Self-driving cars

[The Ghostbuster]

What do you think? Wave of the future? Or just a pipe dream?

[kkt]

Could happen.  It can take over for daily grind driving for me anytime it wants, as long as I still get to drive the twisty country roads.

[tradephoric]

Automation is working in the airline industry.

There are about 1.1 million motor vehicle fatalities worldwide each year.  Compare that to 750 commercial airline fatalities each year.

[SignGeek101]

It'll happen... I hope, since I don't like driving. I'd say I'm the last generation who really had to learn how to drive.

Whether people will trust the self-driving cars is the thing. Some people will choose to drive, so a manual setting should be there too.

[Takumi]

I think it'll be a long time before they're adopted en masse, but I think they eventually will be. I love driving, so I don't plan on ever owning one.

[Dr Frankenstein]

Despite the hype, it's not nearly ready. Google's car still works off of a GIS database for part of its functioning (speed limits, location of stop signs and traffic signals, etc.); its computer vision system can detect lane markings and obstacles, but cannot detect a traffic signal it doesn't expect, and can't read signs. It probably can't navigate a private parking lot. Only a few cities are modeled in its database, and building such a database with pan-american coverage (let alone keeping it up to date) seems like a humongous task, in my opinion. It would require keeping constant tabs with hundreds of governments and agencies of several levels, not to mention any private entity that owns a public roadway. Oh, and I wouldn't trust a self-driving car to navigate me through some of the more complex construction zones, or to obey a flagger...

And when and if it's ready, I probably won't use it. It's great if you want to get from A to B using the "shortest" route (as determined by a GPS/GIS system; y'all know how that goes) (read: lazy). Often, on road trips, I'll decide to take a less efficient but more scenic road or clinch some routes. On many occasions, I'll see something interesting while driving and make a photo stop there. That would require constant on-the-fly reprogramming in a self-driving car, if that's even possible. I mean, how in the Hell is someone going to clinch NY 9N in a self-driving vehicle?

And I've mentioned before that some jurisdictions are likely to authorize self-driving cars on their roads, as long as there's a licensed driver ready to take over the wheel in case of failure. If it comes to that, I might as well drive the damn thing by myself, lest I won't be alert if my intervention is needed. Would you let an unreliable driver drive you around?

Automation is working in the airline industry.

There are about 1.1 million motor vehicle fatalities worldwide each year.  Compare that to 750 commercial airline fatalities each year.

Sure, but airspace near airports is tightly controlled, traffic control devices follow a worldwide standard set by the ICAO (whereas a state can't even follow its own MUTCD, not to mention counties and towns), pilots are much more compliant than drivers and traffic is much less dense. The environment is much more predictable.

[The Nature Boy]

Self-driving cars in rural areas seem like a nice sci-fi fantasy too. You think Google is going to be able to accurately map some dirt road in the middle of nowhere? Or even some paved country road in the middle of nowhere? And accurately follow any changes that happen to said road?

[Henry]

I remember seeing this on The Jetsons...oh wait, those must be the flying cars!

At any rate, I think it would be great to have self-driving cars, which would reduce the need to go to a rest stop on a cross-country trip. I'm sure they will come in handy when I'm way past retirement age.

[Brandon]

Automation is working in the airline industry.

Autopilot is more like cruise control.  The licensed pilot is still required to be there, awake and alert.

[jeffandnicole]

It'll happen. 

Hell, the airline industry went from non-existent in the early 1900's to airports small and large all over the country within a few decades.  Imagine sitting around in 1905 if someone asked you if you would be willing to fly 30,000 feet in the air, or would you rather just stay on the ground in your horse and buggy?  Technologies evolve, and no doubt the self-driving car will evolve relatively quickly at this point.

[formulanone]

No question it will happen, but give it twenty years for mass appeal, reasonable prices, public fear to die down, and infrastructure to be ready for the ever-changing transformation in our "daily driver". The manufacturing industry prefers incremental change, but public demand has to be high for it to be anything more than a concept. I also get the idea that manufacturers would prefer to offer driverless vehicles to forms of mass transit first; the customer base is less fickle, I'd imagine.

I wonder what sort pride people will take in them; as much as I enjoy a noisy, clattering vehicle, with its associated noises, switches, levers, knobs, and pedals with a bit of glee, there's still something neat about seemless technology such as this. But it won't really feel the same if there's not much say in the process. I also fear that eventually human-operated vehicles will subject to more rules and standards, so as not to upset the computer-driven ones.

