What Will It Take To Make Automated Vehicles Legal In the US?

ashshy writes Tesla, Google, and many other companies are working on self-driving cars. When these autopilot systems become perfected and ubiquitous, the roads should be safer by orders of magnitude. So why doesn't Tesla CEO Elon Musk expect to reach that milestone until 2013 or so? Because the legal framework that supports American road rules is incredibly complex, and actually handled on a state-by-state basis. The Motley Fool explains which authorities Musk and his allies will have to convince before autopilot cars can hit the mainstream, and why the process will take another decade.

Comments

2013 or so?

[Anonymous Coward]

So Elon Musk is planning to revive the DeLorean?

Re:2013 or so?

[TheCreeep]

You hear that *whooosh*?
That's the sound of a 1.21 gigawatts joke travelling back in time going over your head.

Re:2013 or so?

[roman_mir]

I see what you did here. The time-traveling is just too dangerous. Better that you devote yourself to study the other great mystery of the universe: women!

Re:2013 or so?

[Richy_T]

Where we're going, we don't need a driver.

Re:2013 or so?

[roman_mir]

That's about as funny as a screen door on a battleship.

Re:2013 or so?

[roman_mir]

Yeah, but shark still looks fake.

Re:2013 or so?

[Maritz]

Hm, I thought that would be spelled JIGGAWATTS

For Starters

[NotQuiteReal]

How about a compelling reason?

Re:For Starters

[The New Guy 2.0]

Autodriving is a protection against accidents caused by human error.

Re:For Starters

[Anonymous Coward]

So we can all get wasted?

Re:For Starters

[graphius]

safety is the "compelling" reason, but that is a bit to ambiguous. Why should I pay more for a self driving car when I have not had an accident since I was a kid? Unless it becomes law, self driving cars* will be a gimmick. Kind of like heated seats, kind of nice, but not necessary for your average joe. And then there is a large portion of people who like driving... I think self driving cars will become more johnny cabs than personal vehicles. And maybe the Maybachs of the rich and busy. *Let's assume all/most of the bugs are worked out and self driving cars are safe.

Re:For Starters

[MozeeToby]

$60 billion dollars are spent on truck driver salary's in the US. If automated vehicles achieve a 1% improvement in fuel economy (which is ludicrously conservative) you would save the economy another $45 billion in fuel costs. Not to mention the hundreds of millions of hours of wasted time, tens of thousands of deaths, and hundreds of thousands of injuries that could be possibly be prevented or at least reduced.

Re:For Starters

[WarJolt]

Controlling a car is an NP-complete problem. We'll save lives by implementing the heuristics better and more consistently than a human mind can. There will be car error, but let's hope we can minimize that to almost zero. I doubt we will get rid of accidents entirely. Hopefully those accidents will be less deadly though.

You're not going to find a magic fix that prevents car accidents, but it's hard to prove that a car can always perform better than a human.

Re:For Starters

[ShanghaiBill]

Why should I pay more for a self driving car ...

Why do you assume they would cost more? The sensors required for a SDC should not cost much when mass produced, and the money saved on insurance will likely more than compensate for that.

Re:For Starters

[aztracker1]

Easy... get one of these automated cars driving into nascar... if it makes it through a season at least placing in the top 5, and no accidents, it's safe.

Re:For Starters

[NatasRevol]

Insurance premium reductions?

Re:For Starters

[Chas]

There will be car error, but let's hope we can minimize that to almost zero. I doubt we will get rid of accidents entirely. Hopefully those accidents will be less deadly though.

Hope. This is kinda where the argument FOR this sort of thing fall down. We *hope* it'll fix the problem. Realistically, nobody knows if it's possible, or even plausible. But hey! Let's do it anyhow!

So, essentially, you're trying to sell it as a "magic fix".

And you said it yourself. "You're not going to find a magic fix".

Re:For Starters

[K. S. Kyosuke]

Controlling a car is an NP-complete problem.

Sounds like a rather random argument to me. Surely that has never prevented people from using sub-optimal solutions to any similar problem, even before we knew that such a thing existed. Not to mention that I find it non-obvious why not hitting people should qualify as an "NP-complete problem". It's not like "NP-complete" is some magical synonym for "difficult". For example, one trivial solution would be always driving very slowly, but what has that to do with computational complexity?

Re:For Starters

[0123456]

Unless it becomes law, self driving cars* will be a gimmick. Kind of like heated seats, kind of nice, but not necessary for your average joe.

Let me guess: you've never had to start your car and drive home at forty below zero, have you?

Heated seats are, at least in part, a response to the increased fuel economy of modern engines, which don't produce enough waste heat to warm the cabin rapidly. I believe some diesel-engined cars even have electric elements in the heater for cold weather, because the coolant takes so long to warm up.

Back on topic, my guess is:

First we'll see decent 'hands free' cruise control for highway driving.
Then we'll see automated trucks driving between truck stops just outside cities, where a human driver will handle the few miles to their destination and back to the truck stop with a new load of cargo.
Then, some time a few decades from now, see actual self-driving cars that can drive anywhere without human assistance.

Re:For Starters

[meta-monkey]

But the jobs? Won't somebody think of the truck stop hookers?

Re:For Starters

[spire3661]

Removing human ego from driving is an extraordinarily compelling reason.

Re:For Starters

[0123456]

Easy... get one of these automated cars driving into nascar... if it makes it through a season at least placing in the top 5, and no accidents, it's safe.

How often does a deer run across a NASCAR track? How often are NASCAR races run in a snow storm? How often do NASCAR tracks have randomly placed ice patches?

Racing is easy for a computer to do, compared to driivng through town in a blizzard.

Re:For Starters

[sexconker]

? Or just use trains to move goods across the continental US (safer, more efficient, etc), instead of the subsidizing the trucking industry.

We already do use trains. We use them when we need to move lots of shit from A to B if A and B are connected by railways.
The problem is that we often have to ship from A, B, and C to D, E, F, G, H, etc. and the most efficient way to do so is not by rail.
If you built out rail to the point that there were enough major lines and stations for transporting shit comparably to roads, you would end up destroying the efficiency advantage.
If you do things the right way, you batch your shipments onto boats, planes, and trains, and from those endpoints you use trucks to move shit to their disparate final destinations. Guess what we do now.
If there is efficiency to be gained there is profit to be made. The shipping industry isn't jerking off over building more rail because they know it would be a boondoggle, not a boon to profits.

Re:For Starters

[spire3661]

We know it is possible and plausible. The biggest barrier to this is POLITICAL. WE could start converting HOV lanes to automated-only tomorrow. We could start establishing autonomous only roads in the next few years. Autonomous driving is a reality NOW, if we adjust the driving grid to support it properly. You start with dedicated roads and work your way from there. Waiting to solve all the outlier variables is dumb.

Re:For Starters

[menkhaura]

And how often does a NASCAR car turn right?

Re: For Starters

[bill_mcgonigle]

UPS makes millions of dollars a day by implementing approximations to an NP-hard problem. Perhaps they should just stay home and naval-gaze.

Re:For Starters

[gstoddart]

Autodriving is a protection against accidents caused by human error.

If that is the rationale, then the car needs to be 100% automated, under all circumstances, with all liability going to whoever made the damned thing.

Many of us have been saying for quite some time ... if we're liable for the actions of a robot, or the automated car is suddenly going to transfer control back to you to solve the problem ... these things will continue to be nothing more than novelties.

If you expect me to be driving in an automated car, I shouldn't reasonably need to be even awake, because any failover to human more or less needs to assume it isn't possible to do that safely.

And why the hell would I pay insurance on my car if it's not being operated by me? You think I'd take the liability over for Google? Why would I do that?

So, if it isn't 100% automated 100% of the time ... it's a half-ass solution which is going to have corner cases in which bad things will happen, and whoever made it will act like it was your fault.

Done properly, the auto insurance industry goes out of business with autonomous cars. Done improperly, there's still the illusion that the meat-sock which should essentially be a passenger is responsible.

In which case, the meat-sock might as well drive their own car.

Re: For Starters

[K. S. Kyosuke]

Again, what does that have to with the problem at hand?

Re:For Starters

[Jerry Rivers]

And how often does a NASCAR car turn right?

Fairly often given that there are 5 Nascar road courses.

Re:For Starters

[Anonymous Coward]

Every time they're at Sonoma (California) and Watkins Glen (New York), that's when they turn right.

Re:For Starters

[Stan92057]

"Autodriving is a protection against accidents caused by human error."

Horse crap!! Its about stock options, corporate growth,PROFITS

Re:For Starters

[joe_frisch]

The compelling reason is that self-driving cars could free up tens of billions of man-hours a year in the US alone. People could use the time they spend in cars for entertainment or productivity. It would be one of the truly great labor saving inventions.

For that to work though, the car would need to be truly autonomous and that gets into tricky legal issues.

Re:For Starters

[Obfuscant]

Not to mention that I find it non-obvious why not hitting people should qualify as an "NP-complete problem".

If your idea of "perfected" is "not hitting people", you've set the safe-driving bar very very low. I'd say "not hitting bridge abutments", "not hitting large animals", "not running off the edge of the road", "not hitting a tree", and even "not almost hitting a human and being saved only by fast action on the part of the human" are considerably important parts of the autonomous driving requirements.

If all you want is "not hitting people" as the answer, then simply not allowing people anywhere near the roads is an adequate solution. No autonomous vehicles necessary. Notice that most limited-access highways already prohibit hitchhikers and other pedestrians, and yet there are accidents on those highways.

