A fascinating account of a feature we'd now recognize as 'cruise control'. Despite the technical details being different than modern systems, many of its behaviors are familiar -- like the surreal rollercoaster-tow hillclimbing -- while some have been rectified in more modern systems, like better engine braking to regain the target speed.
I believe the UX of vanilla cruise control is a good fit for what it does: it's a fire-and-forget, opt-in mode that's disengaged with a tap of the brake or its own button. It's simple to reason about: do I want the car to gun it at a constant 70 mph, or no? You can run a quick mental judgement call and decide whether to engage it or leave it off.
This is completely unlike any other 'assist' feature that came later: ABS, automatic emergency braking, and especially adaptive cruise control. And, in my opinion [1] this vast difference between adaptive and classic cruise control is a safety risk -- features that override driver behavior in case of a condition the car identifies need to have no false negatives, need to be on by default and opt-out instead of opt-in, and need to not be disengaged by trivial inputs like lightly tapping the brakes.
In Tesla's particular case, the opt-in UX was even worse, because its automatic emergency braking wouldn't engage if autopilot was off [3] -- a decision contrary to sane user expectation, and completely unlike cars of other premium automakers. And the fact that their emergency braking still can't detect all stationary hazards in a travel lane is tragic.
I'm not convinced that classic cruise control is a good case study (or anecdote) to compare emerging driver assist, autodriving, and safety features with. It simply has the exact opposite semantics than every other feature that we have since put in our cars.
I'll agree that the way you think of ABS should be unlike how you think of cruise control, but you shouldn't lump it in with automatic braking or adaptive cruise control. You just never need to think about ABS. It removes the need for the driver to weigh up how much or how fast to pump the breaks in an emergency situation, meaning that if you need to stop, you just slam the pedal, as it should be.
Conceptually, ABS means that the driver has a simple interface to make the car behave simply on the road. The other driver aids you mention give the drive a simple interface to make the car behave in a _complex_ way on the road, which is very different. The fact that there are smarts in the implementation is neither here nore there.
The same applies to other electronic stability packages introduced 25-20ish years ago which prevent / encourage skids depending on circumstances.
> You just never need to think about ABS. It removes the need for the driver to weigh up how much or how fast to pump the breaks in an emergency situation, meaning that if you need to stop, you just slam the pedal, as it should be.
Maybe if you learned to drive with ABS in mind.
I had to do an emergency brake on the autobahn in Germany a couple of months ago. I instinctively did not slam the brake pedal all the way, and tried to do a controlled brake from a very high speed to a full stop. I just made it and the ABS didn't engage.
Is it really recommended to let ABS take over even from very high speeds?
Yes, it is recommended. ABS pumps much faster and with more control than most, if not all drivers. Modern cars also re-balance braking between front/rear and left/right.
Anti-lock brakes are forbidden in Formula 1 in order to make it harder for the drivers. If ABS were reducing how fast a car could decelerate, that regulation wouldn't exist!
ABS can perform worse on loose surfaces like gravel or deep snow, where best braking is provided by locking the wheels. However, steering control on those surfaces while braking is better with ABS.
>Anti-lock brakes are forbidden in Formula 1 in order to make it harder for the drivers
That's actually not a 100% true. Professional drivers don't pump, they attempt to brake just at the threshold of locking in order to get the most out of their available traction. It's a fairly difficult technique that is dependent on experience more than anything, so for the average trackday racer ABS is fine enough.
If you're making an emergency stop, that is, you need to stop fifteen seconds ago, then yes, you're better off just slamming the breaks and letting the ABS do its thing. If you are not a highly trained stunt driver, I'd bet that its better at not locking the wheels then you are.
If you have more time then that, feel free to do a controlled brake and coast to a stop.
Regarding the learning to drive: I'd highly recommend taking your daily driving car to a defensive driving type course that includes a skid pan. That way you can experience what your car feels like during an emergency break, with ABS, during hydroplaning in a safe environment. This is something that could literally save your life, because you won't be surprised by how your car feels when it happens on the road.
> Regarding the learning to drive: I'd highly recommend taking your daily driving car to a defensive driving type course that includes a skid pan. That way you can experience what your car feels like during an emergency break, with ABS, during hydroplaning in a safe environment. This is something that could literally save your life, because you won't be surprised by how your car feels when it happens on the road.
This is mandatory to get a licence in some jurisdictions. Including experiencing the effect of tire wear, and having the best tires front/back.
