Fully Autonomous Cars Are Unlikely, Says America’s Top Transportation Safety Official

[FrodosPet]

Fully Autonomous Cars Are Unlikely, Says America’s Top Transportation Safety Official

Some people just like to drive, for one thing.

https://www.technologyreview.com/s/602273/fully-autonomous-cars-are-unlikely-says-americas-top-transportation-safety-official/

by Andrew Rosenblum August 31, 2016

Auto accidents kill more than 33,000 Americans each year, more than homicide or prescription drug overdoses. Companies working on self-driving cars, such as Alphabet and Ford, say their technology can slash that number by removing human liabilities such as texting, drunkenness, and fatigue. But Christopher Hart, chairman of the National Transportation Safety Board, says his agency’s experience investigating accidents involving autopilot systems used in trains and planes suggests that humans can’t be fully removed from control. He told MIT Technology Review that future autos will be much safer, but that they will still need humans as copilots. What follows is a condensed transcript.

~ snip ~

The ideal scenario that I talked about, saving the tens of thousands of lives a year, assumes complete automation with no human engagement whatsoever. I'm not confident that we will ever reach that point. I don’t see the ideal of complete automation coming anytime soon.

Some people just like to drive. Some people don't trust the automation so they're going to want to drive. [And] there’s no software designer in the world that's ever going to be smart enough to anticipate all the potential circumstances this software is going to encounter. The dog that runs out into the street, the person who runs up the street, the bicyclist, the policeman or the construction worker. Or the bridge collapses in a flood. There is no way that you're going to be able to design a system that can handle it.

The challenge is when you have not-so-complete automation, with still significant human engagement—that's when the complacency becomes an issue. That's when lack of skills becomes the issue. So our challenge is: how do we handle what is probably going to be a long-term scenario of still some human engagement in this largely automated system?

~ snip ~

I can give you an example I've seen mentioned in several places. My automated car is confronted by an 80,000 pound truck in my lane. Now the car has to decide whether to run into this truck and kill me, the driver, or to go up on the sidewalk and kill 15 pedestrians. That would [have to] be put into the system. Protect occupants or protect other people? That to me is going to take a federal government response to address. Those kinds of ethical choices will be inevitable.

~ snip ~

[longship]

1. One word. WINTER! nt

[metroins]

4. Autonomous snow plows and salters

Roaming the streets before the snow builds up, keeping the streets open.

[longship]

5. Not where I live, they don't.

Can one imagine autonomous semi-trailers on I-80 during winter? With two feet of snow on the Interstate in a couple of hours?

Right! That will work out really well.

When there's a driver behind the wheel, they know when to pull into the truck stop.
Biscuits and gravy is yummy! Autonomous vehicles know nothing about two feet of snow let alone biscuits and gravy at truck stops.

[metroins]

10. Data

A computer can move a wheel within a fraction of a second in a fraction of a millimeter of space.

A computer will always do most tasks better than humans, it just needs to be programmed correctly.

Take 100 skilled truckers, hook them up to sensors, drive a million miles each in the snow. Crunch the data. You have almost every scenario covered and the ones you don't, a human likely would've wrecked anyway.

You are correct, autonomous cars will know nothing about biscuits and gravy.

[Heeeeers Johnny]

2. Hopefully..

this "cutting edge" technology will die like Google glass did.

[Travis_0004]

7. I doubt it

Way to many benefits. It will allo the disabled and elderly freedom of movement, cut back on traffic jams, and save thousands of lives. The day is coming.

[loyalsister]

94. There are cases in which it would be highly irresponsible for disabled people to use them

The same goes for people who have even the mildest forms of dementia. This is not at all the boon for disabled people that some think it would be.

[metroins]

3. He's wrong

Vehicles will communicate.

The autonomous car would know the 80,000 truck was there before it approached.

That man is living in the past.

[FrodosPet]

8. By any chance, do you work for someone in autonomous vehicle design?

I am looking for a job in 3-D object identification - turning point clouds into actionable information. I am also open to vehicle to vehicle communications, or infrastructure to vehicle communications systems.

I've been involved with computers and radio communications since I was in elementary school, back in the late 60s. I've been studying autonomous vehicles ever since the DARPA grand challenge about 12 years ago. So yes, I have skills and experience, along with a strong desire to make auto autos as safe as possible. I am assuming by the certainty of your statement you are in the field and have insider knowledge about vehicle to vehicle communications, and perhaps efforts to mandate the retrofit of commercial vehicles with this technology?

[metroins]

11. I'm on the insurance side of autonomous algorithms

Predicting future impacts on our industry by this emerging technology.

[Lee-Lee]

80. How do you see liability in courts playing out?

No technology will be perfect.

When an accident occurs and someone is killed or injured, who is liable?

The car owner? They didn't cause the accident.

The car maker who configured the sensors and programming in a way that allowed the accident?

The coder who wrote the actual lines of code that caused the failure?

[longship]

9. What infrastructure will they have here in the National Forest in rural Michigan?

All the roads are two lane, without any shoulders. Most are unpaved and less than two lanes. In winter one waits for the snow plow before one even pulls ones car out of the garage, presuming one has such a luxury.

There will be no autonomous vehicles here, EVER!

