be tricky. Still, not paying for gas helps.

Maybe if you live in your car.

Neat. Question: what happens if there's a car in back of the Tesla? Does does the Tesla disregard him and whether or not he has time to stop as well?

Edit: Certainly the top priority is avoiding the car in front of you, and you should slam on the brakes to do it. However, once you stop, the Tesla should release the brakes (if there's an imminent collision from the rear) to minimize the impact to the Tesla driver and the car slamming into the back of him. Would be cool if the Tesla engineers added this (if feasible).

It would be prudent to do so, as head-on collisions are far more dangerous.

I would assume so. If you are following someone, it's your responsibility to have enough leeway to stop if they stop suddenly. And it's much less dangerous to be rear ended than it is to hit something head on.

And slam into the other car? Don't you want to stop to reduce the energy in the crash? Perhaps turn a little to avoid being hit from behind? Is that your idea? Not stopping because the driver behind you isn't safely in control of the car and hitting someone in front of you might mean that someone's insurance company might find you at fault.

Fast breaking removes energy from the collision and if someone is not paying attention there going to hit you anyway.

If your going to respond the the car behind you then it's much better to simply slow down if your being tailgated. Doing anything else is counter productive.

PS: Even a long line of tailgating cars can stop because you can only use the space between you and the obstacle, but the car behind you get's that same distance plus the distance between you at them. This space keeps stacking as you go further back so by 10 cars or so it's a long slow break.

Though the "buffer" gets eaten-away with every added car that has to wait for the one in front of them before they can "react". Assuming they don't see past the car in front of them.

Your model suggests we should get thousands of car pileups if anyone does heavy breaking in bumper to bumper traffic. Clearly that does not happen.

The reason is the buffer between you and the car in front of you is eaten by the difference in your speeds. Going from 80MPH to 70 MPH takes ~0.7 seconds, where starting to break takes ~0.2 seconds. Even if there is just 5 feet between cars the signal propagates plenty fast.

What do you mean by difference in speeds? The assumption I made, simplifying the thought experiment, was that the cars had a fixed-length between them, and they were all going at the same speed.

I don't know about other countries, but where I drive, not only the front is the top priority but it is your responsibility. If a car hits you from behind, it's their fault for all intents and purposes.

That turn the other car attempted is just insane

You sort of hope they didn't even see the Tesla.

Which is still bad, but it's not a terrible decision.

People often let their attention be corrupted by focusing on the wrong thing. The stopped line of cars looks pretty safe to turn through, right? I wouldn't even be surprised if the stopped driver was waving the turning driver on to "be nice".

Despite having set cruise control below the numeric speed limit, the Tesla is still (almost) speeding by driving too fast for the conditions (passing a line of stopped cars). I'm not a big fan of this reasoning, but there's justification for it. In this situation the turning driver should have obviously seen the oncoming car and yielded, but someone taking the same turn in the opposite direction would never have been able to see the Tesla.

> I wouldn't even be surprised if the stopped driver was waving the turning driver on to "be nice".

When I got my license (12 years ago, UK) my instructor very strongly warned against waving / signalling other drivers through for this very reason. I know others around the same time were warned against it too and I believe it's also part of the Highway Code over here (don't judge me, it was a while ago!).

I still do it, though - it seems rude to be deliberately unhelpful or even obstructive even if it isn't 100% safe / legal for the other driver to go. Not to mention the fact that I'd still expect them to use their own judgement even if a stranger was waving them on.

I generally only wave people on in simple cases like two of us pulling up to a stop sign, to indicate that they can go first. For any case where they have to pay attention to other factors, I usually just stop and lift my hands up off the wheel in a sort of indication that I'm not going to make any unexpected moves.

>my instructor very strongly warned against waving / signalling other drivers through for this very reason

Over here in Germany, too.

My instructor told the the people in my class to never wave. If we were 100% sure and really think it's necessary we should make a movement with our head. That way we would be able to claim we had convulsions in front of a court, greeting the judge with a similar movement :)

Other places are different. In California specifically, I saw many people with the obvious right of way actually stopping to wave someone else on. When that person still cannot go for whatever reason, they wave harder and make a big production. The few times I was able to simply walk behind a car after it passed, they invariably had some out-of-state paraphernalia on the back.

I'm surprised there aren't more rear-end accidents from such unexpected behavior, but I guess everyone is used to the culture of hurting oneself to show piety.

As a cyclist I see this daily, and it creates some deadly situations.

People will often stop at 4-ways before you yet insist you (the cyclist) go first. This is annoying but only results in a little wasted time.

The more dangerous situation is traffic in a two lane road, one lane comes to a halt thinking that you want to cross and then frantically waves you in front of them (frantic as they're literally stopped in traffic for no reason). Meanwhile in the other lane traffic continues at high speed, blind to your crossing the road due to the stopped car.

