You get a flying car factory (RAR!)

[Simon_Jester]

The big problem for automated cars on the roads is going to be steering correctly in... intricate places (parking lots, intersections that aren't nice neat crossroads or roundabouts), and avoiding collisions with other vehicles and pedestrians and, potentially, debris. Come to think of it, orange cones and jersey barriers and so on arguably count as 'debris' for that purpose.

Just because your GPS doesn't see it, doesn't mean it's not a problem to drive over it. Lidar makes it physically possible to avoid most obvious obstacles, but... well, the guys driving these things are up to several hundred thousand accident-free miles, but I'm sincerely unsure how well the automation performs in adverse conditions or when driven into the wrong sorts of places.

People are going to get very annoyed if these cars have a habit of driving over traffic spikes that blow out the tires, or disobeying yield signs they can't read and nearly causing accidents.

[Batman]

Have you got any actual evidence for that? Because I, for one, really don't want to die in a fiery aircar crash because some programmer forgot a zero somewhere in the autopilot system architecture. We find incredibly obvious AI errors in games all the time; do you really think the infinitely-more-complex real world will somehow be immune to them?

As a programmer I resent that. AI programing for video games is radically different from the kind of programing we are talking about here. Especially if a lot of the work is done from ground stations as was implied. Bottom line is, no programer is going to be under pressure to make your automobile AI work WORSE than it could so that it poses a challenge to the driver.
This said, I personally do not think a proper flying car AI could be done safely with our modern hardware. But that's another story entirely.

I disagree with both of you. When a game AI sucks, the game sucks, and a company loses money. When an autopilot AI sucks, people die, there's lots of collateral damage, and the company gets the living daylights sued out of it (not to mention possible jail sentences for everyone involved, including management) so they're going to be a LOT more careful about making sure the AI works as intended.
I do think it's doable with modern day hardware. If it's doable without that hardware making up 75+ percent of the vehicle's cost...

[Esquire]

As a programmer I resent that. AI programing for video games is radically different from the kind of programing we are talking about here. Especially if a lot of the work is done from ground stations as was implied. Bottom line is, no programer is going to be under pressure to make your automobile AI work WORSE than it could so that it poses a challenge to the driver.
This said, I personally do not think a proper flying car AI could be done safely with our modern hardware. But that's another story
entirely.

I'll cheerfully admit that I'm not a programming expert and I don't mean to suggest programming video games is the same sort of thing as programming a hypothetical aircar autopilot; my only point is that as far as I know we haven't been able to successfully program any AI that can successfully deal with all the complexities a driver or pilot has to deal with routinely. Obviously, if I'm unaware of some counterexample, I'll retract that statement.

When a game AI sucks, the game sucks, and a company loses money. When an autopilot AI sucks, people die, there's lots of collateral damage, and the company gets the living daylights sued out of it (not to mention possible jail sentences for everyone involved, including management) so they're going to be a LOT more careful about making sure the AI works as intended.
I do think it's doable with modern day hardware. If it's doable without that hardware making up 75+ percent of the vehicle's cost...

Companies do stupid/illegal things all the time, and are sued for them fairly regularly - hasn't stopped the stupid decisions from being made. We can safely assert that commercial autopilots are viable once a commercial autopilot has been demonstrated, not before.*

*Again, not a programming expert, but if a self-flying autopilot were viable with modern technology, don't you think UPS would have built one already?

[Batman]

Err-that would be the part with the autopilot hardware making up 75+ percent of the vehicle's cost? There's doable and there's affordable and economically viable.

[Simon_Jester]

Current state of the art in an automatic car can drive, and on real roads rather than just controlled, carefully organized test tracks.

This does not translate into being flexible, all-weather, or capable of avoiding pedestrians without human supervision, or avoiding small but significant obstacles in the roads. Or responding appropriately to street signs and lane markers in a construction zone.

