Self Driving Cars Having Issues With Poor US Roads

[hotdogPi]

At least in the future, in theory, cars should know where parking is permitted and not permitted, and one won't receive a parking ticket in the first place.  A better Self Driving Car version will automatically move itself when a parking restriction goes into effect!

Then you try to find your car... and you can't find it, because it moved.

[jeffandnicole]

At least in the future, in theory, cars should know where parking is permitted and not permitted, and one won't receive a parking ticket in the first place.  A better Self Driving Car version will automatically move itself when a parking restriction goes into effect!

Then you try to find your car... and you can't find it, because it moved.

There's an app for that!

[Sykotyk]

Vehicle-to-vehicle radio communication protocols are being developed....and the rules of the road the automated vehicles now on the road follow are pretty clear.

But, as I questioned: which one veers? If both cars think they're on the road and following the proper lanes, if one car deviates across the yellow line, which car yields? Both cars assume their course is correct. Sure, braking may solve the problem, but not at speeds and proximity where it's impossible. Aborting the perceived lane may be the only way to avoid an imminent crash, but not if the vehicle must act evasively.

Again, which one veers?

Presumably, the inter-automated-vehicle communications would include a broadcast saying "I'm here, and plan to do X, Y, and Z".  When the two vehicles pick up each other's broadcasts (or otherwise detect one another), they do what they must to pass on the correct side of one another, while staying on the road.  If there's insufficient space, the vehicle projected to arrive at the chokepoint later yields the right-of-way, either stopping or otherwise evading....just as should happen if driving laws are obeyed.

Alternatively....what's the opposite of a game of "chicken"? 

That's part of the problem, though. 'Staying on the road' would require both cars to know where the road is. That's the problem. A self-driving car reads the road ahead of them, not just following GPS coordinates or implanted beacons in or around the road. If it fails to determine where the road is, it can deviate. Both cars think they're on the road. At 55mph, head on, that's 110mph of overall speed for not one, but two cars, to talk to eachother, establish they are both cars (can't wait until people can spoof a car's signal), announce where it's headed based on what it sees of the road ahead, and determine that both cars will clear eachother based on those observations.

Call me skeptical, but I think it's a lot of data processing and communication in a very short window. And again, if both cars think they're on the right heading, which one would purposefully drive off the road it thinks its on? Or, would both apply the brakes 100%, slamming all unsuspecting passengers hard (unsuspecting because, with self-driving, who will be paying attention out of the windshield, even they still had them someday).

A self-driving car REQUIRES that it is reading the road ahead of them. Otherwise, it's not self-driving. It's simply following points on a map or GPS. It's supposed to see and navigate the left turn up ahead, not just know it's there because a GPS software tells them it's there. That's the issue. Perception, communication, announcement of intentions, and calculating how to avoid eachother.

Most of the self-driving cars I've read about drive at a very slow pace to avoid those problems.  I don't think people will see the benefit in car travel if it's at 15mph, or 20mph, or 30mph. People have accepted the risk inherent in automobiles back in the 50s when cars were heavier, deadlier, more powerfully propelled, and less safe (no seatbelts or airbags), and yet people didn't really shy away from getting in the car to go for a ride.

We've accepted a level of a safety we as a society have accepted. Why do you think we keep upping speed limits? If we wanted zero-sum deaths from vehicles, we'd be going in the opposite direction. But, as a society we've accepted that some are okay. And maybe a few more, potentially, or more serious, is okay in exchange for a bit of time saved while traveling.

The benefit espoused by self-aware cars is that it eliminates or vastly reduces fatalities. But, society has already deemed the current level of fatalities acceptable. Or else we wouldn't have ever allowed cars to be driven the way we do now.

[Sykotyk]

At least in the future, in theory, cars should know where parking is permitted and not permitted, and one won't receive a parking ticket in the first place.  A better Self Driving Car version will automatically move itself when a parking restriction goes into effect!

Then you try to find your car... and you can't find it, because it moved.

One hypothetical is that as self-driving cars become more ubiquitous, cars would behave differently. Instead of 'parking' your car, you would simply be let off a a predetermined point and the car would drive itself to a holding area for cars nearby. Then, when you're ready to leave the store, office, etc, you simply signal your car via internet/phone/pager for it to return to pick you up. No more parking garages for every building, or small lots or street parking. The cars would hide themselves into giant, remote holding areas. Ones that don't even require driveways, as they could be parked bumper to bumper and each one move for one stuck in the middle when it's their time to leave and reassemble back into that lane once that car is removed.

