Could driverless cars replace railways?

[cmjcf]

They WON'T be able to spot that cyclist, or that child, or that ice-cream van and take appropriate action in the same way that a human being would.

They can already, and can both take the correct action and assume proper control of the vehicle. In much the same way as trading computers can respond to movements in the market in a fraction of a second, an AV with highly detailed metrics can detect in a fraction of a second whether the vehicle is responding correctly. We already have these systems in the form of traction control and ABS, and the state of the art is only going to advance. We already have cars that can park themselves.

 

[jon0844]

They could do a lot of things, while also failing for a number of reasons.

The idea of them doing 120mph with half a second gap is fine, in theory, expect these are all mechanical objects. One car suddenly has a technical issue and - well, suddenly you can imagine the inquiry that decides it wasn't so safe after all. The other cars might be alerted to the defect in a millisecond, but they still aren't going to stop in time.

And there's the problem. In ideal circumstances, everything is just fine, but you won't have technology certified for use when there are doubts. Any more than you'd begin to run trains with next to no gaps because they're all computer controlled.

A computer can react quicker than a human yes, a computer might recognise risks that a human wouldn't at all. But the computer is still bound by the laws of physics and common sense.

 

[starrymarkb]

.

I believe many modern aircraft are inherently unstable. They need computer equipment to allow them to fly straight and there are already many unmanned drones in existence. However imagine the reception the idea of Ryanair introducing pilot less aircraft would get.

Fighters are unstable to increase maneuverability, commercial airliners are very much stable without computer input.

Ryanair would never go Pilotless, they have several pilot aids disabled on their aircraft anyway (Ryanair can't do an automatic landing in low visibility)

 

[Bob Ames]

Are railways doomed?

Unless they can get me to Glasgow for £11, I doubt it.

 

[bnm]

There may very well be arguments against them, but ill-informed Luddite nonsense such as your examples aren't among them, and don't really help the debate.

And neither does branding someone an 'ill-informed Luddite' help your side of the debate.

 

[carriageline]

I can't really see them being able to get stupidly close on the motorway, and all being able to stop as one unit. Sensor controlled machinery cannot react that quickly, and I think we can react quicker on the whole.

GPS etc idea, what about tunnels and other black spots? Many problems with that, what if traffic is at a stand still inside a tunnel and your car remains hurtling towards it at 120mph?

What would the car do if a fault developed with say the front sensor? I'm sure sensors are common replaced things in cars, although probably a different concept, they won't last forever.

 

[cmjcf]

The idea of them doing 120mph with half a second gap is fine, in theory, expect these are all mechanical objects. One car suddenly has a technical issue and - well, suddenly you can imagine the inquiry that decides it wasn't so safe after all.

Sudden mechanical failure is incredibly rare. In the majority of cases, the signs are there - we're just not able to pick them up. A computer system, OTOH, can report to its driver "Something doesn't feel right, I'm slowing down. I'll check myself into a garage while you're in work." It could then transmit telemetry to the garage, and remote diagnostic software could figure out "This looks like an electrical issue, best arrange it for when our sparky's in." F1 already uses computer-assisted telemetry to diagnose faults while the car is moving at up to 200mph. This is only going to get better and cheaper.

What people here aren't getting is that there are no technical reasons whatsoever why it can't or won't happen. All the rubbish you could make up about how it won't get out of a junction or how it handles at high speed or whether it can stop is exactly that - rubbish. There's more than enough time for things like that to be worked out. Most of the pieces are already there.

The one place where there are problems to work out is in the ethical sphere. If there is an accident and someone is injured, who is responsible? The owner or the manufacturer? What if the safeguards fail? What if they're tampered with? What if the owner hasn't been diligent with their maintenance? Will the whole thing result in "drivers" disconnecting from the reality of their travels?

There's simply no point arguing over the technology. It exists, and it's only a matter of time. The real question isn't over whether we could have them, but whether we should have them..

 

[Xenophon PCDGS]

On a socially-aware point, how will "TWOC-ers" react to the scenario of driverless cars....

 

[yorksrob]

Well, yes and no. I can‘t help thinking we‘ve had all sorts of exciting new technology over the past couple of decades. WAP phones, digital terrestrial tv, internet transactions, I.T in pretty much any large organisation you care to mention. None of them seem to work all that well.Technology has a habit of promising the earth and not quite following through on it.

