The "car train" revisited.

Back in July Sam Stone created a thread devoted to the idea of what he called “the car train” and provoked discussion regarding wireless communication between cars and ad hoc “platooning” of vehicles.

Now this update:

Okay, he’s not the most persuasive presenter of a case but it seems like a good reason to revisit the issue.

Why is he making the case right now?

Obama has already committed to moving in the direction of the electrification of the transportation infrastructure - supporting the concept of extended range electric vehicles (EREVs) and PHEVs strongly. Will he also commit to the connection concept that Larry Burns is promoting? Should he?

I assume he means connected vehicles in the information sense, and not physically connected. I think he’s thinking more along the lines of intelligent routing, avoidance of traffic jams, proximity sensing of other vehicles, etc.

If you knew the location of every car in the city, you could work out optimum paths for each car to take to lower the overall congestion the city and balance out road use. If you knew the optimal paths, you could set prices on the roads dynamically and inform each car of the pricing. Your car could ask you, “Do you want to take your regular route home? CarNet says that your regular route will take 25 minutes and cost 1.20 in road charges. CarNet recommends an alternative route which will take six minutes longer, but save you .40 in road charges.”

Plus, if you knew where every car was, you could greatly improve the efficiency of traffic lights. You could delay traffic to prevent jams, keep the lights green in one direction if no traffic was coming towards the side streets, detemine when and how long to turn on left-lane turning signals, etc.

I imagine there would be a lot of efficiency gains in a truly smart car network.

Yes, connected in the information sense, but by the end of that thread that’s where the conversation was going with ad hoc “platooning”. That is a bit more than GPS routing with traffic and toll information built in. It is, to hear him tell it, so much a safety improvement that it would allow vehicle weights to come down. Not sure I buy that bit.

Yeah, neither do I. Computers aren’t going to stop cars from hitting things. They might reduce some kinds of accidents, but there will still be plenty of accidents.

And wait until the screaming starts over the privacy implications of requiring your car to be trackable every moment it’s on the road.

Oh good lord. I can’t even begin to think of all the problems with this screen.

I disagree, but it would require computers far more sophisticated than are available today.

A computer can’t protect you against a tree falling in the road, or the combination of a windy day and an icy road blowing your car into the ditch, or a child running in front of the car, or a road hazard like a pothole or some foreign object on the road. And if these things can possibly be run on manual control, it won’t prevent you from turning it off and being an idiot anyway, or being hit by someone else who’s being an idiot.

Any time millions of people are being moved around at high speed, there will be accidents.

I don’t have a cite for you, but my sense is that the vast majority of accidents involve some kind of driver error. For example, I would guess that most people who get blown into a ditch on an icy day were driving too fast.

Of the remaining accidents, I would guess that a lot could have been prevented with a superhuman driver. For example, the other day I drove into a small Christmas tree which somebody had left lying in the road. I was delayed in seeing the tree because I had just started a lane change so I had momentarily looked to the side to make sure the lane was clear. There wasn’t enough time to safely swerve. A sophisticated computer would be able to watch forward, back, left, and right simultaneously and would also have much faster reactions.

I do agree that one could not prevent all accidents, though.

I imagine that once computer control gets good enough, it will be required just like other safety equipment.

I agree with this, but I think that it would be quite rare under the circumstances I described.

The problem with you thesis, brazil84, is that you assume that computer control is like human control, only with no flaws. Computers, unfortunately, have more than a few of their own.

I’ve had ideas along these lines too. Why not make it really simple and just put large cushions on the front and back ends of cars, and then have them all drive bumper-to-bumper (literally) in cruise control? I suppose you might need to make the brakes communicate wirelessly so that the head car’s brakes control all the others.

Not exactly. My claim is that computers can be made to be a lot more reliable than humans. For example, consider the problem of adding a collumn of 1000 numbers. A computer can be programmed to do this task perhaps a 1000 times more reliably than a human.

I anticipate you will object that controlling an automobile is a much more complex task than simply adding numbers, and of course you are correct. Indeed, a modern computer cannot handle this task even in an unreliable fashion. However, it is my belief (and hope) that in the future, computers will be sophisticated enough to handle such a task.