We don't even have cable TV or broadband Internet here. It's dialup or cellular here, the latter only in the past three years. And that is hit or miss.

[metroins]

12. Data

Input, output.

A computer can navigate terrain without markers.

There are currently rock crawling autonomous vehicles.

[longship]

17. The only markers here are the edge of the water.

And an autonomous vehicle had better not tread into that territory.

Did I say that most roads here are unpaved. BTW, they are also mainly unmapped. And when it snows here, only farmers' tractors drive here, with a human behind the wheel. That's how many of us eventually get out in winter. We all help each other.

A autonomous vehicle would be utterly stupid here. It would end up in a lake or river.

[metroins]

20. If a human can drive it

A computer could do better. It just needs to be programmed correctly.

Ignoring that, even if they don't go where you live, 90% of vehicles could be replaced.

[longship]

23. Absolute rubbish.

You apparently do not live in the National Forest. This ain't San Fucking Jose. It isn't even Death Valley. It's the fucking national forest where roads are unpaved, narrow, often blocked by any number of impediments. And in winter... Well, let's just say that one wants some intelligence behind the wheel.

And BTW, fuck Ray Kurzweil!

[metroins]

42. None of that even matters

You're talking 1% scenarios.

It could be programmed, but it doesn't matter. Driving as we know will be changed.

[A HERETIC I AM]

84. I agree with Longship. Not only absolute rubbish, complete and utter bollocks.

Here's when automonous vehichles will be able to operate everywhere;

When the standards for highway width, markings, lane closures, construction zones, emergency vehicle movements, cone and barrel placement and EVERY SINGLE OTHER OF THE THOUSANDS OF VARIABLES are standardized.

I drive a tractor trailer for a living, with well over 2 million miles of over-the-road riving under my belt and I can tell you with absolute certainty that no autonomous vehicle now or even on the drawing boards will be able to determine with absolute certainty, in the varying conditions I come across on a regular basis, whether or not shifting to the left (or whatever) 6 inches a split second before necessary, will keep the vehicle from taking off the construction workers leg.


Sure, My thought process isn't as fast as a computer may potentially be, but I understand nuance and shading. And I mean SHADING, as in whether or not there is enough of a shade to say yes or no to altering my direction when I come across a worker in heavy fog.

The first time an autonomous vehicle takes out half a dozen construction employees will set the entire idea back a decade, and it has already been demonstrated that the technology now in place is far less than perfect, not to mention downright deadly in difficult to understand circumstances.

[Egnever]

27. This sort of denial amazes me

they did this sort of thing successfully over 10 years ago.

[longship]

30. Staged competition.

I am aware of the DARPA challenge. It's cool. But I don't see any lakes, rivers, snow, or even rain there.

When one is limited to CA desert one can make certain assumptions that are not operative here. And in fucking winter, that's an entirely different ball game, when the unpaved roads basically just disappear. That's when one wants real intelligence behind the wheel.

[Egnever]

33. The thing is

While the snow may hide the road from human eyes the autonomous vehicle still knows where it is.

Not sure how you can call a dirt road with a drop on one side staged.

Snow will certainly be a challenge but an autonomous vehicle will know when it is drifting and adjust much faster than a human will. It will know when it loses traction much faster than a human will. The list goes on and on. We like to pretend we are driving masters but our reaction times don't come close.

[longship]

37. Great! How in the fuck does it know?

Grass Lake Road is a difficult drive because:

1. It is narrow, less than two lanes.

2. It is unpaved and without any shoulders.

3. Adjoining Grass Lake, the lake is less than a yard from the road. The lake's water level is also higher than the road surface, kept at bay only by the rubble of cut down trees.

4. But Grass Lake Road is the fastest, most direct route from the west to reach Hawkins, MI, which happens to have a rather marvelous place to eat, the Last a Chance Saloon. It also happens to be the only commercial enterprise in Hawkins, so it's got that going for it. (That and the best fish and chips in Michigan.)

5. Grass Lake Road is not mapped. It twists and curves like a snake. No autonomous vehicle could ever manage it. Forget it in winter! You won't find it on Google, either.

[Egnever]

39. They have sensors that read all of that.

The dirt road challenge with the drop off showed they can already do it and again that was over ten years ago.

If you want to claim they wont happen based on some road in the middle of no where traveled by no one you are reaching. The overwhelming majority of the roads in this country are nothing like that.

Are you talking about this grass lake road?


Grass lake road MI

[metroins]

43. Some people

Just seem to think humans are smarter than we actually are.

They don't realize driving takes very little skill, even in difficult situations. It takes time to program situations, but even those difficult situations can be simulated billions of times in a sandbox and correctly designed for.

Put 10000 humans in cars, you'll have 100 accidents. 10000 computers may have 10 accidents. I trust auto pilot in the future.

[longship]

44. The car will end up in the fucking lake.

There are absolutely no visual clues beyond knowing where the road goes. And a minor mistake puts one off the road into some real trouble, like maybe into a fucking lake.

Too many people are trying to live in The Matrix. Autonomous vehicles are a Ray Kurzweil wet dream.

[Egnever]

45. Was that the road? I linked above?

[longship]

53. Nope. Not the same road.

It is in Newaygo county and is not even on many local maps. The only reason why it is passable in winter is because it is on a school bus route. That is the same as the dirt road I live on, which remarkably turns into Grass Lake Road if one keeps on keeping on.