If only it could spare me stupid arguments for the next fifty years about which one is obviously better, and how the opposing side clearly has its head messed up about their ideas.

[Bruce]

What I'm worried about is the self-driving car advocates purposefully using the new technology to delay/cancel good infrastructure improvements in favor of those that would benefit self-driving cars when they're released to the public "in the near future". It's already starting to happen, with some politicians in San Francisco claiming that self-driving cars would make their new transit projects obsolete.

[triplemultiplex]

Self driving cars are inevitable.
The question will be how widely they end up being implemented.

I envision a road system where you have option to let the car do the driving, or put in manual for fun or evading the horde of zombies suddenly taking over the city.  Put it on automatic for the daily commute, but take over whenever you feel like taking the scenic route instead.
Some facilities will require you to leave the computer in charge.  Urban freeways will eventually become the exclusive domain of computer-operated vehicles because they will benefit the most from this technology.
No one's going to bother making self driving cars that can drive down every little dirt track except for fun.

Currently, the number of people killed in traffic collisions is insane. Almost every single crash is caused by human error.  40-some thousand deaths; ten times that in injuries every. single. year.  If anything else was killing 40,000 Americans every year, everyone would be falling over each other to throw money and technology at the problem.  Remove the humans from the driving and you remove their errors that cause crashes.

I can't friggin' wait for cars to start driving themselves before one of you maniacs kills me.
(that's the royal "you" in that sentence, so don't nobody take offense thinking I'm criticizing anyone's driving)

[vdeane]

Better idea to reduce crashes: implement stricter requirements to get a licence.  Right now we give them out like they're candy.  If I had my way, the road test would cover just about everything and be excruciatingly difficult to pass.  Driving in any kind of snow would require an additional test to be held on I-81 south of Watertown in the middle of a blizzard.

[empirestate]

I think they're inevitable, to a greater or lesser extent. And to be honest, I don't mind that at much as I thought I would; I feel that today's automobiles already leave their drivers far too disconnected from the act of actually operating their machine, to the extent that they don't properly devote themselves to the task (and then that's compounded by an apparent general decline in people interacting in accordance with their actual surroundings, in any part of life). In other words, a lot of people get into their car with the mindset that they're just gonna kind of chill out for a while as they go from one place to another, rather than engaging themselves in the process of moving and navigating from one place to another.

So yeah, give me self-driving cars, or else let's go back to actual machines that you actively have to operate, so we can make sure everyone's mind is really in the game.

[J N Winkler]

Better idea to reduce crashes: implement stricter requirements to get a licence.  Right now we give them out like they're candy.  If I had my way, the road test would cover just about everything and be excruciatingly difficult to pass.  Driving in any kind of snow would require an additional test to be held on I-81 south of Watertown in the middle of a blizzard.

I appreciate this is tongue-in-cheek, but the idea of requiring much stricter driving tests lost a lot of the appeal it might otherwise have had for me when I failed such a test.

It is pretty easy for a road enthusiast (driver's license comfortably in wallet or purse) to pile concepts and suggested reading (MUTCD, multiple editions of the AASHTO Green Book and its predecessors, the ITE traffic engineering handbook, a few textbooks on automotive engineering, a selection of NTSB reports into major bus accidents as an alternative to a monograph on accident causation, . . .) into the syllabus for a driver's education course until the result is more "PhD in motoring" than a basic introduction to driving.  At some point people have to be let out onto the road to do their thing, and probably only a small minority will engage with driving as an intellectual project (though this percentage will likely be higher for commercial drivers).

It has been my experience, and that of my friends who have gotten their licenses to drive in jurisdictions with tough tests, that the substance of the test often reduces to working through a checklist.  In South Africa, for example, failing to check your rearview mirror at least once every 20 seconds is an instant fail.