You should slam the brakes and let the ABS take over.
In fact, modern cars with stability control packages do it for you, it is call emergency brake assist. They detect that you engaged in an emergency braking and immediately apply full power with ABS. It means that your instincts are wrong, that most drivers are like you, and that car manufacturers found that the safest thing to do is to override your instincts.
If you are lucky (or skilled) enough to max out your traction, and if you do not need to steer while braking, I’ve found avoiding the ABS will get you stopped much faster.
The ABS pulsing is designed to allow steering input while braking, which obviously means you are not devoting 100% of your traction to stopping.
A properly functioning ABS should be virtually impossible to beat with a single brake pedal, since it is modulating each wheel independently.
But that doesn’t stop me from being paranoid of “ABS drift” where modulating kicks in so hard that the car just seems to keep coasting forward on and on... you sit there with your foot hard on the pedal wondering if the firmware is just confused.
How often are you engaging emergency braking? The last time I remember activating ABS was on a dry highway onramp due to a lapse in attention while in a fairly old commodity car, and yet despite going faster than I should have been going, the car stopped in an ASTONISHINGLY short distance.
What's the alternative? Can you jiggle the brake pedal in such a way to maximize brake power without the ABS kicking in?
Another thing to note is that, especially on good grip conditions, an ABS emergency brake can be less effective than a non-ABS one. A motorcycle instructor demoed this one time, the bike without ABS stopped a lot sooner in an emergency brake situation than the ABS one. (dry, warm conditions though)
Bikes are a different scenario. If you skid in a car and have your wits about you, you can release the breaks, steer into the direction of travel, and regain control. If you skid on a bike, best case is you "low side" (http://www.360stopni.org/wp-content/uploads/2012/08/Rossi-lo...) and follow the bike on its direction of travel along the floor into whatever is in your path. Or you "high side" (https://i.ytimg.com/vi/0jU-iu7-TAY/hqdefault.jpg) and flip over the bike, hit the floor in front of it (relatively), and the bike follows you into whatever is in your path, and likely hits you when you come to a stop against whatever blocks your path.
As such, bike ABS errs on the side of caution and is much less aggressive than the four-wheel counterpart. Remaining upright and stopping more slowly is a lot safer than risking an off and letting physics, and roadside furniture, decide your fate.
This is also why you always, always wear safety gear on a motorcycle. If you've got good leather and a helmet, a "low side" skid can end up as merely a lot of bruises and stiffness (as I experienced myself once at 35mph). Without proper gear, though, it's guaranteed that you're in for a bad time.
Yes you should just slam the brakes. ABS is your best chance to get the minimum breaking distance. And you maintain control over the car.
Even without ABS, I have been trained to slam the brakes in emergencies. Without slamming you would lose too much precious stopping distance that you can’t get back.
If you spike the brakes at the wrong time on snow/ice, you will lose control of your vehicle and end up in a far worse situation than if you manually brake.
Please can you explain what you mean by manually braking? At its core, ABS releases the brakes for a wheel when it detects that wheel skidding. Braking is still manual until the locking point. Having ABS enables a driver to slam the brakes and know that won't result in a skid, where a vehicle without ABS may lock the wheels and skid.
The effect when ABS kicks in when braking on snow or ice is for the wheels to maintain slightly (or more) better traction. Sometimes this is undesirable - you'll brake in a shorter distance on loose snow if you lock the wheels - but you generally keep better steering control.
Played around with Vehicle stability Assist + ABS and it takes a lot to lose control of your vehicle. Hard breaking without a hard turn is not going to do it without lot's of ice and a curved road surface.
Not to forget the one fairly common circumstance where ABS hinders - packed snow and ice where you do need to think about it. ABS can significantly increase your stopping distance.
This is absolutely correct. If you are driving on a loose service (not just _packed_ snow) like snow or gravel, emergency braking is faster without ABS because skidding the wheels causes the material to bunch up in front of the wheels, dramatically reducing vehicle speed.
On a constantly slippery surface (ice, snow, not normally gravel) you may want to disable ABS because the ECU cannot correctly tell the vehicle speed, so may activate ABS incorrectly. However, modern systems to tend to be quite good. I believe it is best to keep systems such as traction control on though.
I learned to drive in Colorado (plenty of snow and ice in the winter), where I was a poor college student driving a typical poor college student car (old, beat-up, no modern safety features like ABS, etc).