It feels extremely rude but you have to ignore the waving person or try to signal them to move on and stop wasting everyone's time.

My uncle died from this. Walked past a stopped car, got hit in the next lane.

> waving the turning driver on to "be nice"

I've seen pedestrians nearly killed by this multiple times. Outside lane traffic stops and waves someone on with no signals in sight, the pedestrian's visibility is blocked by the car doing the waving, they go anyway because someone is telling them to, and then the inside lane traffic nearly hits the pedestrian at 50mph.

The stopped driver was obeying the law by not blocking an intersection. The turning driver ignored the second lane of traffic when making its turn.

I'd guess the stopped car on the right flashed to let them in. Often see people flash a car to pass them without checking whether it would be a safe maneuver - most commonly where a car waiting to turn left into a side road flashes you out of the sideroad, ignoring the fact that there are cars passing them to their right.

It mentions in the video that the weather was dry. Is weather a variable that the autopilot system takes in to account? For example, if it was raining would the car have slowed down to a speed where it knows it can avoid obstacles taking in to account the extra stoppage distance required?

I wonder what the effect on the insurance market will be. That could be a huge driver of adoption.

I wonder if all of this tesla-autopilot hype might increase their chances of being in a collision? At least initially.

I (regrettably) found myself leering over into a tesla driver's console the other day as he drove past to see if he was using autopilot or not and started to think of how dangerous it probably is to have everybody around you looking at you and not the road.

I guess at the very least tesla's autopilot will have some good practice early on.

I know I would definitely have been unable to prevent the accident. I didn't see the car until it was on the road.

Reminds me of the couple of close ones I had either in my car on motorcycle. On the moment you don't think, you just do but after it's a very chilly feeling of realizing that your life was just being played over a couple of seconds.

I know it's not in HN nature, but I absolutely loved the fact one of the top comments on youtube is: "FU YEA TECHNOLOGY!". It actually means that people outside tech environment appreciate course of events that is happening in the tech industry.

Isn't the autopilot feature ment for highways only at this time?

Edit: It's the emergency break feature - not the autopilot.

That just took business away from the police, court system, insurance industry, auto body repair, ...

:sarcasm: Nothing to see here just some sensors, actuators and few lines of embedded code in action. Thousands of kids do it with Lego mindstorm every days.

Indian roads, especially in towns, are where all self-driving cars will fail.)

Push the horn constantly and go to opposite lane is very common strategy, for example.

I wonder how self-driving cars would perform in such a more dynamic environment and how long it would take to adapt the parameters.

This is the basis of irony - to find the way out of Indian traffic deadlock one has to have something more than a a* search algorithm.)

Title is false. This is not autopilot.

Url changed from http://fortune.com/2015/10/29/tesla-autopilot-uber-crash/, which points to this.

Wow, very cool.

It takes two people to drive a car, you and the guy coming toward you. That is one two many.

one too many

Watch Tesla's autopilot almost causing a head-on collision: https://www.youtube.com/watch?v=MrwxEX8qOxA

I'm fairly sure you are not supposed to be using Tesla autopilot unless you are on a highway.

Then don't call it autopilot

It's in beta for a reason, and before using it, you have to accept a disclaimer clearly stating that you should not use it on road without a concrete separation between you and the opposite traffic, and you should not use it on road with sharp turns. This guy was clearly looking for the limit of the autopilot, and found it, nearly at the expense of others.

> It's in beta for a reason
There's no place on the public roads for beta software in safety-critical sytems. In fact, it's highly irresponsible.
Tesla need to verify this off-road and release a production version.

Do you think they didn't verified it off-road? Beta (in this case) doesn't mean faulty or half-assed, but you cannot possibly encounter all real-life situations on your off-road closed track, hence the name Beta to stress that point. As far as I know there wasn't any issue with the autopilot when respecting the limitations stated by Tesla. Now if a driver want to disregard those warnings, He is being highly irresponsible, not Tesla. It's like blaming an auto manufacturer because you are able to do 100mph with your car in 30mph zones.

> Do you think they didn't verified it off-road? Beta (in this case) doesn't mean faulty or half-assed
I know what beta means: it means that it hasn't shown consistent behaviour in a production environment. It means that there are edge cases which aren't understood yet. It also means that there are desired features which haven't been implemented yet. It means a lot things, of which none give a warm feeling if you're sharing the road with said software.
> As far as I know there wasn't any issue with the autopilot when respecting the limitations stated by Tesla
Ok. Do you think the environment in the video is consistent with the limitations stated by Tesla? When I look at the video, I see a car driving irresponsibly fast past a queue of slow-moving or stationary vehicles which then saved its ass by slamming on the brakes. Automatic cars which drive like that should not be on the road.