In Britain, each driver testing center has a set of testing loops on public streets, and any mistake you make is categorized as a minor fault, a major fault, or a dangerous fault, and test pass standard is defined as so many of each type of fault.  When I took my test at the Marston Road driver testing center in Oxford, the standard had recently changed from unlimited minor faults, 5 major faults, and 0 dangerous faults to 15 minor faults, 0 major faults, and 0 dangerous faults.  Often you can be faulted for actions that are misinterpreted as errors.  For example, on my test I asked the examiner to confirm which way I should turn at an upcoming intersection with a circular blue go-left sign, and I was faulted for this because my question presupposed that the intersection in question was a roundabout when it was actually a T junction with a dual carriageway and the go-left sign was in the central reservation.  The examiner said I should have known it was not a roundabout because there was no map-type sign, despite there being numerous roundabouts in Britain without map-type signs.

Tests of this kind often end up being gamed in ways that do not promote safe driving.  In Britain, for example, even experienced drivers are often advised to sign up for a couple of hours of driving lessons (at a cost of about £20 per hour) just to figure out how to take the driving test, and to pay to have a driving instructor ride along for the test itself (as is allowed) to ensure faults are not handed out unnecessarily.  (The principle is similar to having your accountant along when the IRS audits you.)  Driver testing centers have published pass rates and a common trick is to book a test at a center, no matter how remote from your home, that has a high pass rate--often in a rural location.  The Driver and Vehicle Licensing Agency, which administers driving tests in mainland Britain, actually closed a few driver testing centers in Somerset because they had pass rates that were too high.

Driving (as Tom Vanderbilt sketches out in his book Traffic) is one of the most cognitively complex tasks people undertake.  This is why I think any driver licensing regime should be designed with careful regard to its effects on driver formation generally.  Only high-value interventions should be undertaken, and every effort should be made to avoid tipping new drivers into unsafe habits as part of the process of preparing for a driving test.  (In California, for example, the DMV counts excessive caution at stop signs as a fault during the driving test, and I think this prods inexperienced drivers to pull out too soon instead of waiting for a gap long enough to merge into traffic without forcing other vehicles to alter speed or direction.)

More broadly, there needs to be a realization that a driving test only begins the process of becoming a good driver, and the skill of hazard perception and the ethic of being a defensive driver are more important than working through a checklist in a high-pressure test setting.

[triplemultiplex]

Better testing is all well and good but it doesn't change the basic problem of humans operating machines.  Even the expert driver is still human.  They can still get tired, drunk, angry or simply distracted.  Momentary lapses of concentration happen to everybody; no exceptions.

I think about my own driving history and I can point to several incidents where I avoided a problem by using my skill and experience.  But there are a bunch of others where I escaped 'catastrophe' through dumb luck.  Existing self-driving technology found in increasing numbers of vehicles already would have prevented the two minor collisions I was involved in as a younger man.

Another benefit of self-driving cars: no more speeding tickets.  Ever.
And those speed limits for computer-driven cars will get much higher.  We're talking about a 100 mph lane on urban freeways where no manual control will be allowed.  Vehicles can run nearly bumper to bumper as well, drastically increasing capacity and increasing fuel efficiency (whatever that fuel may be) through aerodynamics.  So much capacity on our roads is wasted with empty space between vehicles to compensate for the shortcomings of human operators.

I envision a day where I can manually drive my car a for a few blocks then as I get to the freeway ramp I just tell my car to take me to exit 123 and then kick back and play a video game or read another chapter.  The most stress-free commute you could possibly have.
The sooner this can happen, the better. And it is a very, very achievable goal.

[1995hoo]

What happens when someone hacks the system the automated cars use to communicate with each other in order to avoid accidents or to deal with traffic incidents? (For the latter, I'm envisioning the sort of thing that happened this morning: The Beltway near our exit was closed in both directions due to an electrical incident just off the highway–some sort of utility fire. I presume the idea with the automated car thing is that the car would detect the road closure and automatically divert to another route.)

Also, what happens to the millions of non-automated cars now on the road? Are people to be required to dispose of them by some certain deadline? (I highly doubt any of the three cars my wife and I own could be retrofitted, especially since they're all manual shift.)

[tradephoric]

I just saw this commercial a few days ago.

[triplemultiplex]

What happens when someone hacks the system the automated cars use ...

I think that's the equivalent of saying, "What if someone hacks the internet?"  The cars would be communicating information directly among themselves rather than relying on some sort of centralized system.  Someone attempting to insert false information into this network would be fairly easy to root out as numerous vehicles in the area of a reported problem would contradict the 'hack'.  Being decentralized makes it much more difficult to intentionally cause major problems on the road network.