When school busses must transit, the snow plows will be there. Otherwise, you are screwn in winter here. I make sure that the freezer is full.

And there will be no autonomous vehicles here for decades. Hell! We don't even have cable TV or broadband Internet. My water comes from a well, and sewage goes to a septic tank. When the power goes out I lose everything.

[Egnever]

57. Well when you start talking roads that people can't drive for long periods of time

I am going to agree with you that driverless cars won't be going there any time soon but then neither are regular cars.

You don't seriously believe that they wont produce driverless cars because you can find roads that are impassible?

Thats like saying we wont make donuts because some folks have diabetes.


Obviously you are going to be able to find an example of roads they are not going to traverse. That has little to no effect on most driving in the US.

[longship]

60. Let me give you another example.

The Grapevine on I-5 in SoCal, the Tejon pass. There is no way autonomous tractor/trailers will ever be allowed to traverse the Grapevine, even in summer. It is a very difficult drive even when there is intelligence behind the wheel. And that's a multi lane Interstate!

Another example is the similarly named Cajon pass east of LA, towards San Berdoo.

The insurance companies would never allow it.

[Egnever]

63. Not sure what the point there is

If it was a risky road the car would know and avoid it. If you absolutely wanted to take that road and the car manufacturer or the insurance company deemed it too dangerous the car would balk at traveling it and suggest an alternate route.

Again you are talking a small fraction of the roads in the US and again the vast majority of people will never need to be on those roads. That you think that will stop driverless cars from being on the roads just makes no sense to me.

[longship]

66. Well, it is the major Interstate highway through SoCal.

And BTW, it isn't the only mountain pass.

Let me tell you that you do not want a tractor/trailer behind you on a mountain pass. They cannot stop! They cannot even slow down. That is the time that you want intelligence behind the wheel.

One of the main arguments for autonomous vehicles is tractor/trailers. No insurance company would agree with that. And only an utterly mad person would have one of them transit the Grapevine or the Cajon pass in SoCal, among many other locations.

[longship]

73. It's a fucking eight lane Interstate!

Four lanes each direction. This is one of the two main E/W Interstate highways in SoCal. The other is I-15 which goes over the Cajon pass, equally treacherous in the winter.

The thing is, if one is driving east from the SoCal coast there is no avoiding one of these two mountain passes, unless one takes a much more dangerous route.

However, the near six degree grade on the Grapevine is treacherous enough that the entire major Interstate (all eight lanes!) is regularly closed down, mostly in winter. And get this. There is no other way! Other than I-15's Cajon pass, equally treacherous, on the other side of the valley. I guess one needs to decide whether one wants to go to Reno (Grapevine, I-5) or Las Vegas (Cajon, I-15).

Either way, one takes ones chances, but I don't want an autonomous vehicle anywhere near me when I am driving.

[A HERETIC I AM]

85. Well.....technically speaking......

This is one of the two main E/W Interstate highways in SoCal. The other is I-15 which goes over the Cajon pass, equally treacherous in the winter.

Actually, I 5 (The TEJON Pass) and I 15 (The CAJON Pass) are NORTH/SOUTH routes, not East/West( I'll assume you know this, just a brain fart, perhaps!). And there is a way around the Tejon - Ca 14 (The Antelope Freeway, which climbs from 300' elevation to 2600' or so) North to CA 58, then west through Tehachapi and into Bakersfield, but you still have a long ass way down, not to mention that if it is snowing at the top of the Tejon, it is likely snowing in Tehachapi.

The only way around the Cajon Pass (to get to Las Vegas) is to go all the way to nearly the State line on I 10 to Blythe and take US 95 north through Laughlin, NV and that STILL has a big up and down getting to Needles, CA.

If one wants to go to Reno via I-5, you have two mountain ranges to climb and descend, not just one (The San Bernardino Range and the Sierra's).

(driven those roads more times than I can count)

I agree with the points you are making, Longship. The people who think it will be a cakewalk for the software engineers to simply make it so, are fooling themselves. Entirely too many variables need to be addressed before this technology becomes as mainstream as, say...anti-lock brakes.

As I said, I'm a trucker and even though they have technology demonstrators on the road that could possibly take my job away, I have no fear. Truck drivers are going NO WHERE for quite a long time until every issue I mention in post 84 is fully addressed.

[longship]

91. I have driven the Cajon pass a few times.

And The Grapevine (Tejon pass) once.

Both are rather terrifying to me even in a car. I cannot imagine taking a tractor/trailer over them. That kind of thing sounds scary as Hell. What if you cannot stop? (Well, actually, you can't. Gravity takes control.)

Thank you for your corrections.

And I kind of knew that I-5 and I-15 were N-S having driven them. As you wrote, brain fart. I moved from SoCal some years ago.

[A HERETIC I AM]

93. Engine retarders. ("Jake Brake")

The first time I took a tractor trailer down a long, steep grade I was terrified! Now, grades like the Grapevine are a piece of cake.

It's all about knowing how heavy you are and starting at the top at the appropriate speed for the weight. If the truck is equipped with an engine retarder system (most are these days) then you can go all the way down with barely a touch of the brake pedal.