Re:For Starters

[Richy_T]

OTOH, driverless trains would be an order of magnitude or two easier and... not there.

Not that I'm against driverless cars. I'd love one.

Re:For Starters

[Richy_T]

And the more likely reason will be that that's what you'll have to do if you want to travel on the roads and not use public transport once it proves safer than human drivers.

Re:For Starters

[frovingslosh]

There is no such thing as Insurance Premium Reductions.

Re:For Starters

[jbburks]

This is why the self-driving car will take so long to develop.

If I fall asleep and hit your car and injure you, you will nick my insurance, probably up to the limit of my liability. If that is exceeded, you will probably stop there, or stop after putting me into bankruptcy.

Now, if the Google(r) Self Driving Car(tm) hits you, you will sue Google for $Billions. Especially if you find out that the car hit you because of a *known* bug. That will go over very well in the hands of a non-technical jury. All software ever written (even software for the Apollo Lunar Lander) has known bugs. No company can fix every bug *immediately*. They have to prioritize, write fix, test fix, distribute fix, etc. During that time, an accident could happen. In the current 'deep pockets' liability theory, the expected cost of these lawsuits would be more than the cost of developing the software and the cost of the sensors and hardware necessary to implement it.

Even if, on average, the Google(r) Self Driving Car(tm) is safer than driving yourself, unless something like the vaccine liability fund is implemented, then the car is a non-starter (pun intended). Which is a shame because 1) I want one and 2) it's likely to be safer than the Human Driven Car (tm).

Re:For Starters

[0123456]

Uh, no. If you're using adaptive cruise control, you're still in charge of the car, and should be monitoring it at all times. Clearly some people won't, and will crash, and will be responsible.

A true self-driving car won't even have a steering wheel, so how can you possibly be held responsible for what it does?

Re:For Starters

[Ichijo]

1. Trains don't go everywhere. You still need trucks to get things from the train station to the warehouse.

When the warehouse has its own rail siding, all you need is a forklift. There's no reason why big-box stores can't be built near rail lines.

2. Who said anything about subsidizing anything?

The trucking industry is heavily subsidized.

Re:For Starters

[Defenestrar]

I enjoy driving my six speed stick on a nice windy country road, but it was less enjoyable whenever I had to drive in stop and go for an hour or two each way from work. I'd have paid for something which would let me read a book the same way I usually did while taking public transit. As adoption goes up I'd expect improved transit times as inter-vehicle communication should be able to significantly ease congestion problems. Eventually I can see non-automated cars prohibited from freeways, at least during commuting hours.

Re:For Starters

[CrimsonAvenger]

How often does a deer run across a NASCAR track?

How often does a bison scratch his head with your rear bumper? Happened to me this past week at Land Between the Lakes.

Of course, that particular piece of road is NOT a place for self-driving cars.

On the other hand, might be nice if there was some provision in the software for dealing with someone accidentally triggering self-drive mode when your car is surrounded by a herd of bison....

Re:For Starters

[FatAlb3rt]

Huh? That makes it sound like even if accident rates and fatalities are reduced by a factor of 10, it's not worth it since they're not completely gone. You guys need to not even talk about a magic fix - no one is suggesting that.

Re:For Starters

[Ichijo]

Don't worry, a train can take the load of 280 trucks off the road.

Re:For Starters

[cayenne8]

We know it is possible and plausible. The biggest barrier to this is POLITICAL. WE could start converting HOV lanes to automated-only tomorrow. We could start establishing autonomous only roads in the next few years. Autonomous driving is a reality NOW, if we adjust the driving grid to support it properly. You start with dedicated roads and work your way from there. Waiting to solve all the outlier variables is dumb.

And these are the very reasons I hope the automated car NEVER makes it in my lifetime.

I've spent a decent amount of money over the years , buying cars that are powerful and FUN to drive.

I actually enjoy driving daily....I don't want fucking mandated robots taking that fun feeling away.

If you don't want to drive....take the bus.

Re:For Starters

[Darinbob]

This is true, but I think we still have a couple of decades before we can get there. All we have today are a few cars that can follow preset directions based on highly detailed maps. These current autos aren't even going to be personal autos since they will be unable to take you into your garage or find a parking spot at the local grocery store, they're oriented towards being a shuttle service.

Best bet will be a hybrid, self driving only on freeways or major roads, human driven elsewhere. (but then some idiot will take over on the freeway because it's going too slow, get an accident, then blame it on the AI)

Re:For Starters

[spire3661]

The problem is fun is not part of SAFE driving. If you want to have fun in your vehicle, go to the track, the rest of us want to be safe from people having fun. Public roads are not built for your enjoyment, and the longer we let people like you drive, the more people will die.

Re:For Starters

[nukenerd]

If I'm driving a normal car and my brakes fail to operate properly and causes an accident, am I liable or is the car manufacturer?

If it is a design fault, the manufacturer. However, the application of brakes is vastly simpler both in concept, and their physical implementation, than an entirely automated car. Just consider that the brakes of an automated car responding to an "Apply" command is just a small sub-set of what the whole automation would be.

For one thing, the car/brake manufacturer is not involved in the decision of when to brake, and that is the hardest part to arrange.

Re:For Starters

[eightball]

I'd like to see code that can "not hit people" equally or better than human drivers, but that also manages to hit stationary and larger than human targets worse than human drivers.

Re:For Starters

[nukenerd]

The sensors required for a SDC should not cost much when mass produced, and the money saved on insurance will likely more than compensate for that.

Thanks for the laugh.

Insurance companies take any technical change whatsoever as an excuse to raise premiums.

Re:For Starters

[eightball]

Nay, a true self-driving car'll hae a name like Kitt - a true scotsman

Re:For Starters

[sootman]

Speaking of NASCAR, life-sized these would be awesome, and a great demo. A dozen races on every track with full 33-car fields without a single crash would be a good first step. Hell, run'em 'till the tires blow and see how well they cope. No humans at risk, right?

Re:For Starters

[Obfuscant]

Autodriving is a protection against accidents caused by human error.

This is called a "tautology". Of course, you mean "direct human error". Since humans will be programming these autonomous vehicles, and maintaining them, then "human error" can occur through poor/buggy/insufficient programming or failure to maintain the vehicle properly.

What "autodriving" is not a protection against is accidents caused by unanticipated interactions between different autonomous systems or failures of those systems.

Re:For Starters

[itzly]

Many airports have driverless trains.

Re:For Starters

[itzly]

So ? When I'm driving, I can only hope that other people are paying attention, watching the road instead of their smart phone. That's not any better than hoping that the computer will make the right decision.

Re:For Starters

[itzly]

It could also save time by eliminating traffic jams. If everybody punches in where they are, and where they want to go, a large scheduling system could give everybody a jam-free slot.

Re:For Starters

[nukenerd]

1. Trains don't go everywhere. You still need trucks to get things from the train station to the warehouse.

2. Who said anything about subsidizing anything?

1) I am in the UK where trains used to go within about 5 miles of just about everywhere except in the Scottish Highlands, and the railways operated local delivery trucks for the doorstep delivery of goods. Places that warehoused, produced, or consumed bulk stuff were located near to railways and had their own sidings for loading. There were vastly fewer trucks on the road, and what there were were quite small. Roads were pleasant places for people living by them, children playing in them (yes), cyclists using them, horse riders using them, and motorists using them.

The system worked well until the railways were all but crippled by poor maintenance by the ends of both WW1&2 (co-inciding with the knock-down sales of ex-army trucks to de-mobbed soldiers setting up road haulage companies), followed by nearly all but the railway main lines and main stations being closed in the 1960's.

2) Trucks are subsidised in the UK by paying vastly less in road tax than would be proportional to the road wear they cause, the distance they travel and the strength to which bridges need to be built. They are subsided by the road tax on cars, especially ones like my wife's, who only drives about 600 miles a year.

Re:For Starters

[ShanghaiBill]

Insurance companies take any technical change whatsoever as an excuse to raise premiums.

False. Most insurance companies will give you a premium reduction for installing a speed tracker in your car. You can also get a premium reduction on your home insurance by installing fire and burglar alarms.

Re:For Starters

[nukenerd]

130 million car commuters in the US spend an average of 280hrs/year driving to and from work. That's $1.2 trillion dollars per year at median salary and another $2.6 trillion dollars frozen in vehicles that sit parked for 8hrs/ day.

So what would be different with self-driving cars? Come on, give it some spin!

and another $2.6 trillion dollars frozen in vehicles that sit parked for 8hrs/ day.

Their commutes take 8 hours each way?

Re:For Starters

[DroolTwist]

I think correct terminology needs to be applied here. I'm of the opinion that any accident involving texting while driving already involves a 'driverless car'.

Re:For Starters

[squiggleslash]

Because driving fucking sucks? Is that seriously not a good enough reason?

"What did you do today dear?"

"Oh, I guided a fast moving metal box between two painted lines on some tarmac, during which time if I made any errors I was likely to KILL SOMEBODY which may at best have been ME, so I had to concentrate on that for thirty minutes. Then I did some work. Then I spent another thirty minutes of my life that I'll never get back guiding the metal box back home."

Yeah, that sure sounds like a good use of my time. Great idea, surburbanism imposing assholes, use building codes and zoning to destroy downtowns across the country and force everyone out of cities, so that homes and businesses that serve homes and, well, pretty much any destination you're likely to need to go to, needs a fucking car ride to get there.

Re:For Starters

[dgatwood]

Yeah, I've herd of bison. Why do you ask?

