Driverless Technology

[jpd80]

The engineering way to do it is optical recognition (stoplights are well standardized and easy to recognize), but this could all be crowd-sourced, ala Waze's traffic monitoring. A mixture of crowd sourcing and engineering seems like a more likely resolution.

 

A lot of this "no self-driving cars" argument sounds a lot like the "a computer could never beat a grand master at chess" hubris, which was true right up to the point that that computers began reliably and repeatedly beating grand masters.

 

Personally, I don't know if self-driving cars are truly inevitable, but I don't think it'll be the "four way stops" and "stoplights are out" problems that will stop them. Those are engineering problems, and engineers know how to overcome engineering problems. What could very well smother the baby in its bed are the "soft" problems I mentioned above--how does a computer make a choice between two bad outcomes, and who is responsible for it when it does?

No, it will be the stupid things like kids running out from behind parked vehicles,

things that no driver can see and anticipate ahead of time that will be a huge hurdle.

 

While you have fallible humans on the roads and entering places where they shouldn't,

no amount of computer tech will stop those terrible tragedies we read every day,

having a driverless car involved in those issues will just worsen raw emotions.

 

Maybe what we need is places where autonomous works well in say peak hour crawl

where humans suffer frayed nerves, let all the cars talk to each other and sort it out

while drivers sit back and do other things, rather than trying to kill each other..

Edited February 25, 2016 by jpd80

[fordmantpw]

No, it will be the stupid things like kids running out from behind parked vehicles,

things that no driver can see and anticipate ahead of time that will be a huge hurdle.

 

That's actually easier for a computer to solve than some of the things RJ mentions. I'm pretty sure this has already been 'solved.'

 

 

Maybe what we need is places where autonomous works well in say peak hour crawl

where humans suffer frayed nerves, let all the cars talk to each other and sort it out

while drivers sit back and do other things, rather than trying to kill each other..

 

I could see this as a good starting point for autonomous driving. The problem, though, is the shear number of cars that are already on the road. Yeah, this sounds like a great idea if all the cars can talk to each other, but when 99.9% of the cars don't have the technology, they can't talk to each other.

[akirby]

The engineering way to do it is optical recognition (stoplights are well standardized and easy to recognize), but this could all be crowd-sourced, ala Waze's traffic monitoring. A mixture of crowd sourcing and engineering seems like a more likely resolution.

 

A lot of this "no self-driving cars" argument sounds a lot like the "a computer could never beat a grand master at chess" hubris, which was true right up to the point that that computers began reliably and repeatedly beating grand masters.

 

Personally, I don't know if self-driving cars are truly inevitable, but I don't think it'll be the "four way stops" and "stoplights are out" problems that will stop them. Those are engineering problems, and engineers know how to overcome engineering problems. What could very well smother the baby in its bed are the "soft" problems I mentioned above--how does a computer make a choice between two bad outcomes, and who is responsible for it when it does?

 

The problem, as Richard points out, is that it's impossible for the vehicle's computer to get all of the information it really needs reliably. Databases are never 100% accurate. Optics are not 100% reliable especially in bad weather.

 

How would a driverless car handle this scenario? Interstate is blocked by an accident and officers are instructing drivers to make a U-turn and drive the wrong way back to the previous exit and exit using the on-ramp. But you have 2 driverless cars sitting there blocking the others because they can't move.

 

I think we can get 80% of the way there pretty easily, and probably 95% within a few years and maybe even 99% in 10 or 20 years. But I just don't see resolving that other 1% to the point you can have completely driverless cars in all situations.

 

I think it's far more likely to have driver assisted cars that can take over the driving in certain situations (freeways, less congested well marked city streets, etc.) and maybe driverless in certain specific situations.

[Bob Rosadini]

Well I have to say a good thread. And a lot of good points made as to the viability of "driverless" vehicles. I would have to say though, at this point in time I see far more negatives being raised than positives. No doubt we have a lot of technology that can be applied to taking control away from the driver.

 

But again, other than "Monkey see-Monkey do" do we need this at this point? Or do we need this at an accelerated rate?

 

And as for me not owning a car and calling "Uber" so I can go get a few groceries, not everyone lives in NYC!

 

I say there are far more fish to fry than spending a lot of money so I can read a book while driving to or from my job.

 

Speaking of taking the humans out of the element, friend of mines son is an airline pilot. Listening to him, you get the feeling more and more pilots are losing the feel for what it takes to fly! I recently read a story describing a recent crash in which the crew totally fugged things up when they were trying to get control of the airplane back. Scary. Kind of substantiated what my friend was saying,

[akirby]

This is very similar to auto-pilot with a lot of the same benefits. I also remember one of the first Airbus planes flying itself into the woods.

 

The only real "need" would be accident prevention and therefore reduced fatalities from the technologies used, regardless of whether the vehicles are actually driverless or not.

 

Everything else is just gee whiz stuff - but that's what people say they want, especially millenials.

[rmc523]

Speaking of auto pilot, I saw a Model X today:

 

[630land]

One one hand, yes, we all have hand held computers.

 

But then, where are all the flying cars predicted in "The Jetsons" and old Popular Science magazines?

 

"I was late for work, my car had to reboot, and load some updates. Had trouble finding a good signal.

This and will "Big Gubment" control the signals running the cars? Who is going to pay for the infrastructure?

 

Kids just cannot take their eyes off of Tinder and You Tube. If they don't want to learn to drive, will the want to learn to actually hold a job?

Edited February 25, 2016 by 630land

[Biker16]

 

How does the car 'know' there's a signal there in the first place?