Now that I can afford to drive a reasonably modern/safe car and live in Florida, I still can't bring myself to not instinctively pump the brakes when making an emergency stop.
A common belief is that ABS always makes braking better. That ignores the situations where it can make things worse and you should probably turn it off.
It's not simply a question of excess speed. One very common situation is where housing estate minor roads have ice or snow, but other routes are cleared or gritted. Most would probably drive in that. ABS can leave you with nearly no braking and in a far worse position than if you'd just turned it off. You might be moving at 10mph or less but slowly sail, with no control, from the minor road into a major road at a junction, or into parked cars.
Have you got a specific case for [3]? I couldn't find anything in the original "A Tragic Loss" post that would indicaate that AEB gets disabled when autopilot is off. This is the first I've heard of this behaviour and I'd like to find out more. I was thinking about getting a Model 3...
My source for that assertion is from this different, April 2016 incident [1]. Relevant quote from the article:
"In contrast, Tesla says that the vehicle logs show that its adaptive cruise control system is not to blame. Data points to Simpson hitting the brake pedal and deactivating autopilot and traffic aware cruise control, returning the car to manual control instantly. (This has been industry-wide practice for cruise control systems for many years.) Simpson's use of the brake also apparently disengaged the automatic emergency braking system, something that's been standard across Tesla's range since it rolled out firmware version 6.2 last year."
> > In Tesla's particular case, the opt-in UX was even worse, because its automatic emergency braking wouldn't engage if autopilot was off
> Simpson's use of the brake also apparently disengaged the automatic emergency braking system
I don't think the second statement necessarily implies the first. The automatic braking system may be a secondary system to the autopilot that is also disengaged when the driver is actively using the brake system. Indeed, that is how I would expect it to function, as if the driver is actively dealing with the current condition, it is probably a bad idea to wrest control away from them. For example, how do you propose a driver would expect to stop automatic braking that happens erroneously? A quick press of the brakes or gas would be my initial guess, and what I would first try. Given that an erroneously acting auto-braking system can also cause accidents, rapid disengagement capability is important.
It is one thing to disable cruise control if I've gently tapped the brake pedal to deactivate. With every cruise system I've encountered that happens well before the brakes bite. I'm fairly sure I wouldn't want emergency automatic braking off achieved by the same trigger. (No experience of Teslas to be certain).
It is entirely another another if automatic braking or other safety features deactivate when I am scrubbing speed by "slamming on the brakes". That seems like a very bad idea. If my foot slips from the brake pedal that's one common scenario where I would indeed like emergency automatic braking.
My point is mainly that most these assist systems are meant to work in the case the driver is not actively controlling that aspect of the vehicle. At any point the driver is trying to take a specific action, such as steer, accelerate, brake, etc, the car should cede control to the driver. If the driver is actively using the brakes, it makes sense to the auto-braking system to detect this and not engage. The last thing you want to do is wrench control away from someone that is actively operating a vehicle.
There may be cases where that rule could be relaxed, and the system could engage, but it would need to be beyond all doubt.
> In Tesla's particular case, the opt-in UX was even worse, because its automatic emergency braking wouldn't engage if autopilot was off [3] -- a decision contrary to sane user expectation, and completely unlike cars of other premium automakers. And the fact that their emergency braking still can't detect all stationary hazards in a travel lane is tragic.
AEB is very much running when autopilot is off, I occasionally trigger the warning when going down the parking garage(car thinks I'm heading head-on into another car).
Also after driving rush hour from Oly->Tacoma->Seattle for the last 3 years you can pry Adaptive Cruise control from my cold, dead hands. 75k mi and not a single hiccup or false positive.
The mental overhead of having to constantly evaluate vs intervene when someone merges into the lane/etc can't really be understated.
When you read this kind of text from old magazines is that you see that humanity is really getting better: "What it does. This is not easy to explain to women and the mechanically innocent. Not that the gadget is particularly complex;"
Peretend I said something politically incorrect and insert your outrage below this comment. Be sure to use some derogatory adjectives, unrelated to anything previously said, to describe me.
You don't even really need examples of casual writing to discern that this attitude is nothing new. I mean, look at stories from throughout history, from the debates of Roman senators to Salem Witch Trials and back to Chinese oracle bones-- since when have people ever avoided ad hominem attacks, fallacious logic, and overwrought outrage on topics they barely understand?