Re: For Starters

[dgatwood]

Perhaps they should just stay home and naval-gaze.

Yes, they should stay home and gaze at warships.

Re:For Starters

[Calydor]

IF $height>220cm THEN drive straight into it.

Re:For Starters

[Reapy]

Right on. I live in NJ and see so many shit driving accidents on my way home, it'll be awesome when computers take over and do it right all the time.

For the people that still enjoy driving manually, I could easily see 'manual drive' roads cropping up once automation takes over, roads that are built only for fun and interesting driving rather than the functional roads that end up being fun by accident of on account of the terrain it was built on.

Re:For Starters

[discomike]

I think we'll also see mandated vehicle-to-vehicle transponders even in non-autonomous cars and mandated retrofitting if you want to be able to drive your old car on "autonomous-enabled" roads.

This would make the system a lot safer, maybe you can even add control gates at on-ramps checking that your transponder is working properly before letting you on.

Re:For Starters

[0123456]

If I remember correctly, rail was also used a lot more for cargo until the rail strikes in the 90s, which convinced many companies to move to a more reliable transport method. If the railways go on strike... your cargo is stuck. If Joe's Trucking goes on strike, you call Bob's Trucking and he collects it for you, instead.

Re:For Starters

[sl149q]

Assume a solution reducing the number of people killed in traffic accidents from 30,000 per year to some other smaller arbitrary number. Say 10,000.

Those 10,000 might be killed from a different type of accident. But overall there are 20,000 less people killed.

So do you convert to that solution or stay with the current solution? This is as much a moral decision as anything else.

But be prepared for some parties to the two solutions to be trying to emphasize problems.

For example drivers unions will go to their graves (literally) claiming that the world will end if driver-less vehicles are allowed.

To the extent that car sales will drop if a driver-less taxi model takes off see all parts of the car sales and repair economy start to complain.

There will be disruption and that will generate a lot of negative feedback.

Re:For Starters

[0123456]

No-one's going to pay for a 'self-driving' car that requires them to take over at a second's notice. For it to be viable, it has to be limited to cruise control, or handle all conditions, without exception... there's no sane middle ground between them.

As we've pointed out many times before, we've seen in aviation what happens when you rely on the driver to handle cases the computer can't, when the computer does handle 99.9% of cases. It typically involves crashing into the sea and killing everyone on board.

Re:For Starters

[sl149q]

First, many people will use a driver-less taxi. That allows each "car" to service multiple people during the day. That allows for a factor of N reduction of vehicles required (where N is the average number of people serviced per day.)

Second, for people that do want to own their own driver-less car, the car can driver itself empty to a distant parking lot. And because the "driver" of each car remains with the parked car, the cars can be quadruple parked. So density of cars in the (already cheaper because it is farther away from dense down town area) parking lot can be higher.

Third, it is likely that lower cost and efficient smaller cars (think one or two passenger) will be used. Again they are more efficient and achieve a higher parking density.

Re:For Starters

[graphius]

I understand this scenario, but outside large cities it is not nearly as attractive. I think public transit also can fill a lot of this niche if it is run properly.

Some people just want easy transportation to get them to their destination. The Johnny Cab type door to door service could become very inexpensive, so there is no reason to take on the overhead of owning a vehicle. This type of service works very well in densely populated areas very well.

Other people, like me, enjoy driving and live in less congested areas. An autonomous vehicle makes less sense. Again, for my commute I can choose to use the Johnny Cab, but for my personal vehicle I want the freedom to go on more back roads etc.

Re:For Starters

[graphius]

Once I was driving when my radiator froze. (Yes I had good antifreeze, but it was beyond Damn cold) I had to drive with my window open so my windshield did not fog and freeze. Yes it was brutal, No I did not have heated seats.

My point is that there is not really a compelling reason for consumers to trade in their cars for a self driving car except in a few fringe cases. (Granted others would LIKE a self driving car for various reasons, but many people like heated seats)

Re:For Starters

[graphius]

What if the act of driving (especially on a curvy back road) IS my entertainment?

Re:For Starters

[tnk1]

I think they should probably do something anything to reduce the variables so it can get out there. Automated car lane, automated car "trains", anything to get it out there and start proving parts of the technology.

People won't be able to live without automated cars once they are out there and people see how much safer and more convenient they are. You might even allow them to exceed certain speed limits as they don't have to deal with human reaction times. Those features would make automated cars and automated-only lanes and highways a clear winner.

Just imagine driving on your local urban interstate without left lane bandits or people who are rubbernecking or who have slowed to a crawl because there is a little rain on the ground. Not to mention fewer ridiculous accidents because someone was a moron and tried something stupid. Accidents are a huge slowdown factor where I live, followed by people who can't figure out that a passing lane is for passing and not for constructing a rolling roadblock with the travel lane cars.
 

Re:For Starters

[graphius]

You are assuming that fun driving includes excessive speed, and testing the edges of traction. What about a nice pleasant day driving in the country, shifting gears, listening to the engine growl, feeling the corners.

I don't really enjoy racing, but I do like driving. No I do not want to take a cab everywhere I go*

*During one of the discussions (on Slashdot or elsewhere, I can't remember) on driverless cars someone said we already have them and they are called cabs...

Re:For Starters

[tnk1]

Isn't it already about that? Nobody is building cars for free, as far as I know.

If they can honestly demonstrate that automated cars will get into fewer accidents than human operated vehicles, then I'd say that they've earned their profits, wouldn't you?

Sure, people will still get in accidents, and perhaps there will be a glitch here and there... like Toyota accelerator pedals being stuck down... but even mechanical cars used to be rather dangerous and temperamental and people still drove them in the millions.

Re:For Starters

[tnk1]

Trying to "save" jobs by stalling progress is a good way to make sure those jobs disappear suddenly and catastrophically later.

Re:For Starters

[binarybum]

Can you clarify? This was modded as a 3 so there must be some people out there that understand what you meant. At first glance I thought you meant a compelling reason for pushing forward with self-driving cars - but that would clearly have been flame-bait.

Re:For Starters

[Dorianny]

You mean like GM's ignition switch fiasco. Killed plenty of people and they knew about it for years.Liability due to bugs or design issues is nothing new in the automotive sector.

Re:For Starters

[Chas]

The thing is, driverless trains have a more or less unobstructed right-of-way. This makes it VERY easy to avoid obstacles. And, if it DOES hit something (jackass on the crosswalk, jackass dodging the gates, etc), it weighs in at least an order of magnitude more than the object it's hitting. Meaning catastrophic damage (for its own passengers at least) is kept to a minimum.

Automobiles...don't have any such right of way. Avoiding obstacles, especially in a malfunctioning vehicle isn't necessarily straightforward, and most collisions are with other objects of roughly equivalent mass.

Re:For Starters

[Chas]

The problem is fun is not part of SAFE driving.

And, yet another person willing to bend over and not even request a reach-around in their hurry to self-bugger for the false promise of "safety".

Re:For Starters

[Chas]

I could easily see 'manual drive' roads cropping up once automation takes over, roads that are built only for fun and interesting driving rather than the functional roads that end up being fun by accident of on account of the terrain it was built on.

NEVER
GOING
TO
HAPPEN

The only way such things would be built would be with government money.
They're not going to ignore the existence of a current thoroughfare just to build "fun roads".

The very idea is asinine.

Re:For Starters

[jbburks]

No, that's not what I meant. The ignition switch fiasco is an example of a poor organizational culture getting people killed needlessly.

I'm talking about deaths that might happen during a normal development cycle of 3-6 months to identify the problem,develop, test and distribute a fix.

That is an unfortunate accident, and the injured should be fairly compensated. But not at a punitive level that the consumers of the technology would have to pay for, one way or another. The lawyers who advertise on local TV would be salivating at the chance to sue a company worth $Billions like Google in a local court.

Re:For Starters

[MattskEE]

If that is the rationale, then the car needs to be 100% automated, under all circumstances, with all liability going to whoever made the damned thing.

I don't see why that would be necessary. Effectively you are saying that insurance over the lifetime of the vehicle should be factored into the purchase price instead of allowing people to buy insurance policies. That's not a bad idea but I don't think that it should be a requirement.

Instead this can be treated like any product: you buy an insurance policy to cover damages and go after the manufacturer in cases of negligent manufacturing or design flaws.

Re:For Starters

[Euler]

Exactly. And it goes deeper than that. Who will want a driverless car that refuses to speed? Or dutifully reaches a full stop at all stop signs, railroad tracks, paper bag in the road, etc. If the designers deliberately allow shortcuts for any of these things and a single accident occurs - ever -, then Google will have the honor of paying out the largest liability judgment ever in history.

Re:For Starters

[HuguesT]

Contrary to a somewhat popular opinion among car hotheads, the least reliable component on any car is usually the driver. While on the road, drive safe. If you want to have fun with your car besides enjoying the scenery, go racing.

Re:For Starters

[cayenne8]

I don't care.

I'm here for me....my life.

And so far, my decades and decades record of driving prove that my driving "fun" is still way safer than most people on the road that are *just* driving from point A to point B....so called "safely".

What will it take?

[TechyImmigrant]

Automated vehicles that work?

Re:What will it take?

[i kan reed]

Or that any company is actually ready to sell?

I mean, the ones that exist "work" but none of the companies involved are in any hurry to put them out.

Re:What will it take?

[Iamthecheese]

As far as I know they don't work well enough: They can't yet properly handle missing lane markers, snow, puddles in the road, temporary road signs, or hand signals.