In some ways an automatic pilot is easier because there's no such thing as a construction zone in the sky. As long as you don't get too close to the ground, nearly anywhere you can physically go is safe, in the "you won't immediately crash and die if you go there" sense. You're not going to run into a cloud stuffed with rocks.

Whereas an automobile is always at most seconds from an accident if it fails to register a relevant obstacle or need to change course.

[Esquire]

Err-that would be the part with the autopilot hardware making up 75+ percent of the vehicle's cost? There's doable and there's affordable and economically viable.

That's a very good point.

Simon - does the above apply to landings as well? I suspect there's at least as much uncertainty in landing a car-sized vehicle as in navigating a construction zone in two dimensions.

[Simon_Jester]

For a hypothetical flying car, not necessarily. You need to confirm that the landing space is flat, but that's about it. Especially if you can have it use its antigravity at partial power to sense whether or not the surface is yielding enough to be a problem (so you don't accidentally land on a canvas awning).

Also, you can reasonably assume that the operator is competent to decide whether or not a place is safe to land on- the real point of the autopilot is not to make the vehicle totally independent, it's to remove human error as a factor in causing the vehicles to crash. If the aircar can sense whether it's landing on a flat surface capable of supporting its own weight, and the pilot isn't asleep at the controls (resulting in the cars landing in the middle of freeways and other suicidal places), that will do the job.

[Purple]

I think I'll take the word of the professional pilot over the AI programmer on this one.

I am not sure what you mean to say by that. Basically it's like Batman said. The hardware required for a functioning AI for something that actually flies it self and dodges other flying stuff in incoming traffic is just going to be very, very expensive. And I am not even talking just about the production costs. Someone has to develop this thing and make sure it develops to match the exponentially growing number of vehicles in motion. And the hardware and software components have to be developed in parallel. And you have to make the entire thing as simple to use as a modern navigation system.

I just don't think we have it in us right now. We can't even make a self driving car right.

I'll cheerfully admit that I'm not a programming expert and I don't mean to suggest programming video games is the same sort of thing as programming a hypothetical aircar autopilot; my only point is that as far as I know we haven't been able to successfully program any AI that can successfully deal with all the complexities a driver or pilot has to deal with routinely. Obviously, if I'm unaware of some counterexample, I'll retract that statement.

Basically I agree with you that it would be very difficult to do. I just disagree with the video game analogy. Video game AI is basically programed to accomplish a very limited set of tasks and do so in a way that is challenging to the player but not overwhelming.

A better analogy for you might be building an autopilot for a fighter jet that can not only fly you to and from battle but also actually fight the battle for you.

In some ways an automatic pilot is easier because there's no such thing as a construction zone in the sky. As long as you don't get too close to the ground, nearly anywhere you can physically go is safe, in the "you won't immediately crash and die if you go there" sense. You're not going to run into a cloud stuffed with rocks.

You do however get a host of other environmental hazards and a vehicle that as per OP can probably pull more G's and certainly fly too high for a human pilot to be comfortable with. So you do have some constraints. Also, clouds full of hail.

But basically my greatest worry is what happens when rush hour arrives and everyone sits in their flying car and starts ascending into the sky all at once from their parking spots. Once you clear the initial rush you can just fly off to a different altitude or something.

[Broomstick]

You're not going to run into a cloud stuffed with rocks.

You can, however, run into a cloud full of aluminum (or carbon fiber, since we started building aircraft out of that).

The thing is, the current state of the art for an autopilot is fine handling routine situations. The biggest problem is when things are NOT usual.

In some ways an automatic pilot is easier because there's no such thing as a construction zone in the sky. As long as you don't get too close to the ground, nearly anywhere you can physically go is safe, in the "you won't immediately crash and die if you go there" sense. You're not going to run into a cloud stuffed with rocks.

You do however get a host of other environmental hazards and a vehicle that as per OP can probably pull more G's and certainly fly too high for a human pilot to be comfortable with. So you do have some constraints. Also, clouds full of hail.

Flying through a weather system that produces hail in an aircraft that size could well be lethal.