The guys that have runaways are going too fast for their weight and heat up the brakes to the point they become useless.

But... That's what the runaway truck ramps are for. To save your ass if your stupid quotient exceeds your common sense!

[longship]

95. Thanks!

I figured there was something to make it safe. BTW, both passes are beautiful, just a bit scary, the Grapevine more so in both categories.

[kysrsoze]

76. Audi has already created and ran a fully autonomous racing car. Overall, it's great...

however it's actually a little too careful and doesn't put enough "feel" or drive to win into racing, so it can be beaten by a good driver. But this is just a first attempt...

[Thegonagle]

72. Actually, if the traction and stability control in my car is an indcation,

I react faster. I have tons and tons of practice, but I do it better than the computer. I doubt most drivers can, but I've really developed a feel for the physics of handling a car, especially in the winter. It's kind of like playing sports--with practice, you get really good and your reflexes and coordination develop like you wouldn't believe. I'm not even sure a computer could be taught how to drive through a significant snowfall without getting stuck. My methods definitely involve disabling the traction control though. You need to be able to "power through" rough spots without losing momentum, and traction control can cut the power at the worst possible moment and then you're dead without a shovel.

[kysrsoze]

77. Depends on the car. Some stability control systems have too many nannies or aspects which can't be

defeated. Some stability control systems make you think you can drive like a champ. But yeah, it's better drivers who give car companies a reason have an on/off switch.

[citood]

34. Cool race

You should check out the solar car challenge...been going on for 20 years. But I have no illusion we are anywhere close to a commercially viable solar car on the highway. Same goes for concrete canoe races and other novelty engineering challenges

[Egnever]

38. The solar car challenge is cool.

Not sure why you would think we are a long way off. Some of these look pretty viable already.

And the challenge in Australia should prove even more interesting as they have classes for the cars including a cruiser class that is a two seater. The last one in 2013 had some pretty amazing cars. The next one is only a month away it will be interesting to see how far they have come.

[citood]

86. Let me count the ways

"Not sure why you would think we are a long way off. Some of these look pretty viable already."

1) Each car is being followed by a support vehicle - not real life
2) Most of these cars are too low to have a legal bumper height
3) In real life, most prefer room for passengers
4) These cars are so light, they likely could only be classified as motorcycles
5) A commercially viable automobile would have 300-400 lb of airbags, crumple zones, and crash bars added
6) These things would get stuck in 2" of snow
7) The sun doesn't always shine

You think these are viable - to me its incredibly obvious that they are several decades off, and would require a quantum leap in solar technology. This is how we look at technology differently. Yes, amazing things are possible...but not in a way that would be realistic for mass use. This is why I don't fly a space shuttle to work or go whitewater rafting in a concrete canoe. These are demonstration projects - nothing more.

[citood]

13. Too many contingencies

The other day I was in a construction zone in an urban area, where 4 lanes had been reduced to two. It set up a situation where I had to pay attention to traffic signals ten feet to my right, instead of above me. A self driving car could not do that - just one of several thousand scenarios I can think of that would confuse a self driving car.

[metroins]

14. Yes they can

Rudementary Autoit scripts have been able to perform that task 15 years ago.

But to make it simple for you, require all traffic signals to be retrofitted with a cheap transponder.

Vehicles will communicate with other vehicles and signals. The autonomous car would know about the signal before arriving.

https://www.technologyreview.com/s/534981/car-to-car-communication/

Mercedes has already implemented the technology.

[citood]

16. You don't understand how these cars 'read' traffic signals

The car has to expect the signal...ie know its there due to pre-mapping of the route. In a scenario as I described, the software will not know to even look.

Now I view your notion of requiring transponders to be a cardinal sin - if self driving cars cannot handle the road as it exists, we should not be spending billions of dollars retrofitting the road to accomodate self driving cars.

Twenty five years ago, tests were done based on the idea of placing a wire in the road...an electric 'track' concept that is used today on factory floors, but was never realistic for the road. So I have to ask - how is outfitting every traffic control device with a signaling device any more practical?

If I'm lucky, I've got 40 years left on this marble - and I'm positive I'll never see a fully autonomous car on the consumer market.

[metroins]

22. They are already here

The Google car drove 1.3 million miles without an accident. A typical human won't make it 60k without some kind of contact. The Google cars accident was at 2mph.

Volvo and Mercedes currently have autonomous vehicles that operate in cities. Consumers can purchase them.

Uber is launching autonomous taxis in Pittsburgh.

https://www.google.com/amp/s/www.washingtonpost.com/amphtml/business/inside-pittsburghs-drivers-seat-move-on-ubers-autonomous-car-testing/2016/09/02/c5d2462a-6fc7-11e6-9705-23e51a2f424d_story.html?client=ms-android-sprint-us

Ford will have a full vehicle by 2020.

By 2030, 40% of vehicles are expected to be autonomous.

[citood]

28. You do understand that a platoon of Google street view cars pre-mapped

The routes withe cameras and LIDAR, and a squad of data crunchers extracted the data from the point clouds over a period of weeks, before the 'self driving' car could drive, right?

That's not real life.