And really, it would be hard for a hacker to cause more devastation on the roads than we are doing right now.  40,000 is a LOT of deaths.

Naturally, things like construction or the occasional incident would have to be accommodated, but there will be fewer and fewer of the latter as self driving vehicles come to dominate the roads.

Also, what happens to the millions of non-automated cars now on the road? Are people to be required to dispose of them by some certain deadline? (I highly doubt any of the three cars my wife and I own could be retrofitted, especially since they're all manual shift.)

There is going to be a transition.  Much longer than it could be if everyone agreed this was the way to go.
At first both manual and computer driven cars will occupy the same roads and those will only be certain roads at first; freeways & major surface arteries with limited or no driveways I image.  But as more people make the switch and the technology proves itself reliable, more and more roads will allow computer-control.  This would probably cover the majority of our driving lifetimes.  Self driving vehicles will probably move in small packs while operating among manual driven vehicles.  They may be restricted to certain lanes at first as well, but slowly be allowed in more lanes and more places as they start to represent more and more of the traffic.

As more self drivers come online, certain roadways will be closed to manual driven vehicles.  When that starts to happen, then the real cool stuff can begin as well.  That's when you will start to see the 100 mph lanes and groups of semi's moving as one big road train with specially designed aerodynamic vehicles leading the pack.  I can imagine there being pilot openings of self driver only facilities where they close a freeway to manual operation for one day and everyone can experience what it's like without having to compensate for human error.  From there it will catch on fast.

Retrofitting will probably be impractical for most people and we will instead be encouraged to 'upgrade' when we buy new vehicles over time through incentives like a lower registration fee or toll rate where applicable.  Governments might subsidize the price difference since it will end up saving us billions over the long term.  Insurance companies may drag their feet at first, but it won't take long for them to be giving out huge discounts for owning a self-driving car.  In fairly short order it will be more economical to buy a newer car that can drive itself than to stick it out with an old school machine.

Stuff like this will help us make the transition.
I'll get into my thoughts about how self driving cars will handle inclement weather in another post, because that has to be part of this discussion.

[The Nature Boy]

Do you guys foresee there being areas where self-driving cars are just impractical? Rural areas are not well mapped in some cases. You can pull up almost any rural area on Google Maps and spot some errors that would prove terrible if a self driving car were to blindly follow it. How do you compensate for it?

[empirestate]

Do you guys foresee there being areas where self-driving cars are just impractical? Rural areas are not well mapped in some cases. You can pull up almost any rural area on Google Maps and spot some errors that would prove terrible if a self driving car were to blindly follow it. How do you compensate for it?

There will be such areas for quite a long time into after the introduction of self-driving cars, and there will still be a need for manual control for the foreseeable future, perhaps indefinitely.

At the same time, there will be certain facilities where self-driving vehicles are compulsory.


iPhone

[vdeane]

Question: how does one clinch things in a self-driving car?  It's hard enough to map out clinching trips on Google Maps, especially with respect to taking a parallel local road instead of the interstate (it's easy to run out of shaping points fast) or loops (my oh my are those hard... I often just don't map them at all!).

What's the point of even having car insurance with a self-driving car?  The current purpose of car insurance is to protect other people in case you cause an accident.  Either people will stop having it (not likely... what the insurance industry wants, the insurance industry gets), people will pay lots of money for something that's useless, or it will become like health insurance and start paying for routine maintenance like oil changes and inspections.

As for the 100 mph lanes... while there's certainly potential, somehow I think the "slower is always safer" crowd will find a reason to kill it.  Plus cars would have to be modified to sustain that kind of speed.  I'm uncomfortable driving my civic faster than 80 as it is due to wear on the engine (5 speed manual; the tachometer approaches 4000 rpm at 80 mph) and tires.

Finally... road trips will be a LOT less fun in a self-driving car.  Personally, I actually like freeway driving.  It's relaxing and a great de-stressor (at least as long as the road isn't crowded with idiots).  That would go away with a self-driving car... it would be more like watching roadwaywiz videos than driving, and IMO driving is more fun.

[J N Winkler]

Question: how does one clinch things in a self-driving car?  It's hard enough to map out clinching trips on Google Maps, especially with respect to taking a parallel local road instead of the interstate (it's easy to run out of shaping points fast) or loops (my oh my are those hard... I often just don't map them at all!).