That's something not usually taken into account with flying car scenarios. What constitutes hazardous weather is a lot different for something that flies as opposed to drives. Simon's statement "As long as you don't get too close to the ground, nearly anywhere you can physically go is safe, in the "you won't immediately crash and die if you go there"" is untrue. One only has to look up the numerous accounts of turbulence causing injury to people on airliners to demonstrate that and something as small as an aircar will be affected more by the same sort of phenomena.

An autopilot will take you where you tell it to go, whether it should or not. I'd be concerned that people come rely far too much on a device that doesn't have the capability to tell you when you shouldn't go somewhere, and those same people will not be able to handle the situation if that device experiences some sort of failure.

[Simon_Jester]

That's a fair point; the autopilot would have to be slaved to a central database that tracks turbulence and bad weather, and can route aircars around dangerous storm systems.

[Purple]

That's a fair point; the autopilot would have to be slaved to a central database that tracks turbulence and bad weather, and can route aircars around dangerous storm systems.

That creates its own mass of issues. For a start, it limits the number of vehicles greatly as expanding the system to accommodate more of them now means a substantial investment in ground hardware and probably satellites as well if you don't want to build a network of radio towers just for these. Than there is also the issue of lag. I don't need to tell you why it's important that this system runs in real time. And having a computer that can process the sheer amount of data parallel for thousands if not millions of vehicles under those conditions is going to be expensive as hell. If anything, a centralized system makes the whole thing less viable and more difficult due to the fact that rather than paying the price in small bits with each vehicle you have to make giant infrastructural investments before you even field the first one.

[Borgholio]

That's a fair point; the autopilot would have to be slaved to a central database that tracks turbulence and bad weather, and can route aircars around dangerous storm systems.

That creates its own mass of issues. For a start, it limits the number of vehicles greatly as expanding the system to accommodate more of them now means a substantial investment in ground hardware and probably satellites as well if you don't want to build a network of radio towers just for these. Than there is also the issue of lag. I don't need to tell you why it's important that this system runs in real time. And having a computer that can process the sheer amount of data parallel for thousands if not millions of vehicles under those conditions is going to be expensive as hell. If anything, a centralized system makes the whole thing less viable and more difficult due to the fact that rather than paying the price in small bits with each vehicle you have to make giant infrastructural investments before you even field the first one.

Actually it's not hard at all. There are already apps that rely on a central database to provide road conditions and traffic information such as Waze and Google Maps. Going one step further and including turbulence / high-altitude weather isn't that big of a deal.

[Purple]

Actually it's not hard at all. There are already apps that rely on a central database to provide road conditions and traffic information such as Waze and Google Maps. Going one step further and including turbulence / high-altitude weather isn't that big of a deal.

The issue is not getting a hold of the data. That's easy. The issue is processing the data into individual flight commands for thousands if not millions of vehicles in real time. Unless when Simon_Jester mentioned routing the cars he simply meant telling them to blacklist certain areas as opposed to actually controlling them remotely.

[Borgholio]

Unless when Simon_Jester mentioned routing the cars he simply meant telling them to blacklist certain areas as opposed to actually controlling them remotely.

Yeah, we should probably get him to clarify. Because the way I see it, this system just has to say to the AP, "There is a weather system at these coordinates. Turbulence reported at these coordinates. Safe and stable flights reported along this route. Light air traffic at this altitude. Program this route at this altitude to your destination. (Nearly identical to a GPS app which reports where the traffic is and gives a route around it).

But if we're talking about full remote control...then yeah you're quite right, that would be a pain in the ass to manage.

[Broomstick]

We don't have remote detection of turbulence right now. Sure, we can say this sort of weather system is associated with that sort of turbulence, but that's fancy guessing, it's not knowing for sure. Currently, the most reliable turbulence reporting is from pilots reporting on actual conditions they are passing through.

Now, I suppose you could outfit all those AI autopilots with a reporting feature to let the "central database" know the actual conditions but there you go adding further complexity.