Now let me pose a question - if we can both agree that a completely autonomous car has to be self driving for the ENTIRE route (heck Google is bragging about removing the steering wheels): What about the areae Google can't pre-map, much less make regular re-maps to detect any changes such as construction, potholes, new signs, signs removed, etc. I recommend you go to Google street view mode and check out all the places that the Google car does not go - apartment complexes, some university campus roads, commercial properties, parking garages, military bases. I used to live in an apartment complex in Manhattan, Kansas and commute to Fort Riley daily...and I would occassionally go to the mall or the grocery store. None of these were places the Google car can even map - which meane no-go territory for a self driving car.

Its not already happening. You are falling for the hype. The day I can use a dart board to randomly find a start and stop point for a route, and launch a car with no pre-mapping - that is the day we really have self driving cars. What Google has done to date is only marginally more impressive than the robot train at the airport.

[metroins]

40. The places you mention

Are already autonomous.

Audis, volvos and bmws already navigate parking lots autonomously.

The Google car is also separate from Google maps.

You're using words in the industry, but you're failing to see how advanced the technology is. If you really want to see where it's headed, read Allstates report on the impact of the industry.

http://www.wsj.com/articles/driverless-cars-threaten-to-crash-insurers-earnings-1469542958


We're talking consumers in 5 years, market shift in 15.

[citood]

46. So a Google car...or Audi etc could drive around Ft Riley

Tommorrow?

I don't think so.

And I didn't say all parking lots are off limits...but if all parking lots aren't available, there may as well be none available if this is going to be commercially viable. I pointed out where the Google car can't go as an example of how many private and government owned properties are not available to be mapped...and yes these vehicles need a map to do more than self park.

[metroins]

51. Autonomous cars can go where not mapped

You keep relying on maps....they don't need maps.

Tesla vehicles don't use maps, most of them don't...

I don't know why you keep harping on maps.

[citood]

59. Tesla doesn't believe their cars are autonomous

So why should you?

Tesla can stay in a lane on the highway...which is around 5 percent of driving. Tesla has zilch to do with autonomous vehicles.

[Egnever]

41. They are mapping the streets already

In real life. Using the data from google maps and waze that people use every single day.

As far as streets signs and construction that is what the sensors are for they don't rely entirely on maps if they did they would be hitting pedestrians every day.

[citood]

48. Look at where the Google car doesn't go

Gated communities, commercial parking lots, mitary installations. And no, Waze won't cut it.

[Egnever]

58. Sure it will

People in gated communities use waze and google maps as well. But they don't just rely on maps to do the driving. That would cause all sorts of issues it is just one of many data points.

Aside from that Google has most gated communities mapped they just don't have street view.

[citood]

62. You misunderstand my use if the term 'map'

The Google project is sending at least five LIDAR laden cars down a route and creating a 3 dimensional view...and then software is used to identify and map the 3 dimensional location of every stripe, pothole, sign, traffic signal; blind spot, crosswalk, etc. It creates a 3 dimensional virtual 'track' for the car. Same way a locomotive can't move without the track, neither can the Google car. Waze doesn't do anything like that.

[Egnever]

65. Something that is easily remedied

when people can't drive their driverless BMW in an area you can be sure there will be a clamour to finish that mapping. Also don't forget once the cars are on the streets they become continuous mapping machines including all of the data you mentioned above.

I have no doubt the first driverless production vehicles will also have the option for humans to take over in areas the car will not work for one reason or another. That wont stop them from being on the road. Every car from then on will be helping to update those maps.

[citood]

67. The company I work for is mapping

For a completely different purpose. There us no room for passengers after installing the equipment, air conditioners, and computers. This is more hi tech than Waze...so no, your average passenger car will not be mapping...thats before we even think about what happens if somebody using crowdsourcing like that deliberately screwed up the map like a joke wikipedia entry.

And no - the first autonomous vehicles won't have backup drivers because that isnt autonomous at all. Vehicles with self driving functions have backup drivers. The difference between the two is as wide as the grand canyon.

[Egnever]

68. So now we are going to be arguing semantics.

they aren't truly autonomous cars...

ok I will agree with you then we wont see fully autonomous vehicles likely for a while. It won't be because we can't it will becuase of legislation and the public's slow movement to trusting them.

Semi autonomous vehicles will be here very soon though and for most uses will perform the same as a fully autonomous vehicle would.

[citood]

69. A little more than semantics

Google is suggesting the steering wheels will come out of cars...not me.

[Egnever]

70. They will

but you are right it will take time because of public trust issues not because they can't

[citood]

71. At least 50 years

If it were really feasible, the public trust issue would evaporate. But it isn't feasible.

[longship]

25. There are no fucking signals in the national forest.

We don't even have paved roads here, let alone two lanes. And there's winter here, when the roadways utterly disappear under drifts. That's when you want intelligence behind the wheel.

[NutmegYankee]

54. You're not going to convince non-technical people.

Technology is magic to them and always will be.

[Egnever]

18. Yup this is no different than Ted Stevens calling the internet a series of tubes.

He has no clue what he is talking about.

[kysrsoze]

75. IBM exec from the '40's: "I think there is a world market for maybe five computers."

This predictor doesn't have a clue. Like it or not, driverless cars are coming. Technology on this isn't quite there yet, but it's coming. Insurance companies are already considering the impact to their bottom lines of cars which don't crash as often, and fewer drivers due to shared rides.

