Self-Driving Cars

[dualstow]

@ Hoost, that is cool! I love the roller coaster idea.  Didn't know about kenesins. I love it.

2. NASA's record - well, faulty O-rings are not the programmers' fault. There will still be mechanical difficulties even if the cars are being steered and braked properly by machines.

Forget O rings.  If you include non-manned things, NASA had a string of screw ups in a row e.g. satellights which did not respond, etc.

Hmm, that is troubling. Maybe it's just my personal bias, but I still think we'd be better off than with human drivers.

OTHER IDEAS
------------------
If automated driving proves safe, will we use lighter materials to build cars? What if a tree falls on one of them?
Will thieves find it easier to rip into a lightweight car to steal that pp gold you were just going to sell to the dealer?
Will naughty kids overturn light cars just because they can?

[Tyler]

1. Human override -- this could definitely be a problem if it's allowed. Yes, a lot of airliners have crashed because the human pilots panicked and tried to override the (correctly flying) autopilot.

Legally speaking, human override will be required for quite some time. States and cities will allow self driving cars on highways, but are hesitant to take away human control in dense areas until the technology is proven.

And practically speaking, I imagine self driving technology is still far away from being able to navigate things like parking garages and dense city areas full of pedestrians. If a car will always patiently wait for someone to cross a street, cities like San Francisco and New York could be completely shut down by pedestrians who ignore crosswalks with no fear of repercussions. How this affects infrastructure and behavior patterns is still unknown, and there will be tradeoffs.

I'm interested in how cities will decide to make up for their lost red light camera income. I'm guessing they'll start charging for certain driving routes, which will distort the utopian hyper-efficient routing vision. Things could get worse before they get better.

[l82start]

Not sure if this is what you had in mind or not, but reading your post I was reminded of this article and one of the ideas featured in it.

http://arstechnica.com/cars/2014/06/ide ... r-commute/

The team from Berlin had a similar type of idea, although it doesn't seem any hard details are fleshed out.

Berlin's plan took inspiration from the microscopic transport networks within our cells that move proteins around. In this system, the "cargo" gets moved by proteins called kinesins that travel along structures called microtubules. The kinesins can be coupled and decoupled from their cargo, and team Berlin’s idea, therefore, is to decouple the passenger cabin of future cars from the bits that handle propulsion. It also wanted to take a leaf from the way ants communicate with pheromone trails that reinforce with lots of traffic, replacing the ant pheromones with data culled from Foursquare.

That quote is toward the bottom of the page.  Sounds like they're supposed to vote on a winner this fall, so maybe there will be more to it coming soon.

its not the same source as i saw but the idea is very similar, self driving cars are going to have to exchange a full spectrum of information to each other anyway, so most of the information necessary for coupling and uncoupling will already be shared by self driving cars. the advantage seems to be coupling them together takes up far less road space = less congestion and offers aerodynamic, shared load, lower fuel use  (and possibly safety) opportunities..
it sounds like it is still a blue sky idea but an interesting one

[Libertarian666]

Not sure if this is what you had in mind or not, but reading your post I was reminded of this article and one of the ideas featured in it.

http://arstechnica.com/cars/2014/06/ide ... r-commute/

The team from Berlin had a similar type of idea, although it doesn't seem any hard details are fleshed out.

Berlin's plan took inspiration from the microscopic transport networks within our cells that move proteins around. In this system, the "cargo" gets moved by proteins called kinesins that travel along structures called microtubules. The kinesins can be coupled and decoupled from their cargo, and team Berlin’s idea, therefore, is to decouple the passenger cabin of future cars from the bits that handle propulsion. It also wanted to take a leaf from the way ants communicate with pheromone trails that reinforce with lots of traffic, replacing the ant pheromones with data culled from Foursquare.

That quote is toward the bottom of the page.  Sounds like they're supposed to vote on a winner this fall, so maybe there will be more to it coming soon.

its not the same source as i saw but the idea is very similar, self driving cars are going to have to exchange a full spectrum of information to each other anyway, so most of the information necessary for coupling and uncoupling will already be shared by self driving cars. the advantage seems to be coupling them together takes up far less road space = less congestion and offers aerodynamic, shared load, lower fuel use  (and possibly safety) opportunities..
it sounds like it is still a blue sky idea but an interesting one

They don't have to be touching. They will still get fuel economy advantages from "drafting", as they do in racing, but without having to worry about colliding.