Re:What will it take?

[Obscene_CNN]

The states don't want to give up the money they make off of drunk drivers. It won't happen.

Re:What will it take?

[CastrTroy]

And even for the ones that work, they don't work well enough to make most people pay the extra amount. None of the current systems are really meant to be completely automatic and lacking a driver. This means that businesses that want automated vehicles still need to pay somebody to sit behind the wheel, which is their greatest expense. Individuals who would like to use the system to free up some time also can't benefit because they would still have to be sitting in the driver's seat, paying attention to the road. For me, I will buy an automated car when it means I can sit and watch a movie or read a book and let my car drive me to work. And even then, only when the difference in price over a non-automated car is only a small amount. Because I'm still stuck in a car, and there's only so much I can do there. If I really want to get the most out of my time, I'll reduce the amount of time spent in the car, regardless of whether or not it's automated.

Re:What will it take?

[pastafazou]

Never mind the number of people made redundant and unemployed by automation. Taxi drivers, truck drivers, bus drivers, couriers, and probably a bunch more. Expect the lobbying fees to get into the billions before any politician decides it's time to act.

Re:What will it take?

[ShanghaiBill]

SDC technology is already being released incrementally. You can already buy cars that have automatic lane control, interval control, etc. Automatic collision avoidance will be next, and then automatic braking for stop signs and traffic lights. When full SDCs arrive, people will hardly notice, because most of the technology will already be in existing cars.

Re: What will it take?

[AvitarX]

The money behind most of those things wants automation though. I would expect a lot of lobbying to make sure we do have automated drivers.

Re:What will it take?

[jellomizer]

Money.
Money to pay the Political Parties.
Money to pay for the governments.
Money to pay for the insurance companies.
Money to pay the local governments.
Money for advertising
Money for paying off special interests groups.
If there is left you can put some money into making the vehicle work better.

Re:What will it take?

[FrozenToothbrush]

The vehicles have to be bulletproof so to speak.

Plus a change in mainstream public opinion, which will only come through proven, high quality tech that does not fail even once. Many people are fearful of these things.

People will need to realze that:

• -Your right to drive is still in tact.
• -Computers are safe drivers.
• -Thousands of lives could be saved.

It's unreasonable to expect human beings to be 100% aware and capable everytime they get behind the wheel, and yet this is the assumption. If we take the time and do this right it will be of incredible benefit to humanity.

Re: What will it take?

[Roger W Moore]

True but look at driverless trains. The first of these was London's Victoria line which started in 1967 but it was forced to carry drivers, not because they were needed but because the union insisted. I expect the same will happen with driverless lorries, buses and taxis until the driverless technology is so widespread that every driver can be replaced. Until then organized labour will fight the changes even if the money is against them. This is not altogether bad though - the likely outcome is that they will slow down automated driving until they can prove it works which is probably a good thing.

Re:What will it take?

[suutar]

The problem is that the average driver will be like "yeah, everyone else needs to be replaced with a computer, but I'm fine."

Re:What will it take?

[CastrTroy]

The point is that you still want to limit the amount of time you spend in the car. If you had an automated car, you still wouldn't want to commute 2 hours each way to work even if fuel was free and the car required no maintenance, because you'd be required to spend 4 hours a day in your car. Sure you could do certain things in your car that would make it more interesting than actually driving, but you'd still be confined to your car. You couldn't use that 4 hours to go for a hike, go skiing, play boardgames with your kids, or do lots of otherwise fun things. Sure you'd be able to keep yourself busy, but you wouldn't have complete freedom of what you wanted to do.

Re: What will it take?

[TechyImmigrant]

Automated driving of a bus is probably and order of magnitude simpler that automated collection of cash from bus passengers.

Re:What will it take?

[0123456]

If they are fully automated you could save a lot of money by sharing a car with your buddy.

We, uh, could do that today. We don't, because it's so freaking inconvenient.

I mean, it's not like me and Bob both want to go to work in different places, from different places, at the same time, right?

Re:What will it take?

[gnu-sucks]

You're right, I've been thinking this for a long time. If we do it right, we will eliminate drunk driving and killed-by-sms. Wouldn't that be nice.

As far as my rights, I'm just fine giving up manually navigating LA traffic... I can't imagine anything better.

Re:What will it take?

[Euler]

Ah, but that last step is the biggest one: remove the steering wheel and take a nap on your way to work. Driver assistance is definitely a good thing: better sensors, alerts, controls. At all points in time, the driver is engaged in the current state of the vehicle as a hot failover. To achieve what self-driving cars allude to with no human fail-over is orders of magnitude above that. In fact, to make cars safer, they should demand more attention from the driver even if artificially induced. This is actually something considered in controls for aircraft, trains, and nuclear reactors: positively engage the operator to prevent complacency.

Re: What will it take?

[Roger W Moore]

Really? Most underground systems already have automated collection of cash from passengers. I've yet to see automated driving of a bus though.

Re: What will it take?

[TechyImmigrant]

Yes. But that's not a street bus. What's to stop people walking on a bus and ignoring the machine. I think that's a hard problem without a human present to enforce it.

With an underground you have machines and people employed to watch the machines. On a street bus the driver both drives and enforces payment.

Re: What will it take?

[Roger W Moore]

What's to stop people walking on a bus and ignoring the machine.

A networked camera. Have the computer send an alert when it thinks someone has not paid and an operator can review the video footage to check and then you can send someone to meet the bus at the next stop. They used to use a similar system on the toll roads around Chicago.

Alternatively use a system like they do in Geneva: the driver does not check any tickets but every so often they have ticket inspectors on a bus or tram that will. They just hop on at a random stop and check everyone's ticket and anyone without a ticket gets a large fine. The same system is used on the LRT in Edmonton, Alberta.

Typo detected

[The New Guy 2.0]

The expectation is in 2023... 2013 is an error and it's already happened.

Re:Typo detected

[Jason Levine]

The expectation is in 2023... 2013 is an error and it's already happened.

Unless the self-driving car is a Delorean. Then the first self-driving car will hit the road in 1955.

Re:Typo detected

[RavenLrD20k]

Ok...which chronology are you talking about? Ours or the car's? Because by our chronology, the first self-driving car would have been converted to be train-propelled on rails in 1885. Just because it was converted to work on rails doesn't mean that it wouldn't be trivial to convert it back...had it not been destroyed in 1985; which is still ok because it should make another appearance sometime next year.

Re:Typo detected

[Jason Levine]

That one doesn't count as self-driving car because it needed a train to push it!

Typo

[Anonymous Coward]

should be 2023 not 2013

Re:Typo

[nukenerd]

2023, not 2013

Something like this would, of course, never happen in the code controlling these cars.

Musk will have to tame

[craighansen]

both his biggest existential threat and time itself to make it happen by 2013.

Re:Musk will have to tame

[DR0P TABLE users]

Achieving this by 2013 might be possible of the next model comes with a flux capacitor...

A lot of bribes

[nine-times]

Q: What Will It Take To Make Automated Vehicles Legal In the US?

A: A lot of bribes for people at various levels of government.

Wait, did I say "bribes"? Sorry, I meant "lobbying and campaign contributions". I have a hard time telling those things apart.

What will it take?

[gnu-sucks]

Let's hope what it takes is: Really good automated vehicles.

I think this is one technology that we don't really want typical google-style beta testing (think gmail) with. Let's wait for things to mature a bit before they go mainstream.

We need a vendor we can trust here...

[The New Guy 2.0]

Car-controlling tech requires that the people involved in the car, in other cars, or owning property near a road must be able to trust the vendor offering the service. Will Google have enough resources to drive every car?

It will take a lot of real world testing.

[91degrees]

They're not goiung to be accepted until they're always better than human drivers. At the moment there are all sorts of minor issue that need to be addressed and dealt with, as well as a lot of real world issues that have yet to even be noticed.

MIT Technology review has some examples of this sort of issue.

It doesn't matter to the general public if the car is statistically safer. If a car *once* fails to spot a traffic light, even if no harm is done, this will be seen as a fundamental failure of the technology.

What will it take? Time Travel

[weilawei]

But can we hit 88 MPH first?

A working automated vehicle

[Hrrrg]

Call me when they can make an automated car that car drive in snowy conditions when no lane landmarks are visible. Or one that can turn into a lane of busy traffic that currently requires you to make eye contact with another driver to get them to slow down and let you in. Then we can worry about legalizing it. Legalization is trivial compared to the technical challenges. Personally, I suspect that there won't be a truly automated cars in my lifetime.

Re:A working automated vehicle

[astro]

Or one that can turn into a lane of busy traffic that currently requires you to make eye contact with another driver to get them to slow down and let you in.

This piece, at least, was one of the earliest-solved, easy problems. Wirelessly networked cars still having human drivers have been talked about in theory and experimented with for years - before this totally driverless thing became... a thing.

Re:A working automated vehicle

[jopsen]

Call me when they can make an automated car that car drive in snowy conditions when no lane landmarks are visible.

Should you drive in a snow storm? Either way, with GPS, etc this is not impossible, though I suspect the car would refuse to drive automatically under these circumstances... But keep in mind not all cars on the roads today can be driven under snowy conditions, try driving without a roof :) he he..

Or one that can turn into a lane of busy traffic that currently requires you to make eye contact with another driver to get them to slow down and let you in-

Is this even technically legal to do that? If you're behind a stop or yield sign, you cannot proceed forward if your interfere with ongoing traffic in any way. In practice it can be a bit different as someone should be nice and help you in; in which case an automatic car could move in too... Acting on eye contact or perceived signals like hand weaving does not hold in court and in case of an accident you would be fully liable.