I guess I'm OK with is so long as we always have the option to take over and do it ourselves. I wouldn't mind throwing the auto-pilot on for a long boring stretch, but I love to drive and don't ever want to have the option take away.

[Lee-Lee]

81. How will that work until the 50+ years have passed until all the older vehicles are gone?

You can't just snap your fingers and suddenly all vehicles are autonomous and using the same system to communicate, it will be decades before that is even close.

How will the emerging technology work in the many decades where the roads have a mixed fleet on them?

[Orrex]

82. And...

The 80,000 pound truck would also be automated, so it wouldn't be running in the wrong lane in the first place.

I'm not impressed by luddite fear-mongering, and autonomous cars seem to bring out the worst of it.

[Travis_0004]

6. Cars unlikely to ever replace horses

I'm sure somebody said that 110 years ago.

I can see some people who love to drive, and they should have that option, but I would imagine that at least the highways will one day be automated cars only.

[NutmegYankee]

15. Most of your technical trained people like engineers know this.

Airplanes for instance have enough automation to almost fly themselves, but the pilots have to be ready to handle system failures and freak weather events which the automation cannot deal with. No matter how good the computer, the human can observe the situation and still process the information faster as to a course of action. For instance, the A380, the most automated state of the art plane in the sky, had a catastrophic failure of an engine that took out much of the hydraulics and wiring on the left wing. The pilots had to take over and hand fly the crippled plane back to the runway.

Automated Cars may run fine until a system fails, and then that will require human control to deal with the issue and put the car in a safe spot. Cars are subjected to far more abuse and less maintenance then airplanes. Systems are going to fail. Your advanced lane camera is just a field mouse's teeth on the wiring away from failure...

[citood]

19. I am an engineer

And a fairly advanced mechanic. And by happenstance I have dealt with some of the mainstays of autonomous cars such as LIDAR and algorithms to recognize objects in point clouds.

And I absolutely agree with you.

People are letting their imaginations run wild with this...but in reality this is several decades away, if not longer.

[NutmegYankee]

26. I'm also an engineer.

Non-engineers often have this magical view of technology. It's does get annoying at times.

I do wonder how the technology will react to small debris coming adrift. The scariest incident in my life driving was a truck ahead dropped a 4x4 post off its bed and it bounced up to my windshield height. I swerved and it missed my car but if I hadn't it would have been fatal. I had noticed the 4x4 loose on the bed and had dropped back just in case. Would automation do that? I've seem bookcases and couches dropped right in front of cars on crowded highways. It will be interesting to find out.

[citood]

32. I wonder about those of us who don't live in southern California

I live on a dirt road and can't imagine keeping cameras and LIDAR lenses clean...and I have no idea how these systems would handle freezing drizzle, blinding snow, all that crap that flies off tractor trailers in the rain, etc. And I can't get over one hurdle - why? Most people enjoy driving...or at least doj't hate it enough to pay extra for self driving car. There isn't even a market for this technology - at least a consumer market. The company I work for is in fact working with this type of technology for an industrial purpose, and I suspect that is Google's end game as well.

[metroins]

47. Communication

Your car senses it.

Swerves. Communicates with the vehicles next to it which also correct to allow room.

We are already on the way. We're talking 15 years. 15 years ago my jaguar xkr had auto cruise control...now becoming a standard feature on regular vehicles.

2020 Volvo, MB, Tesla and Audi will be quite hands free. It may take another 5 years for others to catch up.

https://www.mercedes-benz.com/en/mercedes-benz/innovation/car-to-x-communication/

[NutmegYankee]

52. Will it swerve, or will it brake?

Will it even sense it? A 4 inch square doesn't reflect back much on radar. And it's not much bigger than leaves dropping from trees in the fall. Will cars swerve to avoid leaves?

[FrodosPet]

35. If you know anything about point clouds, I would like to touch base off the forum

I am currently studying PCL (Point Cloud Library). OK, what I SHOULD be studying instead of hanging out in DU and technical journals reading about all the different facets of auto auto technology and social impact.

I would like to send you some contact information through PM, if that is OK?

[Egnever]

21. Well they should have a chat with all the engineers working on this for the major car manufacturers

most know this

Over a million miles driven by google cars on roads where they are the only autonomous vehicle and yet they have only had one at fault accident and it was a 2 mph fender bender.

This guy doesn't know what he is talking about.


The funny thing is in his scenario the prudent choice is to kill yourself instead of the 18 people but the human will make the wrong choice in that scenario nearly every single time. Somehow he thinks that is a preferable scenario. I am betting when most people are confronted with that scenario without being in that scenario they would chose the correct choice but when they were actually in that scenario survival instincts kick in and they make the wrong one every time.

Surely there will be many that won't trust the cars at first but that will change quickly.

[NutmegYankee]

24. At low speed and in urban areas, it's probably going to do OK.

Out in the rural areas where deer pop out for a suicide stroll with no chance to avoid, incidents will happen. And when the car takes that damage and the sensors gets FUBAR'd, the driver will have to take over.