I also question if you really want a central database. Sure,that's convenient and efficient... as long as everything keeps working. If you have a power failure at the central office, though, your whole system will crash. I'd rather see a somewhat less efficient in money terms but more resilient network. Have distributed points where traffic is coordinated, with some sort of hand-off protocol as an aircraft passes from one to another. That way you're not going to lose the entire system (barring something apocalyptic) and you can even potentially have the nodes cover neighboring areas in an emergency.

Too often people envisioning these systems fail to imagine that in the real world shit breaks and power failures happen. If you're talking about aircraft, you system has to be able to handle that sort of system failure at least to the extent you can get everybody safely back on the ground because it's not like you can park motionless for a week up there waiting for repairs.

[Borgholio]

Well taking Waze or Google maps for example, they still download the route to your device so you can navigate even without a connection to the server. So long as the AP has a GPS signal and a pre-loaded map in memory it should be fine if anything server-side goes wrong or if the radio signal gets lost.

[Broomstick]

The AP still has to be able to keep track of other air traffic somehow, either on it's own or having the information relayed.

Maybe we need some sort of AI that functions like birds in a flock - flocking birds and schooling fish don't have a central controller, it's all based on individuals following a few simple rules based on the actions of those nearest to them.

Of course, this will really crank the poo-flinging monkeys who just HAVE TO go first, go direct, or otherwise won't put up with swoopy maneuvers to avoid crashes while getting where they're going.

[Borgholio]

Of course, this will really crank the poo-flinging monkeys who just HAVE TO go first, go direct, or otherwise won't put up with swoopy maneuvers to avoid crashes while getting where they're going.

Look at the bright side - with autopilot, you can flip them off without worrying about taking your hands off the wheel.

[Sky Captain]

I think if you use some sort of anti gravity system instead of thrust or aerodynamic lift to stay in the air a lot of problems relevant to normal aircraft would go away. For example stalling a wing and falling from the sky would not be an issue. Turbulence also may be less of a problem in a vehicle that don't rely on aerodynamic surfaces to fly. Ability to safely fly at any speed from hovering to max design speed would make air traffic managment easier. You could set speed limits in areas with heavy traffic to reduce chances of collisions. For example have a 70 km/h zone in over cities below certain altitude while at higher altitudes in certain corridors it is allowed to go at max speed.

[Me2005]

Ninja'd by Sky Captian!

I'd be inclined to go for it, combined with the facts that these provided cars come prebuilt with autopilot and use counter-gravity for VTOL and hovering. Zor may need to clarify, but if they're using counter-gravity to attain the altitudes they can attain and not wings or rotors, then they're nothing like modern aircraft. They might not even experience turbulence the same way, instead feeling it the way you would feel a gust against a passenger car on the road.

Another huge benefit that seems to be viewed as a detriment is the 3d space to fly in - instead of X kilometers of roadway, you've suddenly got X*10 kilometers of driving lane; without using any available airspace not over roads. Overall traffic density would plummet. Simple pre-programmed rules like "at X speed and heading you travel at Y altitude" could prevent many kinds of collisions - if everyone is always on the same vector at any given altitude, it's hard to crash into other drivers.

Maintenance would be the biggest bear I think, but a well-written maintenance agreement or mandatory lease could probably absolve the liability there. I don't see major auto manufactures bearing responsibility for car crashes today being caused by negligent drivers or poor end-user maintenance, so I don't think that'd be a problem here either.

[Broomstick]

They might not even experience turbulence the same way, instead feeling it the way you would feel a gust against a passenger car on the road.