 

 

What happens when Google thinks there's a stop sign and there's now a signal?

 

the Sensor can See in the dark, through snow, rain, and can see 360 degrees at all times.

 

 

fast forward to 7:31

Edited February 25, 2016 by Biker16

[Biker16]

Today in NC, I had to drive home in pouring rain through some areas with blackouts, along a road with quite a few switchbacks and Raleigh's wonderfully marked roads (sarcasm level 11).

 

An autonomous car would have at best blown through the blacked-out signalized intersections. At worst, it would have rolled down a hill due to absolutely no way to read already poorly-marked roads sitting under an inch of water.

 

Because of variables like what I drove through in addition to others that have been described, I have no faith in autonomous automobility taking over before papilgee4evaeva the 3rd is born.

 

people said the same thing about automatic elevators.

[RichardJensen]

 

the Sensor can See in the dark

 

being able to 'see in the dark' does not translate into being able to identify what it 'sees' as a malfunctioning traffic light.

[RichardJensen]

 

people said the same thing about automatic elevators.

 

Are you comparing driving a vehicle with moving cabs up and down a fixed shaft?

 

Because if you're going to use the "people said the same thing about..." logic, then the underlying problem sets should be at least plausibly comparable.

 

Otherwise your whole point is vacuous: "Yeah, well, people said the same thing about taming horses." "Yeah, well, people said the same thing about smelting iron." "Yeah, well, people said the same thing about nonstick cookware." "Yeah, well, people said the same thing about selling chocolate covered potato chips."

 

It's empty rhetoric.

[sullynd]

Speaking of auto pilot, I saw a Model X today:

Gah! You should warn people before posting things like that. I can't unsee it.

 

It's the Tesla C-Max.

[fordmantpw]

People said the same thing about flying cars.

 

 

How's that?

[fordmantpw]

Speaking of auto pilot, I saw a Model X today:

 

 

Wow, that thing is hideous!

[RichardJensen]

 

Wow, that thing is hideous!

 

It's like someone inflated a Model S to like 2000 psi.

[SoonerLS]

I think it's far more likely to have driver assisted cars that can take over the driving in certain situations (freeways, less congested well marked city streets, etc.) and maybe driverless in certain specific situations.

I think that is the most likely scenario, with the self-driving systems acting something like an autopilot--you take care of the takeoffs and landings (driving on surface streets), but the car handles the long-distance cruising (highway driving). I just think it's the liability and the "least bad choice" problems that will keep them from getting there more than any engineering limitations.

[papilgee4evaeva]

people said the same thing about automatic elevators.

 

As Richard already said, a fixed guideway is an irrelevant comparison.

[akirby]

 

As Richard already said, a fixed guideway is an irrelevant comparison.

 

Put tracks on every road and it will work perfectly!

[Biker16]

 

being able to 'see in the dark' does not translate into being able to identify what it 'sees' as a malfunctioning traffic light.

 

yes it does, it can see the light and that that light isn't working.

 

 

I am not going to argue with someone who believes they know more than the CEOs of every major and minor auto manufacturer and google.

[papilgee4evaeva]

yes it does, it can see the light and that that light isn't working.

 

 

I am not going to argue with someone who believes they know more than the CEOs of every major and minor auto manufacturer and google.

 

Unless the light gives off some sort of special signal that indicates to the vehicle that it's a traffic light, a sensor won't be able to tell if it's a powered-down traffic light or an overhead sign. Or a tree limb. Or a football.

[RichardJensen]

 

I am not going to argue with someone who believes they know more than the CEOs of every major and minor auto manufacturer and google.

 

The position of these individuals hardly qualifies them to speak to the limitations of AI.

 

I'm qualified.They aren't.

 

It would be better for you to ask why these unqualified people feel free to speak about the possibilities of AI.

[RichardJensen]

 

Or a tree limb

 

That was exactly the situation I was envisioning: Your autonomous car coming to an inexplicable stop because it identifies a tree limb as a defective stop light---

 

And, conversely, what about situations where the signals are themselves obscured? Or where there is no overhead signal (as in many small towns) where the pole mounted signal in a non-functioning state would be interpreted simply as a light post or tree.

[630land]

Again, who's going to pay for all this? And no way will these auto pilot cars cost $20,000.

 

When "know it alls" go on about how "by 2020, we will all be in driverless cars!" I mean really? 4 years?

 

And where will these all be parked? What about rural areas? Battery life? Potholes? Poor people?

 

They talk like it will be "Utopia", but we all know this ain't a perfect world

Edited February 26, 2016 by 630land

[fordmantpw]

Again, who's going to pay for all this? And no way will these auto pilot cars cost $20,000.

 

When "know it alls" go on about how "by 2020, we will all be in driverless cars!" I mean really? 4 years?

 

Newsflash...very few new cars cost $20k, so I would not expect an autonomous car to cost $20k either.

 

If (and that's a big if) they could make this all work, I really would not expect the cost of cars to increase that much. $2-3k maybe. The main cost would be in the sensors since the software is essentially free once it's developed. Much of the software development is being paid for by technologies that are in use today, or being added in the next few years (lane keeping, adaptive cruise, etc.).

[Biker16]

 

Unless the light gives off some sort of special signal that indicates to the vehicle that it's a traffic light, a sensor won't be able to tell if it's a powered-down traffic light or an overhead sign. Or a tree limb. Or a football.

 

yes it can discriminate between limb, bird and or Traffic signal

 

 

All this information is being collected by the vehicle.