Self-Driving Cars

Can this topic be revived? After all, more information comes out from time to time.

For example, we're starting to see more concept designs for cars that look nothing like cars. After all, if they'll drive themselves totally (at some point), then why not have them be like a mobile living room?

Why not have them be mobile living rooms? Space reasons that's why.

For the forseeable future these automated conveyances are going to be using existing transport infrastructure, meaning roads and parking spaces. Which means that for practicality they're going to have to be approximately car sized and car shaped. And while that doesn't entirely eliminate the concept of the mobile living room such as mobile homes, it does illustrate it's limitations.

Though I can forsee something with a re-configurable interior that would allow for various permutations. Normal seating for three or four people if a whole family or a bunch of friends is going out but switching things around if only one or two people are in the car for more room and more comfortable seating arrangements. Maybe even reclining enough that commuters can snooze on the way to and from work.

I didn't mean literally be a mobile living room. I meant be designed more for comfort. Like instead of all seats facing forwards, you could have seats arranged in a circle or at the corners of the car, or be designed to face a TV/computer screen, in something roughly the size of a traditional car.

Hilariously late to the discussion, but I'm not sure how to feel about self-driving cars. On the one hand, there's an opportunity to reduce accidents made due to human error. Excellent!

On the other hand, like others have said before I'd like to be able to manually override the system. What happens if the system should experience a failure mode unforeseen by its designers? What if something decides to pack it in and the self-driving system's capabilities are degraded or compromised entirely?

I find the comparison to autopilot particularly relevant, as there have been several occasions where modern automation has ended up contributing to dangerous incidents rather than preventing them. For example, during the Qantas Flight 32 incident, the aircraft sustained such significant damage that its computers could not keep up with it and reverted to manual control laws^note This degraded the level of protection significantly, as the aircraft would no longer protect against flying it in an unsafe manner. Another example would be several accidents that have involved disagreements between various data inputs, causing contradictory instrument readings.

Keep in mind that these are systems built to exacting tolerances and with multiple redundancies built in. I am not certain that automotive manufacturers would go to similar lengths. I mean, it's not like they'd do no testing whatsoever, or fail to anticipate likely problems, but when it comes to this sort of thing you need to be certain that the margin of safety is extremely high.

edited 18th Nov '14 3:47:44 PM by Flanker66

As has been noted, though, driving a car is easier than flying a plane in many ways. In particular, the answer to the vast majority of emergencies in a car is "STOP." Unless there's an avalanche or a few other very specific problems, stopping the car is the best answer. Though I do agree that a manual override is still necessary.

What'll be the biggest advantage is that the cars will likely be required by law to communicate all information with each other, and likely do so within industry standards. There'll be a tipping point when all cars are required to become self-driving because the remaining humans will be too big a liability.

That's what it is at the end of the day. Of course, there will be huge political forces marshalled against it, but technology will win as it always tends to.

This contradicts actual accident rates in cars and planes. Notably, air tends to be really empty, and roads tend to not.

Self-driving cars sound neat, but they're an overly-complicated solution. We should use public transport, it's better in every way.

Public transport outside of major cities is almost guaranteed to suck. When I lived out in the country, the closest public bus stop was over an hour away if you walked, and this wasn't even that far from Toronto (by Canadian standards.) You might be able to get transit between towns and even within towns if they're big enough, but otherwise, you need another method of getting around, or be willing to walk for hours.

And even in cities, it can be very...erratic. I avoid using the TTC buses for a reason. The subway at least comes every five minutes. The buses can be up to an hour late, on a route where they're supposed to be every 15 minutes on a day with no traffic and good weather.

edited 18th Nov '14 9:55:44 PM by Zendervai

Even in more metro areas like my city(Colorado Springs) the routes can often be so thinspread it's almost impossible to use as a serious mode of transit

,

I'll ditto both. Where I live (South-East Queensland in Australia) there is a chronic problem with lack of public transport. A large part is (especially for train lines) is that the government waits for population to build up to a certain level before they start considering train routes or major busways to the area.