Legalization is trivial compared to the technical challenges. Personally, I suspect that there won't be a truly automated cars in my lifetime.

How old are you ? :) Just kidding...

Re:A working automated vehicle

[Anonymous Coward]

A theoretical solution that requires universal adoption (wireless networking in every vehicle on the road) does not count as "solved".

Re:A working automated vehicle

[nine-times]

Call me when they can make an automated car that car drive in snowy conditions when no lane landmarks are visible.

Call me when they can make a driver from the south who can drive in snowy conditions. Hell, call me when they can make a driver in Washington D.C. drive safely in ideal conditions.

Re:A working automated vehicle

[schlachter]

It's purely a political and legal issue.

The technical challenges are nearly completely solved. What remains is infrastructure improvements, standardization, iterative improvements that will come from wide spread rollout of incremental features towards full autonomy.

Re:A working automated vehicle

[Overzeetop]

"Or one that can turn into a lane of busy traffic that currently requires you to make eye contact with another driver to get them to slow down and let you in."

Piece of cake. When 1 in 4 cars has the driverless technology, they will negotiate the entry directly. Until then, the automatic brakes on your vehicle will activate directly once the automated car in front of you nudges into the roadway, allowing them in. Laws will change to make any accident in this case the non-automated driver's fault - i.e., if you choose to drive a car without automatic braking or disable it on a car with the feature you will be liable for damage to cars entering the freeway.

Re:A working automated vehicle

[jader3rd]

Call me when they can make an automated car that car drive in snowy conditions when no lane landmarks are visible.

Would it really be that bad if people stopped traveling in super unsafe conditions?

Re:A working automated vehicle

[Euler]

People are way too willing to believe the hype; the Google driverless car is basically vaporware. Google engineers were recently interviewed for a piece on Slate.com. It was a remarkably well-written article on a technical level that describes the actual limitations of the technology. In short, it is not really autonomous at all, it follows a programmed route with some ability to detect obstacles. For example, the technology won't even observe traffic lights if they are absent from it's internal pre-programmed map. So forget about any compatibility with construction areas with temporary signals.

http://www.slate.com/articles/...

Re:A working automated vehicle

[schlachter]

Coward! I take it your reaction is from a perspective of ignorance. My assertion is not.

Re:A working automated vehicle

[schlachter]

I guess I can't convince the ignorant masses. All I can say is that I have first hand experience with this and I can tell you that the remaining issues are primarily legal and regulatory. No major technology breakthroughs are needed at this point. Just costs need to come down and some engineering needs to happen.

Re:A working automated vehicle

[jopsen]

http://en.wikipedia.org/wiki/G... Gives accuracy down to 1m... Anyways, most likely first automatic cars will allow you to drive manually too... For corner cases like that.

Re:A working automated vehicle

[zentigger]

... one that can turn into a lane of busy traffic that currently requires you to make eye contact with another driver to get them to slow down and let you in.

Well, if all the cars were automated, the oncoming traffic could be signalled much farther than eye-contact distance, and provide space automatically, such that a traffic merging pattern could be initiated and sufficient space made available for you. You also wouldn't get stuck behind the idiot that is too scared of traffic to merge unless there is at least ten times the required distance.

As a starting point, however, I suspect a gradual introduction of automated cars, say a special lane on the highway, much like a commuter lane or bus lane.

Re:A working automated vehicle

[RyoShin]

An automated car would be smarter than a person and refuse to drive in such a storm. If it is already driving when such a storm hits, it will find the nearest safe pull-off location and do so.

If it's an emergency, then:
1) you should try contacting emergency personnel who would have better-equipped vehicles to come to you; but, if they can't come, then
2) fully-automated cars would probably have some override mechanism you could activate[1], allowing you to
3) drive into the snow storm and maim/kill yourself/others, causing more emergencies. GOTO 1

[1] These would probably come with explicit warnings about shifting all liability onto you if you choose to use it.

Litigious Forbearance

[Loopy]

If anything is gonna kill/delay the automated vehicle market, it's gonna be people suing the shit out of car manufacturers when anything at all goes wrong. And make no mistake, it's gonna be up to the manufacturer to prove it wasn't their hardware/software that caused it.

And unfortunately, the people that would normally argue in favor of being reasonable with new tech will be suffering from inner turmoil as that idea conflicts with the "big corporations are ruthlessly profitable" belief.

It's gonna be interesting to watch, for sure.

Re:Litigious Forbearance

[linear a]

Inevitably sometimes it WILL be the manufacturer's fault. Paying for those liabilities has to be built into the system somehow. Nobody will want to build them until they don't have to face possible billion-dollar judgments against them. Also inevitable that some bad designs will get out that will be obvious in retrospect.

Re:Litigious Forbearance

[schwit1]

The manufacturers are going to need panoramic camera views that show the vehicle stayed in its lane, wasn't going too fast, maintained safe distances to other vehicles and also shows a darwin award winner walking into the path of traffic while texting.

Re:Litigious Forbearance

[vakuona]

If car manufacturers can show that they autonomous vehicles are on average far safer than the alternative, then I can see courts refusing to awards costs in excess of the costs that would be imposed on human drivers who caused collisions.

In fact, this would be ridiculously easy to legislate for. You can just require that autonomous cars be on average safer than their human driven equivalents before a manufacturer can sell them. At the point you are ten times more likely to be injured in a human driven car than in the autonomous equivalent, the judge can easily refuse to award punitive costs (unless you can prove that the manufacturer ignored an issue that made the vehicle particularly dangerous).

Fix date s/2013/2023

[jwag]

The date in the linked article is 2023, not 2013.

identify

[kqc7011]

When the automated car can tell itself that those two shiny things on the side of the road are deer eyes, then maybe and only maybe.

Re:identify

[dugancent]

If I see a deer that not moving, but is 20ft from the road, I slow down because I know it might decide to run in front of my car. It might also decide to stand there and not move. Reaction time has nothing to do with this.

Re: identify

[AvitarX]

I'm guessing it will see the deer's outline, because it won't be limited to human visible light.

Re:identify

[vakuona]

I see you scenario and I raise you the automated car's response.

Car sees deer lurking by the road side. Car blasts deer off the roadside to ensure it doesn't stray into the road.

Kidding aside, there is no reason a car couldn't do exactly what you would in such a scenario. As long as it can recognize that there is a large animal near the road (which it is likely to realise before you would), it can slow down, change lanes etc. It is also able to react long before you would. For example, upon noticing that there is a potential hazard in the road, it could preemptively move the brake calipers as close as it can to the disc brakes to minimise the time required to engage. You wouldn't normally want this in a human driven car because you have to use the feeling in your feet to regulate the brake force. An automated car doesn't have that problem.

If it couldn't prevent a collision, it can also prep all other safety systems to ensure that the collision causes as little damage to the occupants as possible e.g. gets the airbags ready to go etc.

Finally, once the collision happens, it can continue to steer the car out of trouble, which can be pretty important if there are other cars around. Many times, once a car is hit, humans lose control. The automated systems in a car, assuming you haven't hit a wall, can continue to operate until the car comes to a safe stop.

There are many ways for an automated car to do much better than human in most scenarios, including scenarios where an accident becomes inevitable. A computer can also choose to crash in the best possible way to reduce damage to yourself.

Re:identify

[itzly]

An IR camera can see a dear hiding in the grass from much further away than you can see.

feds should allocate a lane for this.

[WindBourne]

Seriously, if our federal highways were to allocate the inner most lane as being for automated cars starting in 2020, AND allowed 20 mph higher speeds, this would encourage faster adoption of automation for cars and trucks. Ideally, the vehicles would require a freq for talking to each other with as well.

Re:feds should allocate a lane for this.

[globaljustin]

agree on dedicated lane...everything else is just PR

They're nuts

[whitroth]

Legal:? Last week, I heard a speech by Vint Cerf, and he says the next iteration of google car will NOT HAVE A GAS OR BREAK PEDDLE, OR A STEERING WHEEL AT ALL.

              mark

Re:They're nuts

[ChoosyBeggar]

That's "brake pedal", Hoss.

1.21 Gigawatts...

[dcw3]

Obligatory Back to the Future reference.

Re:2013 or so? Because we are already in 2014?

[NEDHead]

Free Will may or may not be a myth, but in either case you really didn't 'have' to type something there.

What needs to happen

[linear a]

Not a complete list but I believe these all must exist before broad general acceptance. 1 Autonomous driving becomes better than human driving, including "edge cases" (e.g., junk falling from a truck, ball rolling out in front of car, etc). 2 Some way to deal with the inevitable liabilities. Cars will still kill people. Maybe something built into all car insurance and/or into the price of cars to fund liability payments. 3 Some way to deal with all the insane people in non-automated cars.

2013 is a typo, sorry 'bout that

[ashshy]

OS here. Sorry about the 2013 typo; Musk is aiming for 2023 at best:

http://www.huffingtonpost.com/2014/10/15/tesla-driverless-cars_n_5990136.html

I blame the lack of autopilot for these human fingers.

Re:2013 is a typo, sorry 'bout that

[TheCreeep]

I'm disappointed.
If Musk now said he was aiming to do it by 2013 I'd be more impressed.

Re:2013 is a typo, sorry 'bout that

[eth1]

I blame the lack of autopilot for these human fingers.

I blame the faulty autopilot that's running the Slashdot editors.