 

[jon0844]

Most of the technology is indeed there. Just not all of it, and that's why such things aren't already in use and probably won't be for many, many, years.

Those final pieces of the jigsaw require loads of things that we can't get completely right, including the holy grail; artificial intelligence that might one day mean a computer can learn to do everyday human can (we still have some advantages over a computer! ).

The ultimate solution is a car that combines all of the tech with a real driver, perhaps having sections of road where auto mode is possible. But in auto mode, I'd still say cars will be forced to go at slower speeds than suggested here. It won't be like the Fifth Element!

We're also ignoring legislation that will be needed and the issue of dealing with accidents, as acknowledged, such as when the computers make the wrong decisions that could cause them.

Do we really believe that millions of these things could communicate with each other as would be necessary? Take an accident where vehicles need to divert. The car would need to know which way to go, but it would make sense to divert cars in different ways to lessen congestion. So now all cars need to have their location checked in real time and assessed. Or do you have a central control managing everyone?

And, like the GPS situation, what about when that goes down or a car loses its data connection because of local issues, like a power failure or even vandalism? Imagine the temptation to bring down the entire transport system in any given area by taking out the ground stations that communicate with the cars.

My TomTom uses crowd sourced traffic info, collated centrally, and it works really well but not always. We would need a foolproof system, along with loads of other foolproof systems. I'd say it's easier to make planes that take off, fly and land themselves because it's a controlled environment and so has no comparison with cars.

 

[radamfi]

I suppose that driverless cars could adopt the standard stopping distances as shown in the Highway Code, but with the elimination of 'thinking distance' as that would be tiny. However the braking distance at 70 mph would 75 metres, which is probably bigger than the gaps typically left by human drivers today!

 

[bnm]

What people here aren't getting is that there are no technical reasons whatsoever why it can't or won't happen. All the rubbish you could make up about how it won't get out of a junction or how it handles at high speed or whether it can stop is exactly that - rubbish. There's more than enough time for things like that to be worked out. Most of the pieces are already there.

How does a computer cope with an animal incursion onto the road? How does the computer cope with pedestrians who make a mistake and step into the road at the wrong time? What about criminal acts? What of atmospheric conditions? What about heavy commercial vehicles? What about the costs?

And why are the other things people are mentioning being, on the one hand, dismissed as 'rubbish', yet, on the other hand, you are saying, 'things like that can be worked out'. Wouldn't need working out if they are rubbish. :roll:

Then there's you, on the one hand, dismissing scenarios that may cause an accident or failure in the system, yet on the other, pondering the ethical issues of who is to blame when an accident occurs. All very confused.

You are very dismissive of counter arguments to your proposal, and that does your point of view no favours. Did you expect 100% agreement when you started the topic?

There's simply no point arguing over the technology.

There is plenty of point in arguing over the technology. And doing so doesn't make one a Luddite.

 

[carriageline]

Why is everyone ignoring what I am saying about lower reaction speeds?!

 

[radamfi]

Why is everyone ignoring what I am saying about lower reaction speeds?!

I just mentioned that thinking distance would be eliminated! Braking distance, presumably, would still remain as it is now.

According to the Highway Code, stopping distance = thinking distance + braking distance.

 

[carriageline]

I just mentioned that thinking distance would be eliminated! Braking distance, presumably, would still remain as it is now.

According to the Highway Code, stopping distance = thinking distance + braking distance.

But it wouldn't be. A computer still has to "react", from a conversation I had with someone who is alot more "advanced" in the subject than I, computers/sensors cannot react faster than us. The braking distance of course won't change.

 

[radamfi]

Humans: stopping distance = thinking distance + braking distance
Assuming instantaneous reaction, driverless cars: stopping distance = braking distance

So at 70 mph:

Humans: stopping distance = 21m + 75m = 96m
Driverless cars: stopping distance >= 75m

So there wouldn't be that much of a capacity benefit on fast roads, even if you assume that driverless cars have no thinking time, although hopefully injuries should be virtually eliminated. The main benefit would be in urban areas and for cheap on demand transport for those without their own private cars.

 

[jon0844]

While you might accept that if someone steps out in front of a train doing 125mph, there's nothing a train driver (or computer controlled train) could do - we would expect something to be done on a road.

