Math Says You're Driving Wrong and It's Slowing Us All Down (wired.com)

A new study in IEEE Transactions on Intelligent Transportation Systems mathematically suggests that if you and everyone else on the road kept an equal distance between the cars ahead and behind, traffic would move twice as quickly. From a report: Now sure, you're probably not going to convince everyone on the road to do that. Still, the finding could be a simple yet powerful way to optimize semi-autonomous cars long before the fully self-driving car of tomorrow arrives. Traffic is perhaps the world's most infuriating example of what's known as an emergent property. Meaning, lots of individual things forming together to create something more complex. Emergent properties are usually quite astounding. You've probably seen video of starlings forming a murmuration, a great shifting blob of thousands upon thousands of birds. Bats flying en masse out of a cave is another example, swarming sometimes by the millions through a small exit. And scientists are just beginning to understand how they do so.

Comments

Merge problem

[religionofpeas]

if you and everyone else on the road kept an equal distance between the cars ahead and behind, traffic would move twice as quickly.

Yes, because no one would be merging into traffic anymore.

Re:Merge problem

[geekmux]

if you and everyone else on the road kept an equal distance between the cars ahead and behind, traffic would move twice as quickly.

Yes, because no one would be merging into traffic anymore.

If everyone kept an equal distance and followed a standard merging pattern of every other car, then it would likely solve the merging problem as well. Long ago we were shuffling decks of cards in a much less practical and inefficient way until certain physical moves were found to increase that efficiency ten-fold.

It's also well-known that impatience creates stop-and-go traffic patterns, which is but one of the many human factors that autonomous solutions will be looking to solve.

Re:Merge problem

[religionofpeas]

And a slight speed adjustment allows the distance to return to normal after the merge

The problem is that a road runs at maximum capacity when speed is high and distance is minimal. A slight speed adjustment, like you suggested, has the effect of decreasing maximum road capacity.

That means that the road after the merge point not only has to deal with more cars, but also with a lower capacity to carry those cars. This lower capacity will propagate backwards to the road before the merge point. And that's how you end up with a traffic jam.

Re:Merge problem

[PPH]

That's why we have metered on-ramps. They limit the number of cars that have to weave into the traffic. And if traffic on the freeway starts to slow down, so should the metering rate until the bottleneck relieves itself.

But if metering were actually implemented this way, people waiting would go insane. And it doesn't account for the carpool bypass, which ends up jamming traffic at the merge point anyway. In reality, ramp metering has become a penalty* for solo drivers, nothing more.

*Unless you are fortunate enough to live in a wealthy neighborhood and can call your state representative to keep metering off your own little local on-ramp. Why the hell do rich people get their own on-ramps anyway?

No clickbait headlines

[Kohath]

Math says you're treating Slashdot readers wrong and it's making the internet worse for all of us.

In other words, there's an optimal distance.

[hey!]

That's not surprising. Spread cars out too much and you reduce the roadway's capacity. Put them too close together and you have to slow down to accommodate the driver's minimal reaction time. Having every driver choose his own distance means you can have both effects simultaneously: wasted space and insufficient response time.

Put all these constraints on and it seems obvious that you want to space cars uniformly with the minimal distance consistent with whatever statistical level of safety you demand. Naturally robotic systems will be more efficient since they require less response time -- in fact they can react to events that will cause the car in front to slow.

What would be interesting is to see the exact results they came up with: how far for how fast and under what conditions? What are the significant input parameters of the model? For example I'm sure varying the acceptable probability of a crash has a powerful effect on the optimal distance.

Re:In other words, there's an optimal distance.

[Drethon]

I always try to have the distance to the car ahead of me set so I never have to hit the brakes and when the traffic does slow up, not having more than the usual 2 second gap just before it speeds up again. I feel like this results in me maintaining the highest speed possible. Of course people usually cut in front of me, so I have to slow down more than I would otherwise.

This is only for freeway traffic, city streets and inconsistent stoplights are a whole other ballgame.

Re: This doesn't work, although it might

[religionofpeas]

Don't drive closer to the car in front of you than security dictates, even if there is a tailgater behind you.

You should do exactly the opposite. If someone tailgates you, leave more distance in front of you so you can afford to brake slowly, giving the person behind you more warning time.

Set adaptive cruise control

[OFnow]

Just set your Tesla (or other modern) adaptive cruise control to, say, five car lengths, and just steer. It is far far easier than having to brake/accelerate and the hardware watches even when the driver has zoned out. No worries about hitting the idiot in front. No worries if someone merges into your lane: the car adapts.

You can't fix a system by changing *everybody*

[RobinH]

I work in a manufacturing environment, and changing even a handful of people's behavior is so incredibly difficult and costly ("always pick up one orange nut at a time, then the blue nut, don't grab two at once.") that asking everyone to change the way they drive is just ridiculous. You have to change things so that the desired behavior is the easier behavior. For instance, advanced cruise control that adjusts your distance automatically might be a solution. In our plant, if the process says they should do X before Y, then the only way to ensure it actually happens all the time is to prevent Y from happening until there's proof X happened. People just aren't reliable.