Re:2013 is a typo, sorry 'bout that

[Bob9113]

Read this previous article so you can more fully appreciate the extent to which Elon Musk either is misinformed or is misrepresenting what it takes to drive a car. Computers are amazing at playing chess or doing math. They still can't compose a sonata. Driving a well instrumented path in the most predictable conditions is like the former. Reacting to the unexpected without exacerbating the situation is much more like the latter; and the latter is where most of the accidents happen.

Everyone is waiting for California

[macsimcon]

A few months ago, I attended a talk on autonomous vehicles at the Petersen Auto Museum in Los Angeles. The executive from the California Department of Transportation told us that they’ve met with dozens of representatives from different states and countries, and they are all waiting to see what happens here.

California already has laws allowing the testing of autonomous vehicles, and many manufacturers have enrolled. They counted fifteen companies that were working on autonomous cars, including Toyota, Volvo, and most every car company you could name.

They described the five categories of vehicle automation, and explained that the first autonomous (not Musk’s so called “autopilot” which isn’t) vehicles will hit the road in the summer of 2015.

Re:Everyone is waiting for California

[Anonymous Coward]

I don't have much hope, since self driving cars will be shown to cause cancer in California.

Re:Everyone is waiting for California

[awtbfb]

They described the five categories of vehicle automation, and explained that the first autonomous (not Musk’s so called “autopilot” which isn’t) vehicles will hit the road in the summer of 2015.

Here's the levels. Most high-functioning systems on the market, like the Tesla version, are in the Level 1-2 range.

No-Automation (Level 0): The driver is in complete and sole control of the primary vehicle controls – brake, steering, throttle, and motive power – at all times.

Function-specific Automation (Level 1): Automation at this level involves one or more specific control functions. Examples include electronic stability control or pre-charged brakes, where the vehicle automatically assists with braking to enable the driver to regain control of the vehicle or stop faster than possible by acting alone.

Combined Function Automation (Level 2): This level involves automation of at least two primary control functions designed to work in unison to relieve the driver of control of those functions. An example of combined functions enabling a Level 2 system is adaptive cruise control in combination with lane centering.

Limited Self-Driving Automation (Level 3): Vehicles at this level of automation enable the driver to cede full control of all safety-critical functions under certain traffic or environmental conditions and in those conditions to rely heavily on the vehicle to monitor for changes in those conditions requiring transition back to driver control. The driver is expected to be available for occasional control, but with sufficiently comfortable transition time. The Google car is an example of limited self-driving automation.

Full Self-Driving Automation (Level 4): The vehicle is designed to perform all safety-critical driving functions and monitor roadway conditions for an entire trip. Such a design anticipates that the driver will provide destination or navigation input, but is not expected to be available for control at any time during the trip. This includes both occupied and unoccupied vehicles.

U.S. Department of Transportation Releases Policy on Automated Vehicle Development

Re:Everyone is waiting for California

[Lord_Jeremy]

If the Google cars are level 3, then why do every single video I ever see of them being driven show the person in the driver's seat sitting alert with their hands less than an inch away from the steering wheel? It definitely appears that they're not expecting a comfortable transition time. From some of the other stories I've read here regarding these Google cars, it can happen rather suddenly that the driver takes over because they realize the navigation computer doesn't recognize some road obstacle or driving hazard. They even have a guy in the passenger seat with his eyes glued to a laptop, watching a real-time visualization of what the cars "eyes" see.

never going to happen

[slashmydots]

One: the vehicle needs to never run anyone over ever. That will not happen. You hit one single piece of property or person and it might as well be Nancy Grace's house you hit because she and the rest of the media will tear that technology to pieces.

Two: computers need to be as smart or smarter than humans at visual object recognition and adaptive decision making. Right now they're about 0.001% of the way there.

Re:never going to happen

[0123456]

#2: They don't need to be as good as humans are in the general case. Only in the specific domain case of: Hit the brake before you collide.

So it's OK if they go straight through red lights that aren't in the database and drive along the sidewalk because the database is wrong, so long as they stop before they hit anything?

Re:never going to happen

[KeensMustard]

So it's OK if they go straight through red lights that aren't in the database and drive along the sidewalk because the database is wrong, so long as they stop before they hit anything?

Well, yes. I have not problem with that, given that the purpose of red lights is to prevent collisions.

Humans go through red lights all the time. If it is ok for humans to make an occasional mistake, it's ok for self driving cars as well.

Re:Well....

[Nexzus]

Cash for clunkers removed about 700,000 cars from the road. There are about 250,000,0000 vehicles in the US.

Solar Freakin Roadways!

[axehind]

What will it take?!?!?! Solar Freakin Roadways!

How about roads that assist?

[DumbSwede]

We probably could have automated Highways since the late 70’s. In the 50’s engineers imagined all the brains and machinery for automating Highways in the highways – and would have been incredibly expensive. Now we want to put all the brains and sensors in the car, again (at least for now) incredibly expensive. There must be a sweet spot of compromise for 90% of driving situations that requires only modest changes to our transportation infrastructure and doesn’t require the cars to be capable of handling every conceivable scenario most humans can. Build a series of automation friendly/automation assist sensors, transponders into some long haul stretches of our nations transportation infrastructure and it would probably lead to a snowball effect of getting enough earlier adopters of automated assisted driving that improvements would come even more quickly in cars and roads leading to a truly autonomous era.

Expect a push from the Insurance Industry

[nehumanuscrede]

Though, I'm not quite sure in which direction.

On the one hand, the number of claims should bottom out once self-drive cars are in place and the bugs sorted out.
On the other, they'll have to re-calculate how they determine premium rates since the " human driver " factor will be (mostly) removed.

So, while they won't be paying out nearly as much in claims, they won't be taking in nearly as much in premiums either. Should be interesting.

Speed Traps will no longer be the revenue-cow that many towns rely on. Red-light cameras and similar tech will become a waste of time. How WILL Law Enforcement pay for their Soldier-Wanna-Be toys . . . :|

Hell, these things, once mainstream, will also shift the entire traffic structure around. Stalls, wrecks, weather, and other rubber-neck variables will pretty much go away meaning a much better driving experience. Great for the driver (passenger ?) , but probably not so great for the State / City governments who just LOVE congestion because it pushes the traffic onto their Toll Roads which they seem to be building in greater numbers these days. I would expect to see the Toll Roads become ghost roads ( in those areas where the Toll Roads are a means to bypass highway congestion and not the ONLY means into or out of an area ) as the reasons for utilizing them in the first place will become irrelevant.

Will need to put some more thought into it, but I bet the introduction of the self-drive vehicle will impact quite a bit of modern day revenue-generators which will probably cause a major panic along some lines. lol

Re:Expect a push from the Insurance Industry

[gurps_npc]

Insurance companies will push heavily IN FAVOR of auto-autos.

This assumes that robots are safer drivers than humans (which is an obvious requirement before they legalize it).

The reasons are clear:

1) Car insurances don't want to pay you because someone else hit your car, but they can't prove it. Robot cars decrease this risk.

There is a LOT of money spent by the insurance companies trying to prove fault. It is big business. By reducing the actual risk from other drivers, insurance companies will save billions, even if they never insure a robot car.

Also, insurance companies make money when things become safer - because rate changes are always behind actual risk changes. So more safety always equals insurance profits and less safety always equals insurance losses.

I agree that speed traps and red light cameras will vanish, but I am not so sure about toll roads. In fact, they might grow in power, using the robots to connect tolls. They might simply have a tax charge to drive fast in the state. As in, your robot car will be limited to 50 mph unless you purchase the NJ Fast Lane upgrade from New Jersey Transit.

Re:Expect a push from the Insurance Industry

[ezelkow1]

The one question I have on the insurance situation is when an autonomous car causes an accident who is fault? The manufacturer or the owner? Since the owner was not in command of the vehicle I would think it would be the manufacturer.

So then take the auto-auto scenario to its end conclusion, no one drivers manually ever anymore and the manufacturer is always the one at fault. It may take 50 years from the intro of the auto/manual version, but it would come eventually. At which point auto insurance would be out of business. Surely the insurance industry would be apposed to something they know would eventually put them out of business

Re:Expect a push from the Insurance Industry

[jader3rd]

The one question I have on the insurance situation is when an autonomous car causes an accident who is fault? The manufacturer or the owner? The manufacturer. Unless there's evidence that the binaries for the AI system were hacked.

Re:Expect a push from the Insurance Industry

[31415926535897]

They will make all of the roads toll roads.

Insurance

[kwiecmmm]

One of the biggest issues is going to be insurance and who will pay when one of these cars causes an accident.

While most bugs will have been worked out, there will still be a situation where one of these cars is going to cause an accident. And I have not heard anyone talk about who will pay when this occurs.

Re:Insurance

[Shimbo]

One of the biggest issues is going to be insurance and who will pay when one of these cars causes an accident.

I doubt this will be a big problem. Insurers will sooner or later offer policies for self-driving cars, and if the statistics are good, they will eventually be reasonably priced; you can only price fix for so long. What will take longer is for governments to relax the criminal liability of the driver.

*nothing*

[globaljustin]

We will never have truly *autonomous* vehicles driving on the same lanes in the same traffic as regular diver vehicles.

The problem is *not* technical in nature ultimately (now, the tech is not near sufficient, but assuming it could improve), the problem is *liability* for when something goes wrong.

What will happen: Dedicated lanes on interstates

Like HOV lanes, basically.

The only way it will actually be implemented is in controlled zones where there are much fewer variables...