The real issue, for many, would be whether you actually have to drive yourself to count a route transit as a clinch.  The data entry aspect is fairly simple:  by the time self-driving cars are reality, the clinch mapping sites will almost certainly have been upgraded to accept and interpret GPS tracks generated by a smartphone or other device carried in the car.

What's the point of even having car insurance with a self-driving car?  The current purpose of car insurance is to protect other people in case you cause an accident.  Either people will stop having it (not likely... what the insurance industry wants, the insurance industry gets), people will pay lots of money for something that's useless, or it will become like health insurance and start paying for routine maintenance like oil changes and inspections.

If self-driving cars really lead to a massive reduction in road casualties, insurance would presumably become much cheaper.  There would still be an element of comprehensive insurance that is necessary to cover weather-related damage such as hail.  I think auto insurance would become more like health insurance only if, as part of the regulatory tradeoff for allowing self-driving cars, basic maintenance operations were required to be carried out by personnel covered by a certification scheme.  DIYers of whatever skill level have the potential to be squeezed out by a shift toward self-driving cars.

As for the 100 mph lanes... while there's certainly potential, somehow I think the "slower is always safer" crowd will find a reason to kill it.  Plus cars would have to be modified to sustain that kind of speed.  I'm uncomfortable driving my civic faster than 80 as it is due to wear on the engine (5 speed manual; the tachometer approaches 4000 rpm at 80 mph) and tires.

There are a few points here:

*  100 MPH is unlikely to be allowed on any but a few select facilities, since our freeways are simply not designed that well (3° maximum curvature for a 70 MPH design speed is a bitch).  Some freeways would be easier to retrofit than others and it would take time to roll out the upgrades.  In that time, a great proportion of the vehicles that can't sustain speeds that high would gradually fall victim to fleet turnover.

*  "Speed limits too fast for the car" is a problem that already exists to some degree.  For example, I never took advantage of 80 MPH in WY, UT, and ID on my travels last September, except for short lengths on I-80 in WY and I-15 in UT.  My 1994 Saturn can handle 80 MPH just fine, but the penalty is greatly increased fuel and oil consumption.

*  Driving a Civic at 4,000 RPM for extended periods of time won't noticeably accelerate engine wear, as long as oil cooling is adequate.  The real issue with high-RPM operation is not wear per se so much as it is oil sludging due to high temperatures.  Once the oil control rings become clogged with coked oil, the engine will burn oil at a progressively greater rate.  I wouldn't consider using anything but a full synthetic oil for sustained high-RPM operation.

Finally... road trips will be a LOT less fun in a self-driving car.  Personally, I actually like freeway driving.  It's relaxing and a great de-stressor (at least as long as the road isn't crowded with idiots).  That would go away with a self-driving car... it would be more like watching roadwaywiz videos than driving, and IMO driving is more fun.

As time goes on, more and more state DOTs will put photologs or videologs online, and I anticipate that they will increasingly be used by roadgeeks to evaluate potential roadtripping opportunities for fun factor.  Moreover, lightly travelled facilities (mostly two-lane highways, but including some expressways and freeways) will be the last to be equipped with any enhancement related to self-driving cars that requires an infrastructural element, such as a ban on manual operation.

[J N Winkler]

Do you guys foresee there being areas where self-driving cars are just impractical? Rural areas are not well mapped in some cases. You can pull up almost any rural area on Google Maps and spot some errors that would prove terrible if a self driving car were to blindly follow it. How do you compensate for it?

This issue used to be quite serious in the case of Death Valley, where the original GPS-based mapping was very lazily developed (presumably by using software to "guess" which features are roads by comparing changes in albedo in satellite photos), and included roads that had been abandoned decades previously and were not traversable even by high-clearance four-wheel-drive vehicles.  After a number of heat- and dehydration-related deaths indirectly caused by bad directions from in-car GPS units, the NPS worked with the major GPS mapping vendors to remove these spurious roads.

As part of an eventual rollout of self-driving cars, there would probably be a regulatory requirement not to implement GPS-based self-driving except on the basis of mapping that has received some type of certification as to accuracy.  Aside from this, it is at least conceptually possible to design a self-driving car with logic that continually evaluates the projected route, and calculates a new routing when the road ahead is unsuitable.