BTW, major car manufacturers like VOLVO?

http://www.reuters.com/article/us-autos-autonomous-infrastructure-insig-idUSKCN0WX131

Volvo's North American CEO, Lex Kerssemakers, lost his cool as the automaker's semi-autonomous prototype sporadically refused to drive itself during a press event at the Los Angeles Auto Show.

"It can't find the lane markings!" Kerssemakers griped to Mayor Eric Garcetti, who was at the wheel. "You need to paint the bloody roads here!"

Shoddy infrastructure has become a roadblock to the development of self-driving cars, vexing engineers and adding time and cost. Poor markings and uneven signage on the 3 million miles of paved roads in the United States are forcing automakers to develop more sophisticated sensors and maps to compensate, industry executives say.

Tesla CEO Elon Musk recently called the mundane issue of faded lane markings "crazy," complaining they confused his semi-autonomous cars.

I can't wait till it snows...

[Egnever]

31. This is already something they are addressing

Many players are remapping the US because of these kinds of issues. They are even creating mock cities with intentionally hard scenarios to drive. Google is using waze to remap the cities for them with every single user contributing to that mapping.

http://www.oregonlive.com/commuting/index.ssf/2016/08/self-driving_cars_are_increasi.html

On this day, Mcity is more like a ghost town. While the test track is occasionally opened to the public, how it's being used is more often shrouded in secrecy.

It sits behind an 8-foot opaque fence with strict keycard access. The university won't say who tests autonomous vehicles here; only that its leadership circle includes carmakers Ford, General Motors, Honda, Nissan and Toyota, as well as various transportation and technology firms.

The pop-up city is meant to act as an Everytown USA, where vehicles pass through a variety of intersection designs and traffic signals. Buildings can be picked up and moved. Fake tree canopies and a tunnel interfere with GPS signals.

"It's a really sort of compact place where we can create very complicated and interesting scenarios, but in a way that's very controlled and repeatable," said Greg Stevens, Ford Motor Co.'s global manager for automated driving. Ford has publicly acknowledged it's testing self-driving cars at Mcity and elsewhere.

The faux town is built to be as hard or harder to navigate than a real city. Its builders intentionally used faded and graffiti-covered road signs. A small hill is perfectly positioned to hide a stopped car on the other side.

There are certainly challenges to be overcome but they are testing and improving daily. Uber has them on the road in Pittsburgh as we speak. yes they have a driver in the car as well but when they show millions of miles with no driver intervention it will be harder and harder to argue their readiness.

[NutmegYankee]

36. Trust me, I'm no luddite - I design and maintain one of the most complex objects on the planet.

There is stuff that will not be thought of or programmed for that will go wrong. Hardware failures. Mechanical failures. Big potholes that mangle the wheel linkage (it happens every spring to people).

The people who are the most bullish on this stuff tend to be the least technical, because they can't comprehend the complexity. Most people just use technology and don't know how it works. But then again, most technology fails in a safe manner. The phone dies - new phone. Microwave dies - new microwave. Car dies - get repaired.

At the moment, the basic car controls are mechanical, such that a loss of complete control is unlikely. With automation, all it takes is a sensor to send bad data from a hardware fault and it could be catastrophic. And it will happen. Capacitors wear out. Transistors will wear out and fail. Right now, almost every commercial airplane has 1 or more systems that are tagged out/inoperable. A safe car for unlimited road use would need redundant systems, but with that comes cost. Can people really afford a $100,000 car that requires regular maintenance?

[FrodosPet]

50. They don't have a million unrestricted "in the wild" miles though

How many of those miles were on the freeway at 60-70 MPH during rush hour? How many were on dirt roads? How many were during rain or snow storms? How many in fog? How many during fish fly season?

When they get to a billion miles, including at least 50 million miles in fairly common crappy conditions, I might be impressed.

[Egnever]

61. That will happen in the next year.

billions of miles will have been logged by the end of next year with all of the testing going on all over the world. The uber test in pittsburgh will likely do it all by itself in the next year.

[sakabatou]

29. Not now, but in the future, perhaps

[Buns_of_Fire]

49. I figure as soon as all cars can communicate with each other,

they'll probably start talking about us behind our backs. Could murderous roving bands of autonomous taco trucks on every corner be far behind? (From the trailer for Maximum Taco -- coming soon to a theater near you.)

[FrodosPet]

56. Killer Autonomous Taco Trucks?

We need the Trumpinator more than ever to protect us from wave after wave of robots coming up from Mexico, armed with tortillas and carne molida.

[Egnever]

55. All of these folks disagree with him

http://www.theverge.com/2016/8/23/12603624/delphi-mobileye-self-driving-autonomous-car-2019
Delphi and Mobileye are teaming up to build a self-driving system by 2019

Mobileye and Delphi say the CSLP system will be ready for production by 2019, a few years before other major players in the automotive world say they will be ready to roll out their own self-driving cars. Ford said last week that it will build a fully autonomous vehicle without a steering wheel or pedals by 2021. Apple has hinted that its electric car, which may or may not be autonomous, will be ready by 2019. Uber recently announced it would be testing its self-driving technology in Pittsburgh within a few weeks.

Ford plans self-driving car for ride share fleets in 2021
http://www.reuters.com/article/us-ford-autonomous-idUSKCN10R1G1

"We're saying to partners, we are the winning partner. It's not a hollow promise, it's a real intent," Washington said.

