[morley]

Whenever I see an article about traffic patterns, I inevitably see this exchange:

Person 1: You can fix traffic waves by smoothing!

Person 2: Smoothing traffic waves makes no sense!

It's frustrating. I don't know what to believe. Is there any peer-reviewed, simulation- and data-backed research that puts this issue to bed for good?

[ninkendo]

I think the sad truth is that it doesn't matter how you drive, traffic is going to suck.

Given that everyone requires a safe lead time of (for example) 1-2 seconds, the more cars there are per mile of highway, the slower everyone must drive. (You can't drive 80 miles per hour bumper-to-bumper.) So if density is say, 50% (one carlength of open space per one car), you have to drive a speed such that a carlength is 1-2s, in other words 10-20mph.

But traffic distribution is not uniform, there's exits and entrances, and cars do occasionally need to change lanes. If traffic is going 10-20mph with 1 carlength of space between cars (steady state), and I change lanes, the guy I merged in front of now has to slow down more to leave more room, and this will cause a traffic wave behind him. What happens at an exit when half the cars change lanes? Standstill. No change in driving technique on anyone's part will help this.

I think the only times where the way you drive matters is when the density is kinda sorta high but still low enough for a safe following distance at reasonable speeds, at which point "smoothing out waves" becomes a common sense matter of "don't follow so close", which is effectively the same thing, and something everyone should be doing anyway when the density is low.

[btilly]

You are thinking about traffic in terms of cars per foot. The right way to think about it is in terms of cars passing a spot per second.

A lot of traffic problems stem from the fact that you get more cars per second from quickly moving smooth traffic than from slowly moving dense traffic. Therefore if something bad happens to rapid smooth traffic, you quickly get a phase transition into horrible traffic.

Now what does the transition from horrible traffic back to good traffic look like? The first step is that you have to get back to smooth traffic, and then that has to speed up.

If nobody smooths out traffic deliberately, then you don't get this until the volume of people wanting to pass gets so light that all of the stop spots "evaporate" on their own because there is no pressure on them. Then the road speeds up. By contrast if traffic has been smoothed, the bottom speed steadily increases, and you get back to full speed much earlier.

Plus smoothing traffic will save on your own car's wear and tear.

Oh, and on merging? Go read http://ops.fhwa.dot.gov/wz/workshops/accessible/McCoy.htm. You will find that in the real world when drivers are instructed to merge late and politiely, merging happens significantly faster, with far fewer conflicts between drivers. So how you merge actually does matter. A lot.

[catbird]

Yes, there are many, many papers about this in the civil engineering literature, going back to the 1950s. Cellular automata models, fluid flow models, you name it. If you're really interested to learn more, search google scholar for "traffic flow," "traffic wave" or similar.

Even the wikipedia article on "Traffic flow" is quite comprehensive. [0]

It really annoys me when people come up with a "theory" and write a blog post about it without doing a shred of actual research on the subject.

[0] https://en.wikipedia.org/wiki/Traffic_flow

[mixmastamyk]

The smoothing doesn't increase the road capacity, though there is likely a small benefit to mitigating the negative effects of merging.

The big benefit, which the grandparent concedes, is the reduction of unnecessary fuel usage, wear and tear on the vehicle and driver.

So, I consider it a win regardless.

[wbeaty]

The papers I've seen on this conclude that rear-end accidents are the major issue. Smoothing the waves can eliminate immense rush-hour delays, if this eliminates panic-braking that can cause minor daily accidents. Or as they say, "speed differentials are dangerous." In I-70 Colorado they claim that police pace-cars used to smooth out the fluctuations have cut the accident rate in half.

In theory the waves can represent a bottleneck, since the close-packed regions are low-flow, but the wide empty regions are also low-flow. If the smoothed wave results in traffic flowing at 35-40MPH, then smoothing will increase the flow, since the peak flow rate takes place at 35-40 MPH. But the increase isn't enormous. I think they said ~15%. That's nothing, when compared to the effects of removing an accident scene from a rush-hour commute.

You can play with those online JAVA simulators: set up unstable conditions on the ring-road, watch the initial flow, then after traffic-waves develop, watch the flow again. If enormous waves appear, the flow drops by ~25%, but if a string of small waves appear, the decrease is less than 10%.