[Xan]

You're safer from a collision if you're actually touching the car in front of and behind you than if you're hovering nearby (say, in draft range).

[l82start]

They don't have to be touching. They will still get fuel economy advantages from "drafting", as they do in racing, but without having to worry about colliding.

i agree the whole connect/disconnect part of this seems pretty sketchy at best, i cant quite picture all the factors involved but on the other hand i have done plenty of drafting on a bike and this
seems a lot safer than this

[dualstow]

Legally speaking, human override will be required for quite some time. States and cities will allow self driving cars on highways, but are hesitant to take away human control in dense areas until the technology is proven.

And practically speaking, I imagine self driving technology is still far away from being able to navigate things like parking garages and dense city areas full of pedestrians. If a car will always patiently wait for someone to cross a street, cities like San Francisco and New York could be completely shut down by pedestrians who ignore crosswalks with no fear of repercussions. How this affects infrastructure and behavior patterns is still unknown, and there will be tradeoffs.

Yeah, as you hinted, the cars can already handle pedestrians, but pedestrian behavior is already a problem.
As for parking garages, and also poorly mapped rural roads and places where GPS may be blocked, I have seen articles about antennae that broadcast local location.

Yes, it all seems like a lot of work, but once it's done...

I'm interested in how cities will decide to make up for their lost red light camera income. I'm guessing they'll start charging for certain driving routes, which will distort the utopian hyper-efficient routing vision. Things could get worse before they get better.

Red lights, speeding, parking and all other violation income!
Answer: more taxes.

[Libertarian666]

They don't have to be touching. They will still get fuel economy advantages from "drafting", as they do in racing, but without having to worry about colliding.

i agree the whole connect/disconnect part of this seems pretty sketchy at best, i cant quite picture all the factors involved but on the other hand i have done plenty of drafting on a bike and this
seems a lot safer than this

Not if at least the trailing vehicle is computer-controlled.

[l82start]

Not if at least the trailing vehicle is computer-controlled.

you could be right, the difference between coupled vehicles acting as a unit and uncoupled ones acting as a unit may be negligible, my gut feeling that the former is a more stable and safer system, is likely influenced by how hard it is for a human to have the necessary reflexes to make those split second decisions, there is a reason drafting (aka tailgating) is not legal on the streets and why even the best NASCAR drivers end up in wrecks when doing it on a closed track with other pros.
it would take somebody with a better physics/engineering background than i have plus an understanding of the types of systems that would be used to make a strong case for connecting VS drafting, i really cant picture all the factors involved, but i can see there would be a lot to take into account...

[Libertarian666]

Not if at least the trailing vehicle is computer-controlled.

you could be right, the difference between coupled vehicles acting as a unit and uncoupled ones acting as a unit may be negligible, my gut feeling that the former is a more stable and safer system, is likely influenced by how hard it is for a human to have the necessary reflexes to make those split second decisions, there is a reason drafting (aka tailgating) is not legal on the streets and why even the best NASCAR drivers end up in wrecks when doing it on a closed track with other pros.
it would take somebody with a better physics/engineering background than i have plus an understanding of the types of systems that would be used to make a strong case for connecting VS drafting, i really cant picture all the factors involved, but i can see there would be a lot to take into account...

The main problem with drafting when the vehicle is controlled by a human is... human reflexes! This is not a problem with a computer-controlled vehicle.

[Xan]

The main problem with drafting when the vehicle is controlled by a human is... human reflexes! This is not a problem with a computer-controlled vehicle.

I think the main problem is stopping distance.  Reflexes play a part, yes, but if you're going 80 and the car 25 feet in front of you has an axle break or something, you're going to slam into it really, really hard, no matter how computer-controlled your brakes are.

[hljockey]

The main thing I wonder about is how the computer could know certain things, like the fact that there is a big curve of X degrees up ahead and I therefore need to slow down to Y miles per hour in Z feet. I'm sure they have some technology that does this somehow but I can't help but wonder how foolproof it is. Presumably it's been tested on mountain roads.