The biggest factor in "turbulence feel" is the weight of the aircraft. Aloft, it's the air you're embedded in that's moving, it's as if the road a car is driving on is moving up/down/sideways. It's more analogous to boats, where the effect of rough seas on the ship occupants is more about the size of the boat relative to the waves than the means of propelling the boat

Another huge benefit that seems to be viewed as a detriment is the 3d space to fly in - instead of X kilometers of roadway, you've suddenly got X*10 kilometers of driving lane; without using any available airspace not over roads. Overall traffic density would plummet. Simple pre-programmed rules like "at X speed and heading you travel at Y altitude" could prevent many kinds of collisions - if everyone is always on the same vector at any given altitude, it's hard to crash into other drivers.

It also allows for a greater separation between traffic streams, which will sharply cut down on sideswipe accidents, and if another vehicle goes out of control you can dodge in three dimensions and not just two -- as long as everyone doesn't dodge to the exact same location.

Maintenance would be the biggest bear I think, but a well-written maintenance agreement or mandatory lease could probably absolve the liability there. I don't see major auto manufactures bearing responsibility for car crashes today being caused by negligent drivers or poor end-user maintenance, so I don't think that'd be a problem here either.

It's a different story for aircraft. Thanks to some court cases and laws in the 80's aircraft manufacturers are liable only for 14 years after production... but prior to that there was no limit and the lawyers had fun soaking manufacturers, which almost killed the industry. The lawyers will want to treat these as aircraft when it comes to liability, not as cars.

[Me2005]

It's a different story for aircraft. Thanks to some court cases and laws in the 80's aircraft manufacturers are liable only for 14 years after production... but prior to that there was no limit and the lawyers had fun soaking manufacturers, which almost killed the industry. The lawyers will want to treat these as aircraft when it comes to liability, not as cars.

And this problem I will solve by refusing to sell any lawyers any of my super-fun flying cars unless they all agree not to sue me.

An interesting angle this scenario introduces is that I am the sole person capable of producing these vehicles. Suing me out of existence could well be against the interest of every industrialized nation and major cooperation in the world, providing I play my cards correctly when starting out (sell/lease them to company drivers, military, and government officials to start; sell *just* the power cells to automakers and power companies, etc.). It might also be possible to establish law preventing me from being sued in this avenue as well, prior to going mass-market. Another factor possibly in my favor is that I'm guessing most of the aircraft suits were not against individual civilians crashing them, but company pilots trained by the manufacturer flying paying passengers. I'm flying individuals who crashed *their* air-car into a tree/mountain/hail-cloud/etc., likely against the instructions of the onboard autopilot (which they had to tamper with to allow them to crash).

But I'm dubious they'd be that dangerous in the first place. If the AG unit works as it says on the tin, they auto-descend safely when low on power or damaged and stay in the air by magic the rest of the time.

[Darth Tedious]

Regarding the state of AP development, didn't the Russian space shuttle run its (one and only) unmanned test flight completely under automation?

I realise aerospace is a bit more cutting-edge in tech than aviation or automotive, but that was twenty-odd years ago

[Broomstick]

Another factor possibly in my favor is that I'm guessing most of the aircraft suits were not against individual civilians crashing them, but company pilots trained by the manufacturer flying paying passengers.

You are incorrect, sir. Even in cases where the NTSB found that the pilots were wholly at fault when it came to liability cases in the US the airplane manufacturers were being penalized in case after case.

Consider a maker like Boeing, which dates back to 1910, having on-going liability on every single aircraft ever made by the company. Until 1994 that was, in fact, the case. This applied not only to aircraft originally manufactured and sold to civilians but even to aircraft originally sold to the military and later decommissioned and re-sold to the civilian market. Other long-life manufacturers included Cessna (founded 1911, about half the world total of aircraft still flying today were made by them) and Piper (founded 1927, approximately 90,000 airframes still flying). This is why most of them stopped making piston-engine airplanes in the 1980's and were desperately seeking a way to limit liability on planes the manufacturer hadn't seen in 70 (or more!) years but were still being held liable for in any accident regardless of who was actually at fault.

The US General Aviation Revitalization Act at last limited manufacturer liability, but it's far from perfect.