Ford Chief Technical Officer Raj Nair said the company likely will not offer a similar driverless car without steering wheel or pedals to consumers until 2025 or later. Launching a self-driving car first for ride-sharing is a better way to reach the mass market and make the cars more affordable, he said.

http://www.wsj.com/articles/gm-executive-credits-silicon-valley-for-accelerating-development-of-self-driving-cars-1462910491

GM Executive Credits Silicon Valley for Accelerating Development of Self-Driving Cars
Head of GM’s foresight and trends unit says timetable for autonomous vehicles likely moved from 2035 to 2020, if not sooner

https://electrek.co/2016/05/12/bmw-electric-autonomous-inext-2021/
This will be followed in 2021 by the BMW i NEXT, our new innovation driver, with autonomous driving


http://www.wired.co.uk/article/toyota-highway-teammate-driverless-car-tokyo

Toyota to launch first driverless car in 2020

Toyota will launch its first autonomous car in 2020. The Japanese auto maker said the Highway Teammate would be available just in time for the Tokyo Olympic Games.

The driverless car is a modified version of its premium-brand Lexus GS and has already been tested on Tokyo's Shuto Expressway. Toyota explained that the car "features equipment that enables automated driving on highways from the on-ramp all the way through to the off-ramp."

http://www.motoring.com.au/next-gen-audi-a8-drives-better-than-you-46963/
Next-gen Audi A8 drives better than you

The next-generation Audi A8 luxury limousine will set a new standard in safety for the German brand.

Due in 2017 the A8 will be Audi’s autonomous motoring pioneer. And according to one Audi executive, it will drive better than most human beings.

Stefan Moser, Audi Head of Product and Technology Communications, said the upcoming A8 will be a fully autonomous, self-driving vehicle that doesn’t get distracted, unlike human drivers.

[FrodosPet]

64. These folks have a financial incentive to disagree

Their goal is to get a product to market as quickly as possible, and let the world be their beta testers.

These are NOT objective opinions, nor are they likely to be based on a quest for safety above profitability.

Believe it or not, I know that SOMEDAY they will be there, saving about 90% of the lives that are currently lost or damaged. That is a wonderful thing. But if they are pushed out too soon, before they are ready, before the infrastructure is ready, and before society is ready, that there will be a backlash that will inhibit their eventual adoption and acceptance.

[struggle4progress]

74. We'll just use Windows Vista as the OS and the cars will be fine!

[FrodosPet]

78. Adds a whole new dimension to the "blue screen of death"

"I'm sorry, but I will be late for work. My car is stuck in the driveway. It would not recognize the brakes after the latest security update to the turn signals"

[Tactical Peek]

79. “If enough people see the machine ...

“If enough people see the machine you won’t have to convince them to architect cities around it. It’ll just happen.” — Steve Jobs

[Exilednight]

83. People said the same the same thing about cars when it came time to

replace the horse and buggy. They said the same thing about planes, or any other new technology that comes along.

I remember a time when the importance of a person in business was how easy it was to communicate with them, now the importance of a person in business is how well they protect themselves from the outside world.

Things advance and people change, it's only a matter of time.

[Calculating]

87. I don't believe we'll see truly self-driving cars for at least 50 years

In other words, I'm not too worried about it right now.

[TheSarcastinator]

88. "Knowest Thou That No Man Shalt Split the Atom!"

-- John Dalton, the "father" of atomic theory, 1766 – 1844

[Xolodno]

89. Get a horse! Rock n Roll won't last!.......

Auto insurers already know the day is coming and premiums for auto insurance will drop like a rock. Plus they will also have the data for "at fault" driver. Self drivers will probably see higher insurance costs.

[FrodosPet]

92. The Segway is the future of personal transportation! No one will walk in the future!

I think I saw one being ridden a few years ago.

[Glassunion]

90. "Fully" being the qualifying word. Yes, we are quite a bit off before driving is completely handed

over to our robot overlords. Much like with planes and rail, you will need a human to monitor the vehicle's operation.

There are many things to consider, especially considering all of the non-autonomous vehicles that will be on the road for years to come. The easier tasks to overcome is the vast majority of interstate travel, as most interstates are easier for these vehicles to traverse. It could be sped along with upgrades to our interstate system. Special lanes of smart road where equipped vehicles could platoon along at higher speeds will be incentive for a lot of folks to invest in purchasing one. But this would require a ton of investment.

The hardest part to overcome is the stupid. These vehicles will need to account for that idiot who jerks across 3 lanes of highway to make their exit at the last second, that moron updating their Facebook while they do a no-look merge, the pedestrian who jumps off the curb mid block without looking to see if anything is coming.

The benefit while we are on our way to fully automated systems, is that these semi-automated systems will reduce accidents and fatalities. It's difficult to quantify, but how many accidents have been prevented by driver assist features that are currently deployed in the market? Traction control, lane departure warnings, auto braking, adaptive cruise control, blind spot monitoring (one of my favs as a motorcyclist), collision avoidance systems, cross-traffic assist, driver monitoring systems, etc... Down the road we are looking at turn signal assistant/warnings, perhaps a universal ignition interlock, curve speed assist/warnings, heads-up displays, night vision, speed regulation systems, etc...