On the subject of counter-intuitive effects in traffic engineering, it also seems that adding more lanes to a highway actually makes traffic flow worse. Whenever a car changes lanes on a highway, a small traffic "shockwave" propagates backwards some distance, and more lanes -> more cars crossing more lanes -> more traffic.

Anecdata: I live in DC with its notoriously bad traffic. The highways 395 and 495 have 4-6 lanes each and tons of traffic. On the other hand, the George Washington Parkway has just 2 lanes and carries a large volume of vehicles between 395 and 495, but has very little in the way of its own traffic (except in the event of an accident, of course). Traffic only really occurs on the GW at the junctions with 395 and 495 since the X95s don't have the capacity to receive all of the GW cars at-speed. Whereas the X95s will get random backups for seemingly no reason, and at all hours of the day, backups on the GW generally only occur at the ends.

Playing "Cities: Skylines" on PC and growing your small village into a metropolis will teach you a lot of things about traffic. More options for cars are usually not good for traffic flow.

The drivers in that game don't crash, and otherwise their behavior is greatly simplified, but each vehicle in that game has a start and a destination, and you can follow delivery trucks as they deliver packages and then go back to the Depot to pick up more. It's very fascinating, and may offer us some small window into the kinds of things that traffic engineers deal with without trying to find or write traffic simulation software.

And, of course, there are mods to give the player (the Mayor) greater control over lane usage and where stop signs and traffic lights are placed and how they operate. Doing things wrong will quickly result in localized gridlock.

The biggest flaw in C:S traffic flow, even with the popular traffic improvement mods, is that routes and lane choices are pre-calculated when the vehicle departs and don't really take into account alternative routes/lanes or current traffic load very well or at all. You quickly run into issues where you'll have a 3 lane road that has 99% of the cars backed up in one lane while the other 2 are wide open. Coincidentally it kind of mimics an induced demand problem, but for the wrong reasons.

Of course having every vehicle continually recalculate and optimize its path is computationally prohibitive. Given the limitations it does a pretty good job, though!

The vehicles aren't situationally aware, you're right. Nothing beyond basic "go when the car in front of you goes, and stop when they stop" kind of thing.

That said, there is a non-zero percentage of human drivers that do the same pre-calculated route and lane selection.

I found that frustrating too - all the cars piling up into one lane. I'm sure there's some stuff they could have done to pre-calculate and have some "randomness" in there to use the other lanes. There was a plugin, might still be, that kind of fixed it. Maybe I should fire it back up and see whats all new.

Isn't this how most of us operate? I almost always take the same route to work, regardless of traffic.

Route, perhaps (though I personally like to mix it up), but if the right lane of a 3-lane highway is backed up due to an offramp that isn't yours, I imagine you'd move over one lane to get past it! That's what doesn't happen in C:S, unfortunately.

You are absolutely right about the shockwaves. It is also weird that going fast and tailgating actually create slowdowns and make traffic go slower. By not leaving room to slow down, any minor lane change can cause a huge ripple of hard braking and now there is a backlog until everything accordions forward again. I wish newer cars had a constant beeping when you are 10 feet from the person in front of them at 65 mph. We'd actually see better traffic flow.

As for parkways, the better flow is because they don't allow trucks. Variance in the speed of cars is what causes traffic. Obviously this is true at on/off ramps. There is no way X95 can take two lanes of high-speed flow of cars into their already packed flow. For parkways with a lot more hills, you can still get slowdowns but these happen almost exclusively just before big hills and at exits to bigger highways (and occasionally at busy on ramps with short merge areas).

People always talk about driving with gaps[1], but I've yet to see a simulator show how detrimental/beneficial these strategies are to individual drivers.

It feels like a hawks and doves style ESS (evolutionary stable strategy) problem. It appears to me that the strategy that eliminates the overall negative affect for the group (avoid creating phantom stop signs, by gapping or "always in the middle" strategies) is necessarily unstable.

The benefit to any single driver who does not follow that rule becomes increasingly high as more and more others do follow the rule. Even to the point of ignoring a beeping car I should think. I certainly feel the negative effect driving behind someone who creates a gap as more and more cars merge in front of them. In the pathological case the driver who insists on a safe interval between cars will make no progress at all.

1: (CGP Grey) https://youtu.be/iHzzSao6ypE

You say you've observed it, but it doesn't make any sense. If you go 5mph slower than the speed of traffic, you will continuously open up a gap. You never need to slow down farther than that. The cars merging into the gap wouldn't be merging if your lane wasn't faster than their lane, so there's no reason you would have to slow down even farther. What you perceive as repeated slowing down will only happen if you sped back up to match the speed of the car in front of you after opening the gap.

I don't think you've considered what you are saying. You're not describing a gapping strategy, you're just describing a slower driver (i.e. slower traffic). There is no question that a car consistently driving slower than ambient traffic speed is a net drag on overall throughput. In any single lain of traffic all cars are bound by the slowest car. If you drive 5 mph slower than the car in front of you then the car behind you must drive 5 mph slower still to continually produce gaps behind you. As the video illustrates introducing a slower driver is one of the things that creates phantom stop signs.

Of course you must maintain speed. The point of a gap is to allow for adjustments without the hard breaking that would create a phantom stop sign. That is the overall intent of the system. Move the most people as expediently as is safe from point a to point b. The best way to do this is to avoid speed differentials, being slower then ambient traffic speed creates a speed differential.