It’s not the computer I worry about - it’s the software. Train accidents have happened because software glitches put two trains on the same track at the same time. There have been a number of airplane crashes due to computer error, such as the Airbus A320 that crashed on a demonstration flight because the pilot managed to put the aircraft in an attitude the computer wasn’t capable of handling.

I think you also underestimate the willingness of people to put themselves in the hands of automated traffic control. Cars today have computer control of sorts - electronic stability control, etc. The first thing most reviews tell you is whether or not you can shut it off.

Mercedes was demonstrating its automatic sonar braking system for the press, and the demo car didn’t break and just slammed into the car in front of it.

For these and many other problems, you will not see a hands-off computer-guided auto network for decades, if ever. The problem of automatically guiding millions of cars in the real world where there are mechanical failures and unforeseen obstacles and other events is overwhelming.

Now, a computerized traffic control system that advises drives is another matter. I could also see something like a computerized accelerate-decelerate mode which always drives the car at peak efficiency, but which can easily be overridden like cruise control. There are a host of improvements computers can make to cars and the way we drive them, but we’re not going to be strapping into auto-guided capsules any time soon.

I think what we may see is a sort of decentralized, semi-hands-off driving system where each car automatically follows the road and maintains a safe distance from other cars around it. A centrally controlled system would be very hard to implement and probably too constraining to the drivers.

The point of a car train is that it helps traffic. I always thought the biggest cause of traffic is when cars up front start accelerating, but you’re still sitting there in park. If everyone would just step on the gas simultaneously, the traffic jam would instantly dissolve. (And as cars travel faster, their separation would increase due to bernoulli’s principle, and bumper-to-bumper at 0mph would turn to a normal flow at 70.)

This. No matter how good a computer’s reflexes, the fact that it can’t improvise is a BIG downside for computer-controlled vehicles. Now, we’re certainly making leaps and bounds in this regard, and under certain conditions, I think it would be possible to get a computer system that crashed less than a human. The problem is, crashing less often than people is still crashing a LOT. So you’re not gonna be able to get rid of all the safety features, so no weight reduction, etc. He’s building a chain of effects when the first premise is pretty shaky.

The other big problem is, how exactly would you gradually implement the system? If it’s a centrally-controlled network, EVERYBODY has to be on it, so either you mandate people buying controllers (yeah, right) or the government shells out for them (ahahahah!). Even the less-pervasive info system would only work if it knew where all the cars in the country were at all times (plus bikes, in cities). It’s gonna be a looooooooong time before anything like that would work.

It’s the same problem as social networks have - Facebook isn’t so hot unless a bunch of people are already signed up. This kind of info network would be worse, because it requires TOTAL penetration. It’s not something that can happen gradually.

What I see happening is more an extension of the current traffic-cams that some news sites have. As motion tracking improves to the point where a computer can recognize what’s in the street, I could see cities setting up camera nets to track traffic. You’d probably start with really nasty traffic areas, and over time it could gradually cover the whole city, so that you could first get an idea of where traffic is bad, and then eventually maybe (I say maybe) car manufacturers would start including cruise control systems that could interpret the data and react to the road, traffic lights, and cars nearest you. Even that I don’t see anytime soon, because of problems of format and standardization, not to mention the problems with installing soon-to-be-obsolete (and all computers are soon-to-be-obsolete) in relatively long-lived cars. But, at least it’s a little more organic than some central control system.

… then every car would maintain a constant following distance equal to the distance between the cars when stopped. If one car applies it’s brakes, then every car behind it must apply it’s brakes at the same instant and at the same rate, otherwise, an accident will occur. There is very little room for error in this arrangement: Not only do the computer systems need to communicate with very low latency, but the braking systems need to be calibrated and maintained to exacting standards.

Actually, if everyone stepped on the gas simultaneously, there would probably be fender-benders due to differences in acceleration between different cars.

When I refer to “computers,” I am talking about the whole package. Not just the hardware.

So what?

Once computers are clearly much safer and better than human control, it will probably be mandated.

I’m not sure what we are arguing about. I am talking about computers of the future, not computers of today. Today’s computers are clearly not up to the task.

I think his point is that you’re never going to get a 100% failure-proof system. To work out every bug possible, in every situation possible, for a system that complex would take so long that it’d be obsolete by the time you were done. If it were possible at all.