But, then comes step two. No more car ownership. Giant companies such as Uber or Lyft would own the cars themselves. You would simply pay a usage fee. Want to go to work? Arrange an Uber to pick you up and take you to work. Once at work, the trip is over, your cost has been paid, and the car is free for someone else's use. The burden of keeping your car maintained (and the bigger companies will require it to the utmost degree) will make short term borrowing of the car preferable. No matter how long the trip, you would just summon the appropriately sized car. Even scheduling well in advance when you need it (probably at a discount). Houses wouldn't need garages if no one owns the means of transportation.

Then comes the third leg: any form of revolt or uprising will simply disable the means of transportation. No cars. No buses. No trains. No planes. Either by the government, or the counter-regime trying to take over. Nobody has their own vehicle. So, you're basically stuck to your houses out in suburbia. With the grocery store an unacceptable distance away, even that would become obscenely burdensome.

We already know what regimes do with cell service. Self-driving cars would be the next logical step.

[kalvado]

That's part of the problem, though. 'Staying on the road' would require both cars to know where the road is. That's the problem. A self-driving car reads the road ahead of them, not just following GPS coordinates or implanted beacons in or around the road. If it fails to determine where the road is, it can deviate. Both cars think they're on the road. At 55mph, head on, that's 110mph of overall speed for not one, but two cars, to talk to each other, establish they are both cars (can't wait until people can spoof a car's signal), announce where it's headed based on what it sees of the road ahead, and determine that both cars will clear each other based on those observations.

Call me skeptical, but I think it's a lot of data processing and communication in a very short window. And again, if both cars think they're on the right heading, which one would purposefully drive off the road it thinks its on? Or, would both apply the brakes 100%, slamming all unsuspecting passengers hard (unsuspecting because, with self-driving, who will be paying attention out of the windshield, even they still had them someday).

You somehow assume that human drivers are better than computer. For example you think human would be better at knowing where the road is - not always the case on a rainy night, for example. I, for one, had problems locating traffic light loops in snow storm (I needed lane 5 on a 8-lane intersection, and ended up driving through red light - damn thing refused to give me my green arrow. Pavement marks were under 2" of white stuff). If machine cannot see road marks - what are the chances you would see those? With no LIDAR, no IR vision, just old myopic eyes?  Oh, well..
And somehow you assume that humans would be better at resolving the issue. My bet is that human response in imminent danger - such as highway speed head-on - could very well be a ol' good panic.
Equally, you assume that seat belts would go away (as opposed to "no engine start until everyone wears seatbelt"). And no, I don't bother looking through the windshield when I am on a passenger seat, so same situation of unaware passenger is entirely plausible with meaty driver.
Computer cars do have issues - but I am afraid you're arguing against situations when computer totally wins.

[Sykotyk]

You somehow assume that human drivers are better than computer. For example you think human would be better at knowing where the road is - not always the case on a rainy night, for example. I, for one, had problems locating traffic light loops in snow storm (I needed lane 5 on a 8-lane intersection, and ended up driving through red light - damn thing refused to give me my green arrow. Pavement marks were under 2" of white stuff). If machine cannot see road marks - what are the chances you would see those? With no LIDAR, no IR vision, just old myopic eyes?  Oh, well..
And somehow you assume that humans would be better at resolving the issue. My bet is that human response in imminent danger - such as highway speed head-on - could very well be a ol' good panic.

You do seem to forget who will be programming these self-aware cars. It's not robots. It's people making the decision on how, when, why, and where a car will do what it does. That's the problem. All the technology in the world won't correct for differentiating programming. And if two different manufacturers have two different ways of determining the same thing (I seriously doubt the government will create some open-source-like setup), you run into problems.

Maybe one car looks at spatial location first, while the other car looks at speed and trajectory. It has to know where that car first broadcast its location, and know based on processing time how long it was and where it will be. Rather than using imaging software on a live video feed to determine where it is and adjust accordingly. The same as a human does, not how a computer would.

Determining a 'big blue blob' on the video feed from the front bumper growing larger can indicate the speed and approach of an object, and where it's path leads. But, it leaves 'talking to the other car' out of the equation. Just as we do now.