In a split second, how would a computer on a car react? First it has to establish there's a real threat (otherwise, just seeing a kid on the pavement would have cars slowing down or stopping 'in case' and that would be damn annoying and also screw everything up for other cars around it) and then it has to do something. Does it stop (and tell the vehicles behind to stop, hopeful that the message gets through) or swerve? If it swerves, can it maintain control to stop it hitting something/someone else? And if it wanted to stop, could it stop in time anyway?

Sure, a human might make a mistake - but at least they're in control and can make that call. Now we'd be mere passengers and have no idea what will happen for sure.

Judging a threat is one of the many things a human can do better than a computer. We can look and see what's going on and judge that people walking around on the pavement aren't about to leap in front of a car. Or we can see that ball that rolled out onto the street and wonder if a kid will come running after - thus already slowing down in case, giving a better chance of stopping safely than a computer driven car that only reacts to what has actually happened.

Imagine the processing power for a car to be monitoring multiple people and objects, and making human assumptions about what might happen when there isn't evidence?

At Mobile World Congress in February, I interviewed someone from ARM who showed me their latest processor that can analyse video in real time - and they were very proud of it being able to find and track multiple faces. To work in a car, driven in town, I'd expect you'd need to be tracking far more people and even then you aren't just tracking them, you're trying to work out what they're doing.

As of today, you could develop this to warn a driver to be aware of things - but not to actually take control. I really can't see that changing for many many years.

 

[gordonthemoron]

surely driverless cars would be more practical if they were shared with others (e.g. carclub members) and the cars could achieve a high load factor? That would mean you wouldn't require so many of them and they wouldn't use so much roadspace or electricity?

 

[jon0844]

Why not just introduce more buses, considering them to be a public service worthy of more subsidy. At least you then have some flexibility on the day you need to go in a bit earlier/later.

What about getting people to give up their own personal car that they can jump into at any time of day?

Convincing people of that, as well as convincing them that they no longer get to drive themselves, will be tough. This is something that would really make politics interesting!

I think we'd find it easier to spend the money building more, proper, cycle routes (not the half arsed attempts in many places today) and expand the Boris bike schemes, so people could ride from home to their nearest transport hub - for a bus, coach, train or taxi. And while that would cost a fortune, especially in rural areas, it would likely still be cheaper than making driverless cars replace ordinary cars.

The necessary R&D costs would make such vehicles cost a fortune, and likely take many, many years to break even - if ever. I'd personally expect that you'd need a Government to invest our money in such schemes and we'd also need every system to work together - so all these trials being done by various companies would need to be made into one big trial. We couldn't even introduce a new computer system for the NHS!

 

[90019]

Sudden mechanical failure is incredibly rare. In the majority of cases, the signs are there - we're just not able to pick them up. A computer system, OTOH, can report to its driver "Something doesn't feel right, I'm slowing down. I'll check myself into a garage while you're in work." It could then transmit telemetry to the garage, and remote diagnostic software could figure out "This looks like an electrical issue, best arrange it for when our sparky's in."

I disagree.
I'd say I'm better at feeling that something isn't right with a car than a computer - to the point where I can say for certain that there's a fault and subsequently have it diagnosed properly by a mechanic, while the computer sits there telling us that the car is actually running fine - an example being the slight misfire my Polo currently has.

On the other hand, I do bother to check my car over on a semi-regular basis.
--- old post above --- --- new post below ---

In a split second, how would a computer on a car react? First it has to establish there's a real threat (otherwise, just seeing a kid on the pavement would have cars slowing down or stopping 'in case' and that would be damn annoying and also screw everything up for other cars around it) and then it has to do something. Does it stop (and tell the vehicles behind to stop, hopeful that the message gets through) or swerve? If it swerves, can it maintain control to stop it hitting something/someone else? And if it wanted to stop, could it stop in time anyway?

As an example of this, say you're driving down the motorway, and as you're approaching a bridge you can see a person on the bridge. If you have a person driving the car, they can make a decision abouy whether that person is just crossing the bridge, watching the traffic or about to jump.
What would the computer do?
Would it slow down for every person it sees on a bridge?
Would it not react to a possible hazard until it actually happens? In which case it would have to brake very hard, possibly injuring those inside in the process.
Would it respond to seeing a person in a certain position? If that one, you'd just get people finding out what it is and pretending so as to slow the traffic deliberately.