To think anything else is magical thinking and not connected to reality

Re:*nothing*

[devman]

Driverless cards will have all kinds of sensors and likely have blackboxes as well. If one gets in to an accident data from those senser coudl be used to reconstruct the cause of that accident and assign blame. Insurance companies will love it assuming driverless cars are safer and turn out to be the victim of accidents more than the cause of accidents. Liability can be covered similar to the way it is handled now. Operator indemnifies manufacturer and carries an insurance policy to cover the assumed risk, those premiums will reflect the risk of driverless car being at fault in an accident.

Re:*nothing*

[gurps_npc]

That is a very poor argument.

Liability is an EASY problem, not a hard one. No fault insurance is a real thing that already exists in some states and countries.. It already solves the problem

Your thinking is pessimist foolishness, ignorant of real world solutions.

Liability will do nothing more than hold up robot cars for MAYBE one year.

But no fault insurance, while it could solve it, will probably not be the solution to this issue. Car companies are already responsible for defaults in manufacturing. Given reasonable safety records, car companies will probably end up INSISTING on taking all the risk. It will be a big selling point, that when you buy their car, they give you free insurance. They will claim that their software is so safe, that they can offer that deal, but manual cars are so dangerous, they can't.

Has to be unhackable

[Ann O'Nymous-Coward]

Otherwise, there'll be a whole new class of assassination: Death-By-Driverless.

Re:Has to be unhackable

[russbutton]

Unhackable... Just as unhacakable as banks, on-line retaliers... sure...

If I'm a terrorist, what could be better than to have hundreds of thousands of networked moving vehicles I could take over from half-way around the planet? How much fun would it be to order it to make a car do a hard left turn as soon as it hits 70 mph, which would only happen on a freeway. And imagine being able to do that to many thousands of cars all across the USA and at random intervals?

Sorry. Autonomous, networked, driverless cars is waaaaaay beyond stupid.

Re:Has to be unhackable

[Ann O'Nymous-Coward]

Exactly.

Would they approve this?

[paulpach]

Suppose you tell the current politicians:

"Hey, I got this great invention, it will improve our transportation 10x, but it will require highly toxic and flammable chemicals to be stored in underground tanks every other block in highly populated areas. It will also cause around 1.3 million deaths per year worldwide, and become the #9 leading cause of death".

How many current politicians would approve this?

That is right, if it was for the current politicians, the car would not exist. We would all be still riding horses, which are even worse for the environment and have a fatality rate of 1 per 10,000 riders.

So I don't hold my breath for current politicians to approve self driving cars, even if 1 person gets killed by a self driving car, politicians will scream "think of the children!" and ban this right away, it will not matter if the fatality rate is even lower than manually driven cars.

Re:Would they approve this?

[aclarke]

A similar thing happened around the time of the invention of the automobile. If you're not aware of the early "red flag laws", here's an interesting article about them. It took a couple decades before horseless carriages could legally be driven on the road without what we would now consider ridiculous hindrances.

Hope it's sooner than he thinks

[bhv]

I look forward to the beginning of autopilotmobiles. However, I'll be one of the last to participate.

I picture the entertainment value of the work commute increasing exponentially. For example, if I need to take a lane I'll look for the autopilotmobiles and just cut in front, knowing that it will see the impending doom and hammer on the brakes.

If they make dedicated lanes then there best be a barrier between them and us regulars, it's impossible to keep the single drivers out of the existing HOV lanes.

Re:Hope it's sooner than he thinks

[vakuona]

And the autopilotmobile is probably recording everything and reporting you to the relevant authorities.

And then you get banned from driving.

It won't happen that way

[NReitzel]

The oncoming of fully automated vehicles won't happen the way that being discussed in geekish circles. Governments tend to move with all the speed of a glacier, and insurance companies will go out of business if the number of traffic accidents plummet. (Yes, they will. Water conservation sounded great until a lot of people started actually conserving water, now the water companies are having to jack up rates to stay solvent.)

What will happen is that "safety features" will be added to top end vehicles and work their way down. This is already happening with rear-watch, lane obstacle detection, and others. Insurance companies will like safer cars, as long as they aren't so safe that they are no longer needed. Public safety groups will lobby for these safer cars.

The myriad of state legislatures in the US will be very reluctant to authorize fully automated vehicles. Instead, manufacturers will just keep introducing "features" that reduce traffic accidents, things like lane following and collision detection and braking. Then, as the number of features mounts, the distance between a fully featured safety car and one that will drive itself will become smaller and smaller until it doesn't seem like such a giant leap. In addition, we may find automated vehicles licensed only for certain pieces of highway. It takes a lot of CPU to automate a car, adding GPS is a detail.

Look around, the changes have already started.

It's very simple

[schklerg]

If the laws allow open containers (alcohol) in automated vehicles, all downsides will be ignored and the population will demand them.

My Blind Spot Indicator still isn't perfect

[kannibal_klown]

My blind spot indicator on my car still isn't perfect and I still have to check.

Just today I noticed a black hatchback in my right blind spot that the indicator didn't pick up. I don't know if it was dirt on the sensor, the color of the car vs the blacktop, etc.

So... I don't know how much I want to trust a car that fully relies on that.

Because if I have to babysit the car the entire time, I might as well drive.

Eventually maybe, and hopefully within my lifetime. But I won't be using one any time soon.

Robust Software!

[MagickalMyst]

>What Will It Take To Make Automated Vehicles Legal In the US?

Above all, robost software!

The word "crash" becomes doubly important in this case.

Avoiding Squirrels...

[QuietLagoon]

The self-driving cars have a long way to go (no pun intended) before they should become ubiquitous.

Would rails be a solution?

[Anonymous Coward]

Many cities already have rail-based streetcars and streetcar drivers aren't really doing much driving, stopping, going, the occasional turn, occasionally getting out to switch the track, etc. Seems to me an automated car could be designed to use those same rails. Thus, automated cars could piggyback on whatever legal framework or traffic infrastructure allowed streetcars in the first place. A lighter (car only) rail system could also be extended to major roads where it doesn't already exist, expanding the network. A hybrid approach could see the cars flipping into smart car mode from autopilot mode once the car leaves the tracks to traverse normal roads. Automated cars could even conceivably deposit their passengers at the same stations as both streetcars and subways. I know this idea isn't new but maybe it's time has come?

How about no-fault

[scamper_22]

It is naturally a legal question of who can be held liable should an accident happen.

People want to assign blame to the car manufacturer, to the driver...

There are valid cases that can be discussed on an individual level for each one. If a manufacturer has a bug in the code that results in a crash and did not do due diligence, that is one thing.

But there is another solution that many people don't consider.
NO FAULT INSURNACE

This is how it is in Ontario, Canada. It has it's flaws, but the concept is really good and I think would work really well with automated cars.

http://www.ibc.ca/en/car_insur...

Basically, you get your insurance from your own insurance company and they cover your losses.

Person A and person B get into an accident.
A gets damages resolved from A's insurance company.
B gets damages resolved from B's insurance company.

A doesn't care about getting money from B or dealing with B's insurance.

Generally what happens here is your insurance premiums go up the more risk you take on (fancy red sports car = more risk) or the more claims you've had in the past.

So we could easily see insurance companies evaluating automated cars. Cars with better ratings/systems can get lower insurance premiums...

Again, this does not mean other liability causes are out of the window. A person who disables their auto system and crashes or a does not maintain scheduled maintenance, or a manufacturer who has a bug...those are separate issues.

Re:How about no-fault

[gewalker]

There are about a dozen states that allow (buyers choice) or require no-fault automotive insurance.

There used to be more, but average premiums were clearly higher for no-fault insurance, so some states reverted to traditional tort insurance.

You may be right about it making more sense for insuring robot cars.

Re: Welcome to slashdot

[vakuona]

Very little of the technology we have today was developed due to emergency needs.

The only thing that matters is that someone develops the technology, and it is good enough.

However, on the subject of automated cars, the real difficulties is how to introduce them on roads where you still have "manual" drivers around (I don't mean stick here).

Computer driven cars can be totally predictable. Manual cars, less so - people forget to use their signals, people make late lane changes, fall asleep and drift into other lanes etc.

same as always...

[schlachter]

An economic incentive that causes many companies to put pressure on their representatives...

or

A major tragedy that could have been avoided with autonomy that gets continual news coverage (not that this worked for gun control) ...otherwise, reason and good sense have no place here

Re:same as always...

[nine-times]

A major tragedy that could have been avoided with autonomy that gets continual news coverage (not that this worked for gun control)

Yeah, I'm not sure this works in general. For one thing, we have loads of tragedies from car accidents-- I think it's as low as 30,000 deaths per year in the US now, but a few years ago it was more like 40,000. It's possible that many of these could be fixed by self-driving cars, but the American public doesn't care because we love to pretend we're race car drivers.

But also, in general, there can be all kinds of tragedy and scandal, and politicians will go spend their time on damage control, making it seem like they're doing things, going on TV and saying what people want to hear. And then they'll wait for people to stop paying attention, and they'll do whatever will get them more campaign contributions from rich people.

Public outrage is only effective during an election, and even then only when it threatens to keep politicians from winning their elections. Attention spans are short, and politicians don't care if the outrage isn't wrapped in money.

Re:same as always...

[nukenerd]

we have loads of tragedies from car accidents--... but the American public doesn't care because we love to pretend we're race car drivers.