Equally, you assume that seat belts would go away (as opposed to "no engine start until everyone wears seatbelt"). And no, I don't bother looking through the windshield when I am on a passenger seat, so same situation of unaware passenger is entirely plausible with meaty driver.

You assumed that. I assume that harnesses will be required. Though, if fully automated, the safest feature would be to position the passengers with their backs to the front of the car to avoid whiplash.

Computer cars do have issues - but I am afraid you're arguing against situations when computer totally wins.

I'm playing devils advocate to most.

But, self-aware cars still haven't answered the primary issue: control.

You're content to scan your items at a self-checkout because you have control. You let someone else scan them while you stand and watch because, again, you have control. The biggest issue with ordering delivery groceries is that you can't see the food, or really get a sense of the pricing. You also don't know the quality that you're getting. At the store, you can look at that tub of strawberries and pick the one you want. At the auto-store where it'll be delivered to you either requires a lot of shrinkage, or they'll be shipping the bad produce to you (peapod has this issue).

The problem is control. You ride an elevator not because you trust the machine, but because in the event of failure it's gravity-driven brakes will lock the lift into it's position within a few feet. Even if only one of the brakes functions). But you'd be far less likely to trust an elevator whose computer tries to 'fix the problem' while the elevator is falling to the lobby. You're okay with a little loss of control in exchange for a near iron-clad safeguard against failure == death.

An auto-driving car, either with you the passenger or you the pedestrian, takes 'control' and puts it into a system who will calculate what is the most viable outcome given the situation. If a person driving a car swerves to avoid another vehicle and hits a pedestrian, as a society, we've accepted that. Because someone, somewhere, is in control of the decision. Even if you, the injured party, wasn't at the time. You accept that if you were in that position, you, too, would want to be in control.

A computer would not give anybody any control over the decision. Even if the calculated outcome is deemed to be better for society at large. As a society we've accepted blame, responsibility, negligence, and control to be part of the human experience. Even if it doesn't work out as well as a computer does.

As I mentioned before upthread: if you were to take a pair of scissors and cut a pattern out of a piece of paper, you'd probably feel perfectly comfortable doing so. But, if a computer were to cut out the sheet of paper while you held it, you'd probably be far less likely to participate.

It's a matter of control.

[vdeane]

You somehow assume that human drivers are better than computer. For example you think human would be better at knowing where the road is - not always the case on a rainy night, for example. I, for one, had problems locating traffic light loops in snow storm (I needed lane 5 on a 8-lane intersection, and ended up driving through red light - damn thing refused to give me my green arrow. Pavement marks were under 2" of white stuff). If machine cannot see road marks - what are the chances you would see those? With no LIDAR, no IR vision, just old myopic eyes?  Oh, well..
And somehow you assume that humans would be better at resolving the issue. My bet is that human response in imminent danger - such as highway speed head-on - could very well be a ol' good panic.
Equally, you assume that seat belts would go away (as opposed to "no engine start until everyone wears seatbelt"). And no, I don't bother looking through the windshield when I am on a passenger seat, so same situation of unaware passenger is entirely plausible with meaty driver.
Computer cars do have issues - but I am afraid you're arguing against situations when computer totally wins.

Assuming the car even looks at pavement lines in the first place.  Google has everything mapped, and the cars navigate by those maps.  Restripe the road, and the car won't know about it.  And if its sense of position or the map is even slightly (and, let's get real, Google Maps is more than slightly off) off, it's game over.

One hypothetical is that as self-driving cars become more ubiquitous, cars would behave differently. Instead of 'parking' your car, you would simply be let off a a predetermined point and the car would drive itself to a holding area for cars nearby. Then, when you're ready to leave the store, office, etc, you simply signal your car via internet/phone/pager for it to return to pick you up. No more parking garages for every building, or small lots or street parking. The cars would hide themselves into giant, remote holding areas. Ones that don't even require driveways, as they could be parked bumper to bumper and each one move for one stuck in the middle when it's their time to leave and reassemble back into that lane once that car is removed.

I don't feel like waiting  for the car to come back when I want to leave somewhere.

[US 41]

There's got to be a point at where humans just need to do things themselves and I think driving is one of those areas. If you don't want to drive then live in a big city where you can take public transportation anywhere you want to go. If you want to take a vacation from NYC to Miami then either fly or take a Greyhound. I don't understand what is so hard about this logic. Besides it doesn't take rocket science to drive a car. There's a reason you only have to be 16 years old to get a drivers license.