Back to my own original question - does anybody have a guess how they follow the contour of the roads? My best guess would be GPS but although I think GPS is an amazing contraption it's only as good as the maps and has led me astray more times than once when the maps don't reflect reality. Follow the cars in front of you? Like our parents used to say if someone jumps off a cliff are you going to follow? The street signs? What if they have been taken down? Follow the white line on the side - what if it isn't there?

Really curious if someone has the answer.

[Libertarian666]

The main problem with drafting when the vehicle is controlled by a human is... human reflexes! This is not a problem with a computer-controlled vehicle.

I think the main problem is stopping distance.  Reflexes play a part, yes, but if you're going 80 and the car 25 feet in front of you has an axle break or something, you're going to slam into it really, really hard, no matter how computer-controlled your brakes are.

Yes, that is a possibility but I think the number of such incidents is miniscule. Very few vehicle failures are catastrophic these days. Even though I leave a LOT of room in front of me, I imagine I'd feel much safer with a computer-controlled car drafting the car in front of me than driving manually with my (ab-)normal safe distance.

[Mark Leavy]

Yes, that is a possibility but I think the number of such incidents is miniscule. Very few vehicle failures are catastrophic these days. Even though I leave a LOT of room in front of me, I imagine I'd feel much safer with a computer-controlled car drafting the car in front of me than driving manually with my (ab-)normal safe distance.

I see this type of reasoning all of the time and it makes me cringe.

When designing a huge national system, the question isn't "how probable is this event?", it is "what are the consequences of this event?"

Failures WILL happen.  A tire will blow out.  Some drunk teenagers will string a 1/8" steel cable across the road.  A rock will tumble down a hillside.  Whatever...

Then what happens to the string of lightweight, fuel efficient cars going 80mph, bumper to bumper?

Personally, I like systems with frequent small failures, as opposed to systems with infrequent gigantic failures.

[dualstow]

...
When designing a huge national system, the question isn't "how probable is this event?", it is "what are the consequences of this event?"
...
Then what happens to the string of lightweight, fuel efficient cars going 80mph, bumper to bumper?
...
Personally, I like systems with frequent small failures, as opposed to systems with infrequent gigantic failures.

If you ride Amtrak, you can enjoy both.

[Mountaineer]

...
When designing a huge national system, the question isn't "how probable is this event?", it is "what are the consequences of this event?"
...
Then what happens to the string of lightweight, fuel efficient cars going 80mph, bumper to bumper?
...
Personally, I like systems with frequent small failures, as opposed to systems with infrequent gigantic failures.

If you ride Amtrak, you can enjoy both.

Who is it that is in charge of Amtrak?  Who is in charge of the highway system?  Who is in charge of the Post Office?  Who is in charge of the Veterans Administration?  Who gave the "stand down" orders for Benghazi?  Who is it that says Islam is a peaceful religion?  Who is in charge of border security?  Etc. 

And we want more of that in charge of a self-driving car system?  R I G H T !  Go for it.  It will improve the gene pool.

... Mountaineer

[hoost]

The main thing I wonder about is how the computer could know certain things, like the fact that there is a big curve of X degrees up ahead and I therefore need to slow down to Y miles per hour in Z feet. I'm sure they have some technology that does this somehow but I can't help but wonder how foolproof it is. Presumably it's been tested on mountain roads.

Back to my own original question - does anybody have a guess how they follow the contour of the roads? My best guess would be GPS but although I think GPS is an amazing contraption it's only as good as the maps and has led me astray more times than once when the maps don't reflect reality. Follow the cars in front of you? Like our parents used to say if someone jumps off a cliff are you going to follow? The street signs? What if they have been taken down? Follow the white line on the side - what if it isn't there?

Really curious if someone has the answer.

I believe the google system uses a combination of GPS and lasers which create a high definition 3D model of the environment.

Here's a good video:  https://www.youtube.com/watch?v=bDOnn0-4Nq8

[Benko]

So all maps for the computer controlled cars would have to be upgraded perfectly and timely every time there is road construction or a detour?  What could possibly go wrong?

[dualstow]

I'm pretty sure Mark was not advocating for any kind of government-run self-driving car netwrok.

I think his point is that a network of very close following cars is one that is primed for catastrophic failure, and he would prefer a system that is instead designed to have small failures.