My grandmother was born in the 1920s. There was a self-enforced 'femininity' back then, wherein women told themselves that they couldn't, shouldn't, or didn't need to understand things, and that their husbands or fathers would take care of things.
I'm personally very grateful we have moved past this. My grandmother would break down at the slightest hint of technical discussion. She never understood computers as anything more than a fancy typewriter.
I think parent's point is that if we look forward 50 years even the most progressive of us might not agree that progress has been made (or think that we've regressed in some way), whereas new people will see vast progress.
Do you believe people in the 1840s were better than people in the 1770s? Or people in the 1950s better than the 1920s? I'm not convinced there is any consistent progress, as opposed to a simple preference for the times in which we live.
Inflating by CPI, the option cost $805 at today’s prices [1].
For comparison, the brochure for a model year 2016 Audi A3 reckons cruise control was a £225 and ACC a £575 option i.e. $300 and $763 at current exchange rates.
"A reversible electric motor, a flyball governor, and associated electrical gadgetry."
Making that kind of precision mechanical stuff isn't cheap. Making the "gadgetry" from discrete components isn't cheap. Making it in low volume isn't cheap. Making it keep working when splashed with wet road salt and then frozen isn't cheap.
Plus, it was a premium feature and probably was priced accordingly.
Electronic stuff is cheaper and works better. Even though a flyball governor is a cool mechanism.
It wasn't my intention to imply that it was expensive - within an order of magnitude of current prices shows how relatively cheap it was, given that electronics and sheer scale will have driven down the cost of parts (if not necessarily the end user cost).
I expect the difference is in the vendors' margins (i.e. nowadays the option price is almost pure profit/salesperson's commission).
Other than the obvious and horrifying casual sexism that others have highlighted, the most interesting part to me is the diagram at the bottom. The system was mechanically simple enough to be largely explained in one small diagram that a near-layperson (and clearly not women!) could largely understand. Such things are amiss from popular media today because our systems are so (often perhaps unnecessarily) complex.
> horrifying casual sexism that others have highlighted
Horryfying? You think that this is 'horryfying':
"This is not easy to explain to women and the mechanically innocent."
It's 1958. And yes most women (and especially at that time) tended to be less interested in mechanics. Not to say that there weren't women who were but nothing wrong in stating a point like that. Nobody cared back then either. And it wasn't horrifying that is a big exaggeration.
I watch a TV Show on Jet Airplanes. My wife has zero interest in watching that show or anything related. Ditto for my daughters and for that matter any past girlfriend. Of course there are women who watch and enjoy the same show. But I would love to bet anyone that the number of men far outweighs female interest.
Separately women make jokes about men and sports. That isn't to say that all women don't like sports or all men do. It's just a generalization that is rooted in truth and there is nothing derogatory about saying that either.
This is disingenuous. "This is not easy to explain to women..." is saying that women categorically would struggle to understand it, to the extent that they need to be mentioned as a separate category in addition to the "mechanically innocent." It's not simply observing that fewer women would be interested.
The other flaw in your argument is that casual sexism like this is why women are less interested in mechanics (or at least appear that way). If I lived in the 50s, why would I want to read a publication that casually implies I'm not smart enough to understand it? Candidly, the number of commenters willing to defend casual sexism is a big reason why I've lost interest in Hacker News.
Alright sure, let's assume for a second that you are correct about the relative lack of interest that you claim is rooted in truth. Even then, the statement in the article does not refer to interest at all, but rather it puts a constraint on ability to comprehend something that is totally unrelated to gender, and that is incredibly damaging. Imagine telling your daughters "it will be hard for you to get math because you're a girl." Or imagine someone telling you "Typing will be really hard for you because you identify as male or have a certain skin color," and consider how much damage such blanket statements can do.
Honestly, most things people do in their daily lives are much more complicated that this simple system. I mean, does your wife not use a smartphone or is she unable to figure out how to use a washing machine? The statement in the article, even if it was intended as a joke, is one of plain condescension, and thus quite horrifying indeed.
I drive most of my miles these days with ACC and lane keeping on, and I find that the article is generally still correct when relating to a newly automated function of driving.
I watch more for peripheral threats and have to pay attention to the throttle only when I feel unexpected acceleration or braking, or if there's a tricky condition that ACC might not handle correctly. I hope, as time goes on, those moments decrease in number and danger.
Same here, Fan of ACC & lane keeping. The one thing that bugs me about my car (2018 VW Tiguan Allspace) is that it'll never get "better" at ACC & lane keeping.
