Watch traffic flow better in 30 simulations of a 4-way intersection


Originally published at:


Sure you can get good flow if you allow vehicles to pass through each other (as the simulation does).


30 different types of intersections…


30 simulations of a 4-way intersection

Yes. Many of them are the same design, with different mods. How many of us care how NExt improves traffic efficiency?

But those last few designs were crazy. I wonder how real-world construction costs compare with efficiency numbers?


Mythbusters did a section on this once, in which they too discovered that roundabouts are incredibly effective at throughput. I never really understood why the US didn’t have them (here in the UK, we even have painted miniroundabouts which work almost as well as full-sized block ones during periods of heavy traffic (whilst being routinely ignored when there’s no traffic.)


We have some in the northeast (although we call them rotaries). Most people hate them, because they bring out the worst in our already shitty drivers. It’s basically the most efficient way to expose you to the worst that your fellow asshole drivers can dish out: those who will cut you off and endanger your life to shave 7 seconds from their commute.

One problem is that because they are so infrequent, people don’t get the hang of navigating them. I can only imagine what would happen if you throw you average driver used to lighted intersections into a turbo roundabout (about the 2:30 mark in the video) where it appears each driver has to choose from 4 or 5 possible paths through the roundabout.


Think of it as smooth vs turbulent flow.


The biggest problem with rotaries* here in the northeast US is that they are just round-ish islands plunked down in the middle of an intersection. They’re not particularly infrequent here in the Boston area – so we’re used to them – but are still usually dangerous free-for-alls because they’re built so poorly and no one is clear on what to do. In the UK, the entry and exit geometry is very carefully engineered so that it’s difficult NOT to understand and flow along a clear route (this is assuming it’s a relatively newish road; there are certainly plenty of smaller, older roundabouts that haven’t been upgraded and are a bit of a mess).

*technically, rotaries and roundabouts are not the same thing, but that’s a distinction not relavant to this discussion.


This. This is the thing I suspect most USanians do not get about roundabouts, given their limited and often badly designed US experience.
People can drive quite badly on UK roundabouts and still survive and not endanger or piss off too many others, or clog the system up. The worst sin, for me, in multi-lane roundabouts, is the person on the outside lane (closest to the roundabout) who fails to position themselves for (or misses) their exit and suddenly cuts across everyone rather than realising they can just go round again and next time get in the inner (closest to the exit) lane in good time.


There are crosswalks in those videos, but no pedestrians. Roundabouts without traffic lights are great for moving cars but how do I cross the street if the cars never stop?

If you don’t put pedestrians in your simulations you’re optimizing for the wrong things.


Actually, I think the distinction is pertinent to this discussion. The simulation didn’t model a true roundabout, though every traffic circle they showed was labeled “roundabout”. The distinction is that a traffic circle / rotary has high speed entries and traffic in the circle must yield to incoming traffic; they can also have multiple lanes. A roundabout has slow speed entries (nominally enforced with yield signs) and traffic in the circle has the right of way. Roundabouts are higher capacity and safer than traffic circles / rotaries, but a true roundabout was not modeled.


L’Enfant designed them into DC roadways in 1791 and they have been providing insurance claims ever since.

There are many.


The roundabouts seem wrong, at least compared to UK. Traffic already on the roundabout was stopping to let cars onto the roundabout. We only do that if there’s traffic lights, which weren’t in the title, and I think there were no stop lines indicating they were stopped at lights. Wonder what the flow is for UK-style roundabouts.


There was a study several years ago which showed that at (European) traffic circles, total car/pedestrian impacts and conflicts dropped dramatically, due to a decrease in the overall number of potential conflict points. I think it was referenced on Wired. but this was around 2013 or 14 I think.

I was very happy to see the efficiency of Continuous flow and DDI intersections, validating for my previous job with the Utah DOT. A lot of the highway intersection renovations in the state have been phasing those in, and a lot of the public doesn’t believe they increase it by that much. But they do. And they have the traffic data to prove it. I’m still sensitive about it.


hhhhhnnnnnnnnnnnnnnnnnnnnnnnnnngggggggggggggg, fuck ya!


Some of those would be terrible to get thru if you aren’t familiar with them or don’t have GPS. On the other hand late at night, with the sport car, it would be fun to hang it out.


Maybe I haven’t encountered a properly-designed traffic circle. The new ones around here seem like a menace to foot traffic (which means pedestrians avoid them, which I suppose does probably reduce fatalities.)

It’s too bad we don’t have some kind of animated graphical representation of how pedestrians interact with traffic circles to correct my misconceptions :slight_smile:


That’s possible. And of course, simulations don’t account for human behavior or interaction. But here’s the Federal Highway Administration page regarding roundabout safety advantages.




So how do we get to see what happens for novel designs?


Did anybody notice in the first simulations left turn traffic was happily driving through each other? Was time spent dealing with the crashes, lifeflight landings, cleanups etc factored in to the efficiency?