I think we cannot agree because we don't have the same definition about who should be responsible of what. This is probably because Tesla call this feature Autopilot, while it is just an Autosteer (they also use this name sometime). Autopilot gives the impression that the car should be autonomous and be able to drive itself, while this is really not the case. The car can only do basically 4 things:
-Follow the road lines
-Follow the car in front (this is what happen in this video)
-OCR on the road sign to adjust it speed
-Emergency breaking
All the rest is currently the responsibility of the driver, including reducing the speed in dangerous situations (like this one), taking the wheel in sharp turns, taking the wheel in cities, doing stops at traffic lights etc.
The Tesla is NOT (yet) an autonomous car, and if you drive it with that expectation, it can be very dangerous indeed.

The didn't call it self-driving, right? They called it something different told you to always keep your hands on the wheel.

It depends how 'autopilot' is defined. There are airplanes that have autopilot but all they can do is to keep the plane level. And there are others that can take off, land and do everything in between.

Nitpick, but there's no autopilot for takeoffs.
http://www.askthepilot.com/questionanswers/automation-myths/

Damn it, I knew this. Thanks for spotting it. =]

I remember a few months ago taking a test dive and the Tesla rep was very quick to correct me when I called it "autopilot". He kept saying it is just a safety feature. This is a very important distinction because many states laws' will have problems with autonomous vehicles. Even at the municipality level; Some cities require you have one hand on the wheel at all times. The laws aren't ready for what's coming. So until then..."Safety Features"

Tesla a calls it autosteer not auto pilot

> Tesla a calls it autosteer not auto pilot
They certainly use the term[1]: "Along with the new Autopilot feature...".
[1]: http://www.teslamotors.com/presskit/autopilot

"You're not supposed to use autopilot during takeoff sir." "Then don't call it autopilot"
What exactly about the term autopilot makes you think it must be for all aspects of the journey? Real autopilot does not work like that, so where did you get the idea that car autopilot does?

It's not exactly the best road for Autopilot.
>Autosteer works well on standard highways when there are clear lane markings or a car directly ahead to follow. It does not function reliably when a road has sharp turns. [...]
Keep in mind, Autopilot is a (public) beta, and as such, it asks you to keep your hands on the wheel at all times.

Autopilot disengaged( with a loud beep and visual ) and the driver didn't have his hands on the wheel like he is supposed to. Consequentially the car started to drift left. In addition autopilot can only be used on a highway.
Several mistakes made by driver, so the blame is on him.

Except that's not the autopilot almost causing anything, it's the autopilot disengaging that almost causes a collision. The autopilot is disengaging because this is, in fact, not a completely autonomous vehicle, and the autopilot feature is mean to be used on a highway, which that road is not, and similarly to cruise control with the human remaining alert and ready to take over, which the guy recording obviously was not.
So, good evidence that we can't go halfway with these things, and need to take humans out of the equation.

This is why all other car brands with this feature automatically disengages a few seconds after you let go of the steering wheel. And no - this does not make the whole feature useless, it's both a great convenience and works as an extra safety net to keep you on the center of the road if you put your focus on adjusting the aircon temperature for a second. The disengage in the video also comes very suddenly, the steering wheel starts steering very quickly to the left before there is even any warning sound.
Tesla should recall this feature as is, first of all calling it "auto pilot" invites people to trust it too much, plus not enforcing the by law required hand on the wheel is just asking for abuse. People are stupid and the scenario on the video could have killed both the negligent driver of the tesla and the innocent driver in the other vehicle.

You make good points. Although, disengaging when you release the wheel sounds like it could do as much harm as good. If anything, I'd rather it discourage this behavior using a nag alarm, possibly followed by automatically slowing/stopping the vehicle somewhat safely. As you say, people are stupid, and I can see some idiot trying to grab a beer out of a cooler in the back seat within the grace period (or deal with a dropped cell phone, pebble in a shoe, spider on the window, etc) and ending up careening off the road or into oncoming traffic.
I agree that this feature isn't useless. I'd love to have it, myself. As you say, it does need work to idiot-proof it, and possibly a re-branding.
But still, I would much much rather see full autonomy, as soon as possible. Wouldn't want to see us stay bogged down in the halfway stage. And I'm a little concerned that the first fatal accident caused by inappropriate use of driver assistance systems like this will unrightfully poison minds against fully autonomous systems, and take us a huge step back.

Except that you can clearly hear the autopilot disengage before the car swerves to the left. And the swerve doesn't start until it is physically impossible for a collision to happen. And the slope of the road isn't enough to make the car swerve like that. And you can't see the lower half of the steering wheel (where the driver's hands might be). And [...] ad infinitum.
Just a "Youtuber" trolling for clicks.

Reply on news.ycombinator.com