How'd you program the morality of auto-cars?

[salm]

Underground parking lots? Campgrounds? Moving a car onto a sports field to pick up an injured 12 year old and drive him to hospital? Move your car into the backyard so you can use the headlights to light up the back of the house and fix a tarp over the gaping hole in the roof a falling tree has made? There are all sorts of reasons to move a car that are not specifically drive from venue A to venue B and then stop precisely our the front, and anything that cannot be precisely predicted and allowed for by a programmer who has probably never seen your house, your hometown or a member of your family needs some manual controls.

Do you really think underground parking lots and things like that are unsolvable problems?
Now, I wouldn´t care that much if specific things like lighting you backround weren´t possible with cars because the advantages would still outweigh the disadvantages.
Things like parking in an underground parking lot where no gps is available should be solvable rather easily. I mean, the car needs a bunch of sensors to operate in other areas where no gps is available like tunnels or on ferries, so it will definately be possible in underground garages.

I don´t think that telling a car to cross areas that are not specifically designated for cars on user input shouldn´t be too dificult either. Just give the user some some sort of interface via a map on a touchscreen or telephone, enter safety mode, so that the car moves at a max speed of 10km/h and let the car move to the point specified by the user. Or if a map isn´t usefull like in on a frequently changing camping ground let the user input the data via touchscreen similar to a touchpad in a game on an iPhone.
I am sure there are better ways to interface in such a situation but it doesn´t seem very difficult to come up with a solution that would work for such cases.

[madd0ct0r]

I think all parking would be done using sensors. Civilan GPS is not accurate enough at that scale.

[Simon_Jester]

For maneuvering cars in unusual conditions, manual override is likely to remain very desirable. It's not that the robot car can't theoretically do any one thing a human driver needs. It's that the combination of all the reasons humans move cars other than "GPS-guided Point A to Point B" is pretty significant.

There is also the issue of what happens to a robot car that gets partially stuck in a snowbank, spinning its wheels. Or skids some distance on icy roads. For purposes of negotiating foul weather, especially snow, I've had a number of cases where I used direct driver control of the vehicle to accomplish something that would be totally impossible for a robot car- the robot car would have to sit there helplessly making 'ping central office' noises.

When you really, really need someone to ram their old Volvo through a snowdrift so they can give your car a jump-start so that you can start your 500 mile drive where you have to be at work the next day... You're thankful that the person who owns that Volvo can in fact ram it through the snow drift.

[salm]

Robot cars could still get rid of the steering wheel and other parts that are now necessary for controlling a car. They could drive by wire (or WiFi) and all the heavy mechanicle parts could still be kept out of the car.
For situations in which the car gets stuck and can not analyse what the situation exactly is you could implement emergency programs. Get stuck in a snowbank and the user just has to push the "got stuck in a snow bank" button which will make the car try certain things to get out.

However, I think that certain functionality you are able to have with todays cars will simply not be available in robot cars just like horses had certain functions that are not available in todays cars.

[Sky Captain]

Robot cars could still get rid of the steering wheel and other parts that are now necessary for controlling a car. They could drive by wire (or WiFi) and all the heavy mechanicle parts could still be kept out of the car.

I think something like small drive by wire joystick could be implemented for manual override rather easily. It wouldn't be as convenient as steering wheel and pedals everyone is used to, but certainly much better than having computer just stop a car in some weird situation a human can easily solve.

However, I think that certain functionality you are able to have with todays cars will simply not be available in robot cars just like horses had certain functions that are not available in todays cars.

Problem is functionality that would be lost with fully robotic cars is very important to many people so I would expect a car with no manual control just would sell very badly. Pretty much everyone would love to have autopilot that can take care of daily commute and routine driving, but few would want to give up functionality and utility a manual control gives when needed. Also from software perspective it would be much easier to make driving control software than can deal with 95 % situations than software that can successfully deal with 99,99% of situations car may encounter on the road.

[Simon_Jester]

Robot cars could still get rid of the steering wheel and other parts that are now necessary for controlling a car. They could drive by wire (or WiFi) and all the heavy mechanicle parts could still be kept out of the car.

I think something like small drive by wire joystick could be implemented for manual override rather easily. It wouldn't be as convenient as steering wheel and pedals everyone is used to, but certainly much better than having computer just stop a car in some weird situation a human can easily solve.

The biggest obstacle there would be that the manual override interface would seem unfamiliar to the early adopters.

It might actually be easier to have a steering wheel and pedals that are themselves drive-by-wire, not mechanically linked to the rest of the car. That's already happening, after all, and it's not like the weight of the steering wheel and pedals is that much as such. Keeping them doesn't really have much opportunity cost.

However, I think that certain functionality you are able to have with todays cars will simply not be available in robot cars just like horses had certain functions that are not available in todays cars.

Problem is functionality that would be lost with fully robotic cars is very important to many people so I would expect a car with no manual control just would sell very badly. Pretty much everyone would love to have autopilot that can take care of daily commute and routine driving, but few would want to give up functionality and utility a manual control gives when needed. Also from software perspective it would be much easier to make driving control software than can deal with 95 % situations than software that can successfully deal with 99,99% of situations car may encounter on the road.

What it comes down to is that people will only accept a loss of functionality if they get back something a great deal better in trade, and don't have the option of getting the best of both worlds.

As you say, people want an autopilot- but if the only choices are "manual control and no autopilot" and "autopilot and no manual control," there are a lot of reasons why people might choose the former over the latter.