The factory defaults are what it's going to be, compared to how some tesla owners describe their cars to handle more situations better over time.
This is a good example of someone who can't explain a simple concept. The author uses pages and pages when a very simple explanation would make a lot more sense.
I think a lot of new technology is explained this way, and that's why some people think certain new technologies have magical properties.
Referring to cruise control as a "robot" is, in hindsight, idiotic.
> Referring to cruise control as a "robot" is, in hindsight, idiotic.
Much like 'AI' the definition of 'robot' shrinks to the pool of things which technology has not conquered. That is to say - Speech Recognition is not AI, Image Recognition is not AI, etc etc. As soon as it has its own name, we no longer call it AI or robotics.
Actually thought it was well written, except the women comment, and it was doing a good job of explaining how the technology worked. Pointing to the same limitations cruise controlle is facing today, hills feeling uneasy, too fast, or too slow as there is no context, etc.
The concept is simple to you because cruise control is such a common feature in cars that you only notice its absence. You understand exactly what a cruise control does and how to control one. Now consider being in the 1950s -- probably you had never even imagined the concept of a device that automatically regulated the throttle position of your car. And also imagine that your overall knowledge base was much less -- the average person of 1950s western civilization was exposed to much less information on a daily basis than the average person of 2018.
One of the base definitions of the term "robot" is "a machine capable of carrying out a complex sequence of actions automatically". Given that definition, I hardly think that calling a cruise control a "robot" idiotic. But what we call robots are far more complex and integrate narrow AI, so our threshold for calling something a "robot" is much higher than what would have passed for a robot in the 1950s.
In the author's defense, virtually nobody understood feedback control systems in 1958. Most people still don't, but describing it as a computer may make it easier to understand today. Also, it looks like Popular Science hasn't changed.
Regular cruise control is a feedback system, comparing the current speed to the desired speed constantly and adjusting the throttle to make the two match. A fixed throttle would result in wildly varying speed based on road grade and other factors.
> It does not invite highway hypnosis. The explanation for this lies in your past driving experience. You lend the Auto-Pilot a duty you have been trained to perform yourself, and this somehow makes you just a little uneasy. It is not an entirely relaxing thing to sit back at 70 m.p.h. and feel the engine get gas from some other agency than yourself; your alertness is more likely to be heightened than lulled. This slight wariness stays with you as long as the device is working, even after thousands of miles have been logged.
This exact same sentence sounds like it could be applied to Tesla's Autopilot (correctly or not I don't know!)
Could be, but isn't. Even with modern cruise control, you're just freed from the accelerator; there's still plenty of driving tasks to keep you alert ("what is that opposite vehicle's intent? Is there something in that shadow, threatening to enter my path? Which of the three closely consecutive exits do I take?"). With the promise of "it drives itself", you're prone to leaving the driver's seat, figuratively speaking (although there are people that have done literally that).
No - and why would it? Its only job is to keep the speed of your vehicle constant, if the visibility is too low to travel safely at the chosen speed, you should disengage the system.
I believe the UX of vanilla cruise control is a good fit for what it does: it's a fire-and-forget, opt-in mode that's disengaged with a tap of the brake or its own button. It's simple to reason about: do I want the car to gun it at a constant 70 mph, or no? You can run a quick mental judgement call and decide whether to engage it or leave it off.
This is completely unlike any other 'assist' feature that came later: ABS, automatic emergency braking, and especially adaptive cruise control. And, in my opinion [1] this vast difference between adaptive and classic cruise control is a safety risk -- features that override driver behavior in case of a condition the car identifies need to have no false negatives, need to be on by default and opt-out instead of opt-in, and need to not be disengaged by trivial inputs like lightly tapping the brakes.
In Tesla's particular case, the opt-in UX was even worse, because its automatic emergency braking wouldn't engage if autopilot was off [3] -- a decision contrary to sane user expectation, and completely unlike cars of other premium automakers. And the fact that their emergency braking still can't detect all stationary hazards in a travel lane is tragic.
I'm not convinced that classic cruise control is a good case study (or anecdote) to compare emerging driver assist, autodriving, and safety features with. It simply has the exact opposite semantics than every other feature that we have since put in our cars.
[1] https://news.ycombinator.com/item?id=16772748#16774258 [2] https://news.ycombinator.com/item?id=12097671#12097911 [3] https://news.ycombinator.com/item?id=12011584