That's what I got out of it, too, although I don't think a move to automation will necessarily mean a move to 80mph speeds. Perhaps speed will be determined by reaction time + braking technology.

Right now, we have human tailgaters who drive too closely, too fast, and who can't be bothered to look away from their phones. If that's what a lot of small failures means, I'll take the occasional catastrophic train-like accident.

[Libertarian666]

Yes, that is a possibility but I think the number of such incidents is miniscule. Very few vehicle failures are catastrophic these days. Even though I leave a LOT of room in front of me, I imagine I'd feel much safer with a computer-controlled car drafting the car in front of me than driving manually with my (ab-)normal safe distance.

I see this type of reasoning all of the time and it makes me cringe.

When designing a huge national system, the question isn't "how probable is this event?", it is "what are the consequences of this event?"

Failures WILL happen.  A tire will blow out.  Some drunk teenagers will string a 1/8" steel cable across the road.  A rock will tumble down a hillside.  Whatever...

Then what happens to the string of lightweight, fuel efficient cars going 80mph, bumper to bumper?

Personally, I like systems with frequent small failures, as opposed to systems with infrequent gigantic failures.

In 2012, the last year for which statistics are available, we had 33,561 highway fatalities in the US. That seems to leave a lot of room for improvement.

If you can make a case that self-driving cars would be more hazardous than that, I'm all ears.

[dualstow]

I love that I'm in a thread in which I agree wholheartedly with Libertarian666.

In 2012, the last year for which statistics are available, we had 33,561 highway fatalities in the US. That seems to leave a lot of room for improvement.

If you can make a case that self-driving cars would be more hazardous than that, I'm all ears.

The big problem from self-driving cars is going to be this: how do we handle the overpopulation problem now that all these lives are being saved? How do we thin the herd? (Sorry, I've been watching that UK show, Utopia).

[Libertarian666]

I love that I'm in a thread in which I agree wholheartedly with Libertarian666.

In 2012, the last year for which statistics are available, we had 33,561 highway fatalities in the US. That seems to leave a lot of room for improvement.

If you can make a case that self-driving cars would be more hazardous than that, I'm all ears.

The big problem from self-driving cars is going to be this: how do we handle the overpopulation problem now that all these lives are being saved. How do we thin the herd?

Um, I think that a reduction in yearly deaths amounting to 0.01% of the population (if we completely eliminate traffic deaths) is not going to cause too much overpopulation. 

[Pointedstick]

Self-driving cars should be great for mechanics! Cars are already so complicated that doing your own maintenance and repair is all but impossible. As cars become ever more complex and computerized, I expect the trend to only continue. Expect routine repair bills of four figures or more…

On the other hand, maybe cars will become so smart that they'll largely maintain themselves until something so catastrophic happens that they have to be towed back to the shop. That's sort of the direction we're moving in anyways.

[Pointedstick]

Nah, computers are still repaired by people. I don't expect that cars will be any different. Of course the people who repair self-driving cars may start to resemble Apple Store Genius Bar people more than grease monkeys.

[Mountaineer]

If you ride Amtrak, you can enjoy both.

Who is it that is in charge of Amtrak?  Who is in charge of the highway system?  Who is in charge of the Post Office?  Who is in charge of the Veterans Administration?  Who gave the "stand down" orders for Benghazi?  Who is it that says Islam is a peaceful religion?  Who is in charge of border security?  Etc. 

And we want more of that in charge of a self-driving car system?  R I G H T !  Go for it.  It will improve the gene pool.

... Mountaineer

I'm pretty sure Mark was not advocating for any kind of government-run self-driving car netwrok.

I think his point is that a network of very close following cars is one that is primed for catastrophic failure, and he would prefer a system that is instead designed to have small failures.

I agree with what Mark said wholeheartedly; his comment is my experience too ... many leave out the consequence part and focus only on the probability part of risk assessment.  I was just looking forward to who will most likely run a nationwide system of self-driving vehicles ... I doubt it would be private industry, mainly because of the potential for "put you out of business" lawsuits.  Thus, it will fall to those who pretty much are not capable of running much of anything successfully, i.e. profitably and competently.  Thus, more tax burden yada, yada, yada. 

... Mountaineer