And providing good public transportation is a much easier problem than creating a self-driving car.

Even public transport still need someone or something to drive.

In fact, self-driving public transport is better given that 1) the drivers no longer need rest and can run for as long as you can afford the cost and 2) they are far less accident prone.

edited 19th Nov '14 12:10:00 AM by IraTheSquire

Absolutely true, not arguing with this. But automated public transport is already a reality. Not so for cars.

It sounds great but unfortunately I feel like there is going to be a lot of kicking and screaming by the masses. Though I'm sure metropolitan cities where driving is a pain in the A will see them commonplace with the next 20-30 years.

Not if the said public transport is on roads (ie buses).

edited 19th Nov '14 2:37:05 AM by IraTheSquire

Cities will adapt this stuff fast, where the hazards are more predictable than the countryside and where driving is less fun and more of a chore (and even a simple AI could negotiate slow-moving rush hour traffic)

@ Lsama: That's because, below a certain level of population density, it isn't worth running public transport, even if you're a local authority, because passenger numbers will never be high enough to justify running/paying to run vehicles.

Just a friendly reminder: This thread is in On-Topic Conversations. Public Transportation in general is not sufficiently related to "Self-driving vehicles" to be on topic. If you want to talk about self-driving public transportation, that would be on-topic.

Thanks...

Once self-driving cars get made, I imagine there will at some point be special roads designed for insanely high speeds. Imagine special 120mph (or faster!) roads made that only self-driving cars are legally allowed to use, with severe penalties for anyone not riding a self-driving car who tries to use them.

I'd quibble with what some folks are saying about flying as opposed to driving, but that is a topic for PMs rather than here.

Anyway,

Agreed and disagreed. Depending on whether it falls into your "very specific" qualifier, I don't think it'd be unreasonable to point out that if, say, a self-driving car tried to just stop on a dual carriageway or motorway it'd probably get smacked by at least one numbskull who didn't expect it. Now, you could argue that's hardly the self-driving car's fault but I'd say that the risk of such an accident is a consequence of only having a simple "stop" command - albeit an unintended one. And, of course, travelling at 70 MPH in potentially busy traffic is not the time for a critical piece of hardware to decide it's had enough and put on the proverbial anchors.

I'm assuming here that the self-driving system has completely failed, or at least is so degraded as to no longer be able to provide effective guidance. Therefore, automatically going into a lay-by or drifting off on to the hard shoulder isn't necessarily an answer - in such a condition, how could it know where the lay-by or hard shoulder is, much less act on it?

You may think that I'm cooking up unlikely scenarios, but it's a by-product of my mindset as a budding pilot and engineer. All the bases must be covered, even those that may seem vanishingly improbable. After all, it's often an unlikely failure or design oversight that causes major incidents.

One potential solution - taken from aviation - is for the vehicle to have an obvious and attention grabbing warning when the self-driving components have failed, with the vehicle reverting to manual control in a safe manner. Of course, this would require the occupants to be ready to take control if necessary, which some may argue defeats the entire point of a self-driving vehicle. My own feelings are that you should be keeping your attention on the road, all singing all dancing self-driving systems or no.

I think I've pretty much exhausted all I have to say on the safety aspect, though - if I can think of anything further to add to the discussion (safety related or otherwise) I'll pop my head in again.

You're thinking about when self-driving cars operate in a vacuum. In the short run they'll have to deal with human irrationality. In the medium term human-driven cars will feed telemetry to communicate their location, speed, and whatnot so that self-driving cars can correct. In the long run, self-driving cars coordinate with each other like a hive mind.

But what happens if that stops working for whatever reason, or is interrupted?

Actually that raises what I think it one issue that self-driving vehicles are going to have to deal with in the future. Are they going to be under autonomous control or use some external system that controls all the vehicles on the road?

Are the chances of that telemetry getting interrupted higher than the chances of human error?