Another one that comes to mind is objects thrown from bridges. How would it react to that?
If you had someone hiding on a bridge who then threw a brick at the car, how would it react?

One more - lets say you have three lanes of motorway traffic, all autonomous and all running at 100mph closely bunched together. A car in the middle lane has a blowout, how would the computer react?
Truth be told, in that situation it doesn't matter how it reacts - by the time it has, due to the closely bunched traffic, there will most likely aready have been a crash and it will already have taken out the car to one side of it. Of course we then have the chain reaction of the cars behind having to slam on their brakes as hard as possible to avoid another collision, again likely injuring the occupants in the process.
Blowouts can't always be predicted, there may be a defect in a tyre that can't be seen or detected, or even just something on the road that damages the tyre enough to make it burst.

...it would likely still be cheaper than making driverless cars replace ordinary cars.

That's a point, who will foot the bill of replacing the ~34m cars on the roads of the UK and pay for the subsequent production of millions of autonomous cars?
Not to mention the environmental costs of disposal of old cars and production of new ones.
To implement fully, you'd hsve to either gradually close off some roads and make them autonomous cars only. But if you do the latter, what if someone without an autonomous car wants to go somewhere that can only be reached with a conventional one, or vice versa? And what about foreign tourists who currently bring their cars over on holiday?

I'm also against partial automation of certain things in cars as it is - like automatic headlights, because I am of the opinion that it causes people to get into the habit of not putting the lights on themselves, so they then get into a car without automatic headlights and drive off without the lights on, and I do see a worrying number of people driving at night with no lights on whatsoever (I think having dashboards that are lit even when the lights are off doesn't help in this respect).

 

[jon0844]

The BBC is reporting problems with air traffic control that is impacting on UK flights. This is nothing compared to the chaos if computer driven cars can't communicate or track other vehicles nearby for whatever reason. There are so many things that could go wrong, either by accident or deliberately. Vandalism will be a major problem, as people try and deliberately play with these vehicles.

Also, my car has the lights always on and new cars all have daylight running lights. It's now mandatory for all new cars, so eventually I think we'll see all cars with automatic headlights.

My parents have a car (Volvo V40) that has automatic full beam, which works perfectly by having cameras that can detect oncoming cars, or street lighting, a vehicle that is approaching head-on (turning on a strong red light on the dash and braking if necessary), reading road signs (and you can optionally set the car to drive no faster unless you floor it), follow road markings (but only by using some gentle control on the steering) and other things, including traffic approaching from the sides if you're at a junction or about to pull out of a parking space. Naturally, it also has adaptive cruise control where you can manually set the distance from the vehicle in front. I believe it also has a sensor to detect fatigue.

Thus proving my point technology can be used to aid drivers, not replace them!

There will be the concern that all of the above, and more, will make drivers less alert and switch off - but I still think there's a giant leap from that to a car that takes over everything.

 

[90019]

Also, my car has the lights always on and new cars all have daylight running lights. It's now mandatory for all new cars, so eventually I think we'll see all cars with automatic headlights.

I like DRLs, but I have come across a few issues with them.
The worst one was a few months ago when I was driving down the M6 at night, and a nearly new Seat Ibiza came past. His DRLs were lighting the road ahead quite well, but because of that, he hadn't turned his lights on, meaning the rear of the car was completely unlit. A few people (including me) tried flashing their lights at him, but I'm not sure he was aware that he had no lights showing at the back.

My V40 also has the lights always on, though I do know how to turn them off, and currently have them off. I run with my lights on a lot of the time, but I'd rather I made the decision to put them on, not the car.

Thus proving my point technology can be used to aid drivers, not replace them!

There will be the concern that all of the above, and more, will make drivers less alert and switch off - but I still think there's a giant leap from that to a car that takes over everything.

That's the problem I have with some of the current technologies - making people think they don't have to pay as much attention, because the car will do certain things for them.

 

[jon0844]

Having been in Sweden recently where it was always the law to have dipped lights, it seems the EU has actually made things go backwards.

Because of the new rules also including energy efficiency, as part of the Kyoto agreement, you have DRL running with LEDs - but only at the front. Thus, cars no longer have rear lights on - which is seen by some people in Sweden as a backward step. I don't think it's necessarily a problem, but it will take some getting used to.