[Revive 755]

We have dirt roads here. 

And while some of the county dirt roads are pretty scary, I have some access roads on my farm that would make you crap your pants.

I think this is near the Harrison/Monona line:




For some unknown reason, there are always idjits that want to see how soon after a good rain any given dirt road might be passable.  I'm pretty sure no one is going to want to send a self driving car down any of these pre-Columbian trails any time soon.

What does "Level B service" mean?

http://www.ctre.iastate.edu/pubs/itcd/reduced.pdf

Basically less than Level A but higher than Level C.

[SD Mapman]

We have dirt roads here. 

And while some of the county dirt roads are pretty scary, I have some access roads on my farm that would make you crap your pants.

I think this is near the Harrison/Monona line:




For some unknown reason, there are always idjits that want to see how soon after a good rain any given dirt road might be passable.  I'm pretty sure no one is going to want to send a self driving car down any of these pre-Columbian trails any time soon.

What does "Level B service" mean?

http://www.ctre.iastate.edu/pubs/itcd/reduced.pdf

Basically less than Level A but higher than Level C.

Wow, Iowa actually cares about warning people about bad roads! Our "minimum maintenance" signs are for overgrown trails... that road shown could pass for a county arterial West River.

[kalvado]

Assuming the car even looks at pavement lines in the first place.  Google has everything mapped, and the cars navigate by those maps.  Restripe the road, and the car won't know about it.  And if its sense of position or the map is even slightly (and, let's get real, Google Maps is more than slightly off) off, it's game over.

Given that discussion started with the article talking about "Poor markings and uneven signage " as problems for the self-driving equipment, I would assume DOTs would not be relieved of road marking hat responsibility any time soon. Accuracy of GPS is not enough for lane keeping anyway, errors associated with radio propagation in ionosphere are pretty tough. Differential GPS may solve the issue, but would create another layer of complexity.

[kalvado]

I don't feel like waiting  for the car to come back when I want to leave somewhere.

Do you feel like waiting for the bus to take you to the parking lot 2 miles away? I've been to such places...

[Brandon]

I don't feel like waiting  for the car to come back when I want to leave somewhere.

Do you feel like waiting for the bus to take you to the parking lot 2 miles away? I've been to such places...

I'm not a fan of it.  Nor am I a fan of using a valet.

[andy]

Assuming the car even looks at pavement lines in the first place.  Google has everything mapped, and the cars navigate by those maps.  Restripe the road, and the car won't know about it.  And if its sense of position or the map is even slightly (and, let's get real, Google Maps is more than slightly off) off, it's game over.

Given that discussion started with the article talking about "Poor markings and uneven signage " as problems for the self-driving equipment, I would assume DOTs would not be relieved of road marking hat responsibility any time soon. Accuracy of GPS is not enough for lane keeping anyway, errors associated with radio propagation in ionosphere are pretty tough. Differential GPS may solve the issue, but would create another layer of complexity.

As I understand it, most driving GPS systems fudge the location to place you on the highway anyway, so we are unaware of the inaccuracy.  For example, parked in my driveway usually places me in the middle of the highway.

I have seen articles about placing magnets or transponders into the pavement as supplements to road sensing. As you stated that is a much higher level of complexity and effort.  I suspect until we get "smart roadways", self driving cars will be severely limited.

[vdeane]

I don't feel like waiting  for the car to come back when I want to leave somewhere.

Do you feel like waiting for the bus to take you to the parking lot 2 miles away? I've been to such places...

I try to avoid such places...

[kkt]

I don't feel like waiting  for the car to come back when I want to leave somewhere.

If you're at a movie, text the car as you get up from your seat and it would probably be there by the time you get to the curb.

http://www.ctre.iastate.edu/pubs/itcd/reduced.pdf
Basically less than Level A but higher than Level C.

Level C being closed to the public and gated.  Why don't they just decommission them as public roads?

[empirestate]

There's got to be a point at where humans just need to do things themselves and I think driving is one of those areas. If you don't want to drive then live in a big city where you can take public transportation anywhere you want to go. If you want to take a vacation from NYC to Miami then either fly or take a Greyhound. I don't understand what is so hard about this logic.

That logic isn't in question; it just won't apply in our hypothetical future the same way it does today. In short, your belief that driving is something we need to do for ourselves, while commonly held today, will be held by far fewer people.