I'm flying individuals who crashed *their* air-car into a tree/mountain/hail-cloud/etc., likely against the instructions of the onboard autopilot (which they had to tamper with to allow them to crash).

In the US the lawyers will still try to sue you for not making your autopilot feature truly foolproof - the problem being, of course, that fools are so ingenious.

[Simon_Jester]

I think part of the solution here is to hire some very skilled lawyers to set up a waiver system in advance. Something like (and yes, obviously this needs to be ten paragraphs of legalese):

"I hereby recognize that the autopilot system is in fact designed to avoid collisions and other air accidents. It is not to be turned off without proper cause. If the autopilot must be shut down due to a malfunction, I hereby agree to take the aircar in for depot maintenance at the earliest opportunity; failure to comply voids the warranty.

If I override the autopilot without cause to believe it is malfunctioning, I am taking my life into my own hands. The manufacturer is not repeat NOT responsible for my controlled flight into terrain, willful disregard of basic directions provided by autopilot or ground control, willful disregard of common sense, or my decision to prohibit my autopilot from swerving to avoid obstacles causing me to smack into them like a bird into a windowpane."

That creates its own mass of issues. For a start, it limits the number of vehicles greatly as expanding the system to accommodate more of them now means a substantial investment in ground hardware and probably satellites as well if you don't want to build a network of radio towers just for these. Than there is also the issue of lag. I don't need to tell you why it's important that this system runs in real time. And having a computer that can process the sheer amount of data parallel for thousands if not millions of vehicles under those conditions is going to be expensive as hell. If anything, a centralized system makes the whole thing less viable and more difficult due to the fact that rather than paying the price in small bits with each vehicle you have to make giant infrastructural investments before you even field the first one.

...My GPS can tell me where traffic jams are, and tries to suggest routes around them. This might actually be easier.

[Then again, to be fair, my GPS kind of sucks at spotting traffic jams, so yeah]

Actually it's not hard at all. There are already apps that rely on a central database to provide road conditions and traffic information such as Waze and Google Maps. Going one step further and including turbulence / high-altitude weather isn't that big of a deal.

The issue is not getting a hold of the data. That's easy. The issue is processing the data into individual flight commands for thousands if not millions of vehicles in real time. Unless when Simon_Jester mentioned routing the cars he simply meant telling them to blacklist certain areas as opposed to actually controlling them remotely.

Yes. Yes I did. Sorry, it honestly never occurred to me that you would think I meant to have a central computer literally take control of the flying controls of each individual aircraft.

Air traffic control "routes planes" all the time in real life, but does not take over the instruments and controls in the cockpit of those planes.

We don't have remote detection of turbulence right now. Sure, we can say this sort of weather system is associated with that sort of turbulence, but that's fancy guessing, it's not knowing for sure. Currently, the most reliable turbulence reporting is from pilots reporting on actual conditions they are passing through.

Now, I suppose you could outfit all those AI autopilots with a reporting feature to let the "central database" know the actual conditions but there you go adding further complexity.

Not really; you just need a "if aircar starts to shake, ping location with "HELP! SHAKING!" flag."

Also, question: is life threatening turbulence likely to occur in unpredictable parts of the sky? I mean, we can predict that aircars shouldn't fly anywhere near a major storm system, where we can expect major updrafts/downdrafts/whatever that would swat them out of the air. That's easy to do automatically. Is it normal for there to be turbulence in basically clear air, that cannot be logically deduced from the ground, but is nevertheless intense enough to pose a deadly hazard to light aircraft?

I also question if you really want a central database. Sure,that's convenient and efficient... as long as everything keeps working. If you have a power failure at the central office, though, your whole system will crash. I'd rather see a somewhat less efficient in money terms but more resilient network.

Okay, you're right, but... seriously, now you're just picking nits.

My phrasing should not be interpreted to mean "I want a system that is expensive, chronically overloaded, and riddled with single point of failure components."

I know I have my faults, but I'd appreciate it if someone would make the assumption that I'm not actually a complete dimwit. Heh.