And, you're right, speaking to my father-in-law, it seems that on his new car (Skoda) he no longer has the main lights coming on automatically, so at night he now has to turn the lights on himself - after years of being used to them being on!

I bet there will now be many drivers (who have never needed to worry about such things) getting caught out at night.

I'd have thought all new cars would also need sensors to turn on the main lights automatically, given it's a pretty cheap feature and has been available for many cars for ten years or more.

Do that and that problem goes away.

 

[90019]

Having been in Sweden recently where it was always the law to have dipped lights, it seems the EU has actually made things go backwards.

Because of the new rules also including energy efficiency, as part of the Kyoto agreement, you have DRL running with LEDs - but only at the front. Thus, cars no longer have rear lights on - which is seen by some people in Sweden as a backward step. I don't think it's necessarily a problem, but it will take some getting used to.

It's annoying, because I do now see quite a few people driving around at night with just their DRLs on, presumably unaware that they only have lights on at the front.
Of course, if you're driving towards them, you don't know that until you've passed themm so nobody coming towards them flashes their lights, and anyone who flashes their lights behind them will often, in my experience, be ignored.

And, you're right, speaking to my father-in-law, it seems that on his new car (Skoda) he no longer has the main lights coming on automatically, so at night he now has to turn the lights on himself - after years of being used to them being on!

I bet there will now be many drivers (who have never needed to worry about such things) getting caught out at night.

I'd have thought all new cars would also need sensors to turn on the main lights automatically, given it's a pretty cheap feature and has been available for many cars for ten years or more.

Do that and that problem goes away.

Well, until the sensor fails, the lights don't come on, and you get people driving around wondering why the road in front of them isn't as light as it usually is (if they even notice in an urban area).

 

[jon0844]

I had sensors on an old Mondeo and it never failed. You could still turn the lights on manually though (the auto mode was a different position).

 

[JohnCarlson]

Sudden mechanical failure is incredibly rare. In the majority of cases, the signs are there - we're just not able to pick them up. A computer system, OTOH, can report to its driver "Something doesn't feel right, I'm slowing down. I'll check myself into a garage while you're in work." It could then transmit telemetry to the garage, and remote diagnostic software could figure out "This looks like an electrical issue, best arrange it for when our sparky's in." F1 already uses computer-assisted telemetry to diagnose faults while the car is moving at up to 200mph. This is only going to get better and cheaper.

What people here aren't getting is that there are no technical reasons whatsoever why it can't or won't happen. All the rubbish you could make up about how it won't get out of a junction or how it handles at high speed or whether it can stop is exactly that - rubbish. There's more than enough time for things like that to be worked out. Most of the pieces are already there.

The one place where there are problems to work out is in the ethical sphere. If there is an accident and someone is injured, who is responsible? The owner or the manufacturer? What if the safeguards fail? What if they're tampered with? What if the owner hasn't been diligent with their maintenance? Will the whole thing result in "drivers" disconnecting from the reality of their travels?

There's simply no point arguing over the technology. It exists, and it's only a matter of time. The real question isn't over whether we could have them, but whether we should have them..

This reminds me of an old joke. During WWII the Germans realised that thy had a shortage of butter but mountains of horse manure. (I have cleaned this joke up a bit.) So Hitler tells his technical guys they have got to find a way of turning manure into butter. Devoted to their Fuhrer the boffins get to work. Six months go by and Hitler is getting a bit impatient. He phones up the guy in charge of the project and asks what is going on. The guy tell Hitler, "Don't worry the theory is sound, we're just working out how to perfect the system. Phone back in six months." Reassured Hitler says goodby and puts the phone down.

Six months later he still hasn't heard anything else so Hitler phones back again. Again the guy in charge of the project tells Hitler, "Don't worry the theory is sound, we're just working out how to perfect the system. Phone back in six months." Getting a bit impatient Hitler asks, "Look this is really important. How far are you on with it. Is the butter edible yet."

"Were fifty percent functional the man tells Hitler.

"Fifty percent functional. What the hell does that mean," shouts Hitler.

"Well" says the man, "It still looks and tastes like horse manure but we have got it to look yellow and spread very well."





For an update on car developments check out below.


http://www.youtube.com/watch?v=2hGjl3zcJMk