No, it is because people would soon get bored if every road accident received as much coverage as, say, someone killed in a mountaineering accident (not to mention how voluminous the coverage would need to be). Generally, public reaction goes up exponetially with the number of people killed at the same time. In the UK, a railway accident that kills 10 people (about the number killed every day on the road) will get national headline coverage for about a week; but one person killed in a road accident will just get a few column inches in a local paper.

Another reason is that people hate the idea of being killed by an entity rather than by another person. They think being killed by another person driving a car is in some way "democratic", whereas being killed by a train, lift, ship, crane is not. Or put another way, when they read of a road accident they believe that if they were there (either as culprit of victim) they could have influenced it differently, whereas they know that in a train accident they could not have done.

Re:same as always...

[nine-times]

No, it is because people would soon get bored if every road accident received as much coverage as, say, someone killed in a mountaineering accident

True. But also, people are sensitive to any comments about the dangers of automobiles. People love cars, they like to pretend they're race car drivers, and the get severely butt-hurt if anyone suggests that maybe they shouldn't be driving. If you take a step back, it's really amazing how many people will argue fiercely that they are great drivers, that they are safe drivers, and that they're in control when they're driving. It's especially noteworthy when placed in contrast to the fact that they can't possibly control what happens on the road, and by the statistics surrounding car accidents resulting in injury and death.

People are emotionally attached to their cars, and they're in denial of the dangers that the car represents. There's no other reasonable way to see it.

Blue Collar VS White Collar

[BrendaEM]

On the one hand we have internet providors effectivly stopping community internet.
On the other hand we have a lot of blue collar people who are going to be put out work.

While I cannot believe that the Teamsters and the Cab driver unsions are just laying down on this issue, in the end, the might not be able to stop driverless cars, because that's what Darpa wants for for their war machines.

Who is going to be the first unlucky person to die for driverless car research?
Who is going to be the furst unlucky person to die from a driverless war machine?

think of the insurance companies...

[SethJohnson]

.....and ban this right away, it will not matter if the fatality rate is even lower than manually driven cars.

You ignore the gargantuan influence insurance companies wield over politicians.

Who do you think got these types of laws passed?

• No smoking in bars
• No sodas sold in big cups
• Mandatory seat belts
• Child safety seats

Those were the doing of an entity who could see that modifying these behaviors would reduce the payouts they make each year. This entity lives and breaths statistics and charges its customers based on anticipated payouts and profits off the difference. By modifying the behaviors while keeping the premiums at the same level, the insurance companies are able to expand their profits. Insurance companies use these profits to control politicians.

Self-driving cars are hugely attractive to insurance companies. If they can overall reduce payouts by some small number, they'll happily pay for the fewer claims made against their customers' self-driving cars. Should cases go to court, they'll have plenty of telemetric data to throw in front of a jury to bolster their defense.

Driverless Trains are Here

[Roger W Moore]

OTOH, driverless trains would be an order of magnitude or two easier and... not there.

Wrong - driverless trains are already here and have been since 1967 which means they have been around longer than most Slashdot readers including myself. However, like driverless cars, the first line which was automated, London's Victoria line, still carried drivers because of the unions i.e. politics. More recent lines, like the Docklands Light Railway (DLR) are driverless and there are plans to upgrade other tube lines to true driverless operation.

Given the politics with driverless trains should we expect driverless cars to be any different? I expect there will be a period of adjustment where they are allowed but heavily restricted as we learn to see how well they cope. Assuming they perform these restrictions will be gradually lifted.

Re:Driverless Trains are Here

[HuguesT]

Underground/Subway/Metro or some other specific lines with zero interconnection running as a loop, yes. They are all akin to a long, horizontal elevator, with lots and lots of sensors and other feedback systems built into the track. They make a lot of sense because they carry a lot of passengers over short to medium length lines over the very same tracks all the time. The longest automated network is the Vancouver skytrain, which is about 70km long.

General-purpose train lines, with something unlike single-purpose engines running on open tracks with interconnection ? The page does not list any. It is still too difficult for automated systems.

As someone who has driven before,

[txoutback]

It's hard to imagine how autonomous vehicles can exist safely amongst vehicles driven by inebriated, distracted, careless, or angry humans.

Re:As someone who has driven before,

[n6kuy]

Well then, the 2nd phase of the project will be to outlaw manually driven cars.

Here's to hoping

[DaMattster]

That it never happens! And I admit, the reasons are entirely selfish. I drive a truck for a living so I'd rather not lose that livelihood. I can, however, see a place for adding automation in the safety area like automatic braking if following distance is too close and even maybe a limited autopilot feature that would still require a human being to be sitting in the driver's seat. I guess I need to face the reality that just about anything can become automated and I'm sure my employer has one keen eye cocked towards the progress in automation. I left IT for the love of the open road. Part of me understands why some companies might really want the automation because truckers are a notoriously difficult lot to deal with - basically macho juvenile adults. I'm sure drivers are the bane of the transportation company's existence. Many drivers have an undeserved sense of entitlement, make demands, and are often rude towards the office staff. Little do those guys know they are basically going to "demand" themselves out of a job because don't think for a New York minute that the big boys like Werner and Swift aren't making keen investments in automation.

Re:Here's to hoping

[Joe_Dragon]

so you did not hear about the auto drive system? are you in the union?

Speed Limits

[Anonymous Coward]

Personally, I can't wait for these to get here.
If they are SO MUCH better drivers than we are, and can react SO MUCH faster, then the speed limits can SAFELY be upwards of 120, right?

First step is identify the problem, if any

[radarskiy]

The first is always to identify the problem. Where and how are they currently not legal? You would think that an article about making automated cars legal would say that they are prohibited in state X because of A and in state Y because of B. However, the article is *completely silent* on current laws that prohibit automated cars.

What indeed

[ArhcAngel]

"What Will It Take To Make Automated Vehicles Legal In the US?"
Simple...The appropriately greased palm of a majority of any politicians publicly opposed to said legislation.

We're really not a decade away for driverless cars

[cyn1c77]

I don't know why the OP thinks that the "legal framework" is an issue? The government has been ignoring legalities for over a decade now.

The lobbyists will decide when driverless cars will be legal.

My guess is that it will all be sorted out right after the mainstream US automakers (Ford, GM, and Chrysler) have their self-driving cars ready to go.

Be as awful at driving as people.

[gelfling]

Just today I was behind someone who for no obvious reason took both hands off the wheel, waved them out the window for about 30 seconds, veered out of their lane and hit a stopped car. Robo-cars I think have to be able to be no BETTER than that for acceptance. I think they're on the wrong track with trying to be BETTER than humans. They should aim for being just as terrible because people don't want to feel the machines are mocking them.

Clearly no one cares who dies right now so that's not a problem.

The 'OFF' switch

[kheldan]

So long as it has a steering wheel, accelerator pedal, brake pedal, and an immutable 'OFF' switch, they can put it in any car they want, but I will NEVER own any car that is just a box I have no manual control over. Ever. Not even once.

Re: Welcome to slashdot

[David_Hart]

Very little of the technology we have today was developed due to emergency needs.

The only thing that matters is that someone develops the technology, and it is good enough.

However, on the subject of automated cars, the real difficulties is how to introduce them on roads where you still have "manual" drivers around (I don't mean stick here).

Computer driven cars can be totally predictable. Manual cars, less so - people forget to use their signals, people make late lane changes, fall asleep and drift into other lanes etc.

In theory, a self driving car should have enough sensors, sensor data gathering speed, processing speed, etc. to be able to "see" objects moving towards them and take appropriate action to avoid an accident.

If an automated car cannot deal with manual drivers, I shudder to think how it would deal with stuff falling off of trucks, animals running into the street, sudden downpours or snow squalls, mechanical problems, etc....

Similar issues to ou favorite tech: Drones

[recharged95]

If Drones ...like the autonomous ones I'm working on can give you any indication, it's going to be some time.

The gov't is having a beef about FPV and manned drone flight. I already know they will have a heart attack on the autonomous ones... sort of says what they're going to think about fully autonomous cars... which uses essentially the same tech and concepts.

Security Implications

[peanutious]

We need to better address security implications before autonomous cars become mainstream. How do you stop someone from programming a car to drive to a destination and then explode? If the driver doesn't have to be in the car, that same single "driver" can continue to cause mayhem with multiple cars.

Maintenance

[rykin]

I'm curious how insurance companies will deal with proper maintenance for automated cars. It's no mystery that bald tires don't handle as well in bad weather as new tires. What happens when an autonomous vehicle gets into an accident because it attempted to stop, but the lack of tread on the tires caused it to hydroplane much more than normal? I imagine insurance companies will want documentation of regular maintenance to prove that the error was not neglect on part of the owner.

Just what we need....

[Benders]

Cars that will allow the US driving public to drive around completely asleep! Over 75% of the driving public drives around asleep at the wheel now! And the car makers keep adding "safety" features and alarms that try to keep those asleep from running over someone else. Is this really progress? We teaching folks that don't even have to be personally responsible even when they are piloting a 3,000 pound weapon!

Docklands Light Railway

[Roger W Moore]

General-purpose train lines, with something unlike single-purpose engines running on open tracks with interconnection? The page does not list any.

Clearly you have never seen the Docklands Light Railway (DLR): here's a map. This is not simple A-to-B track you can catch trains to different destinations from the same track on the same platform e.g. London City Airport has two platforms: one for trains heading to Woolwich Arsenal and the other which has trains headed to both Bank and Stratford and it used to have trains that also went to Tower Gateway but those were dropped before the London Olympics so the routes shown in the map are not static and can be adjusted to match traffic.