So the people predicting "autopilot and no manual control" as the future remind me vaguely of those articles in early 20th century popular-science publications predicting that evolution was making humans have bigger and bigger brains until the humans of THE FUTURE would just be giant heads with tiny vestigial bodies. It's kind of silly because it's based on the idea that not only will present trends continue, but they will continue indefinitely without running into any limiting factors.

[Zeropoint]

The biggest objection I see to the grand vision of automated highways being presented here is that it turns a fleet of cars into what amounts to a half-assed mass transit system. If you REALLY want people riding energy-efficient transport that they share with other people, which operates on fixed routes with no user input, and has a "closed ecosystem" to prevent delays caused by private drivers messing things up . . . build yourself a good subway system.

Don't try to turn cars into a cheap copy of trains. If you want trains, build trains.

[salm]

We allready have a good subway system. It is obviously not enough.
I don´t understand how you´d get to the conclusion that automated cars would be in any way more similar to subways than to todays non robotic cars.
I mean, even if they couldn´t be privately owned for some reason they´d still be more similar to todays taxis than to any kind of train.

[Lagmonster]

Don't try to turn cars into a cheap copy of trains. If you want trains, build trains.

That's a weak argument. Many cities don't have the density or revenue to support a rail system for public transit, or the climate sucks, or maybe just building a subway at all is a herculean feat because the whole city is spread over marshy flats.

Take mine, which is going to take a decade to build just one light rail line under just one part of the downtown core, and they aren't even close to offering a rail line to the south or west of the city at all due to a combination of available land and the sheer amount of parks, forest and farmland between suburban centres. But because a lot of our growth has taken place within the past 50 years, we have lots of wide arterial roads to take advantage of and long stretches of cross-city private roads only useable by buses, emergency vehicles and taxis. For us, with a low population density, lots of roads, and low overall taxpayer population, the self-driving car (or better, bus) is absolutely the better option to fixed rail lines.

[madd0ct0r]

fuck sake guys, talk about the difference between concentrated point to point traffic flows and distributed traffic flows (the latter look more like bare trees - lots of small flows come together to form the main trunk. Can even be double ended trees if you prefer).

a subway or even a bus system will only be able to support distributed flows at certain scales:
1) very short distances, when the twigs are a 10-15min walk
2) country/state size scales, when the twigs are flows out to a specific village (and even then it falls down if you have many people living another twig size along, in isolated farmhouses)

It's an important concept, especially since freight is steadily switching from point to point (distribution center to super market) to distributed (amazon distribution center to house. Super market to house ect). The more distributed the traffic flow becomes the less rail can compete (which is high capital, but can move stuff very cheaply, point to point)

elec cars that plan on traveling the same route for a few hours locking up against each other to convoy in less space seems doable, although I think the benefit from slipstreaming would be low* and the crash conditions may not allow it. Freight distribution vans might benefit more from it.

*I know Tuxedo found 20% for a car behind a lorry, but a chunk of that is driven by the height differential. The car benefits from being about half the height of the lorry, so the swirling vortex starting behind the top of the lorry is pushing the opposite way at the level of the car. The car's basically getting pushed along. When the lead vehicle is the same size and reasonably aerodynamic, a lot of that area isn't effective. you certainly don't get the same fun vortex effect.

[Simon_Jester]

We allready have a good subway system. It is obviously not enough.
I don´t understand how you´d get to the conclusion that automated cars would be in any way more similar to subways than to todays non robotic cars.

It's not that they would be. It's that a couple of people posting in this thread think they would.

I mean, even if they couldn´t be privately owned for some reason they´d still be more similar to todays taxis than to any kind of train.

I quite agree.

Don't try to turn cars into a cheap copy of trains. If you want trains, build trains.

That's a weak argument. Many cities don't have the density or revenue to support a rail system for public transit, or the climate sucks, or maybe just building a subway at all is a herculean feat because the whole city is spread over marshy flats.

Take mine, which is going to take a decade to build just one light rail line under just one part of the downtown core, and they aren't even close to offering a rail line to the south or west of the city at all due to a combination of available land and the sheer amount of parks, forest and farmland between suburban centres. But because a lot of our growth has taken place within the past 50 years, we have lots of wide arterial roads to take advantage of and long stretches of cross-city private roads only useable by buses, emergency vehicles and taxis. For us, with a low population density, lots of roads, and low overall taxpayer population, the self-driving car (or better, bus) is absolutely the better option to fixed rail lines.

Thing is, in that case the self-driving cars are not acting like trains, which is the specific prediction that Zeropoint is responding to. It's quite true that cars (robotic or otherwise) fill a niche trains can't, in places where it'd be uneconomical to build trains. By the same token, nobody wants cars (robotic or otherwise) that act like inferior train impersonators.

[Sky Captain]

The biggest obstacle there would be that the manual override interface would seem unfamiliar to the early adopters.

It might actually be easier to have a steering wheel and pedals that are themselves drive-by-wire, not mechanically linked to the rest of the car. That's already happening, after all, and it's not like the weight of the steering wheel and pedals is that much as such. Keeping them doesn't really have much opportunity cost.

Eliminating steering wheel would allow to place a small foldable table there that could be used to put laptop, eat something or do something else while car drives itself so there would be some gain by removing the wheel which take up lot of space and wouldn't be used very often in a robotic car.