That's not true about the second driver behind me, because people are not pulling into his gap, because his lane is going 5 mph slower than the other lanes. So he establishes it once and is fine.

This is more of a thought experiment to show that you never have to keep slowing down to zero even if for some reason people continually enter your gap. I barely observe people entering my gap at all in practice, because all lanes are generally going the same speed so there's no benefit to it.

Because you insist on the gap, your lane is 5mph slower than the other lane (because people pull in front of you after passing). Assume the other lane is doing 50 (65 mph speed limit, but there is congestion).

That lowers the throughput (number of cars that complete a trip in your lane) by ~ 10%. Also, the faster lane is now probably moving closer to your speed behind you than the 50mph you observe (because you are causing cars going 45mph to cut off cars going 50mph that lane).

The arrival rate of cars won’t change because you chose to go slow. So, if the road can handle 10,000 cars per unit time with normal congestion, it can now handle ~ 9000. That means per unit time, there are 1000 new cars sitting in the stop and go backlog that you single handedly created by creating the gap.

This has been shown time and time by traffic simulations. It only takes a small handful of slow drivers to create massive backups.

Also, the number of traffic fatalities is proportional to the difference in speed between the fastest and slowest car, so, statistically speaking, by being the slowest car, you are murdering people.

The reason I am going 5mph slower in this thought experiment is because a continuous stream of cars is merging in front of me from the other lanes. (What the OP called the 'pathological case'.) They are the reason the lane is slower, and because they are not in the other lanes, those lanes will be 10% faster.

With the pathological case being so benign, the real life experience is even better. You might get a car moving in front of you once a minute, and usually a few seconds of slower accelerating or coasting is all you need.

To me the thing that really slows down traffic is people not moving left to let onramps merge. I think that's because they are afraid of missing their exit because no one is leaving gaps.

> I wish newer cars had a constant beeping when you are 10 feet from the person in front of them at 65 mph. We'd actually see better traffic flow.

I just drove a Ford Edge that had this. It is a little overeager about what it thinks is an impending collision, but it works alright.

The adaptive cruise control is a mess though. In theory, it is supposed to match your speed to the vehicle ahead if they are slower, so you don't have to pay attention. In practoce, it has the potential to be a great contributor to congestion. With dumb cruise control, you can push slowpokes to get out of the way or at least speed up; the adaptive cruise control seems to match speed at a distance where the driver ahead doesn't get the signal that they are obstructing things, and they continue bumbling along.

For highly congested areas the increase in the number of lanes is done to accommodate the number of cars on a given route (optimize for capacity). If you can physically fit four times as many cars on a route, but it results in only being able to move at 1/3 normal speed you are still moving more cars over time than a two lane road.

It's not random, it's poorly engineered entrance and exit ramps. In order to maintain traffic flow, cars need to be able to increase to traffic speeds on on-ramps and decrease on off-ramps. Much have the beltway system put entrance and exit ramps where they definitely don't belong. Thoroughfares for in-city traffic also don't exist, the interstate has taken that role. To compound the problem, often times entrance and exit ramps are the same piece of pavement.

I think a lot of the counter-intuitive effects in traffic engineering are due to traffic arbitrage. If you make a path much better more people will take it until it's as bad as all the alternate routes.

It can actually be worse. This is called Braess's paradox [1]. The relevant part is: "an extension of the road network may cause a redistribution of the traffic that results in longer individual running times."

Further down in the Wikipedia article is an actual example.

[1] https://en.wikipedia.org/wiki/Braess%27s_paradox

If a lane is shut down due to an accident, it is better to have five spares though.

> This suggests to me that to make roads safer, we might want to design them so that drivers feel they need to drive slower and more safely.

Yes: https://en.wikipedia.org/wiki/Traffic_calming

I read an article long ago about incorporating optical illusions into traffic calming. Things like the shape of ditches and the placement of trees on highways and parkways to make the shoulders look narrower than they really are.

Making the road appear more dangerous than it really is makes people slow down and pay attention, without the uptick in danger of an actual hazard.

"Let's train people to ignore danger"

what we need is red camera* at every light and speed trap at every road, only the fear of enforcement keep people honest

*minus the whole shorten yellows for profit, of course

Surely we could more or less eliminate all kind of law violations by total control. But I don't think this is what we want as a society.

I'm not arguing about every single law, only those that have impact on the leading cause of death beyond health issues in the civilized world

but ok, nice slippery slope, I guess.

Let my rephrase it in a better way I hope.

I'm not trying to say that once we introduce methods to fully detect some types of law infractions it will inevitably lead to a totalitarian regime.

But I believe some leeway is necessary for people to feel comfortable and somewhat free. I don't want to be punished every time I make a mistake. I think you can find some sweet spot between zero and total detection. And yeah, probably leaning towards the latter, in correlation with seriousness of a violation (optimal discoverability reaching zero rather means the law should be revoked, like jaywalking being penalized in my country imho).

Also, how would you even know some law is optimal when no one violates it? I mean, we could put radars on every road and set the limit on 30 kph. The studies discussed in the article wouldn't be possible then.

> I'm not arguing about every single law, only those that have impact on the leading cause of death beyond health issues in the civilized world

Definitely speed is the factor. But what about other variables at play like road quality, intersections planning, lights, other traffic regulations? And if tweaking them actually lowered casualties equally well with better economical outcome?

> only the fear of enforcement keep people honest

As for a slippery slope. You vastly generalized your statement yourself tbh.

Similar to a phenomenon in sports - more and better protective gear leading to more aggressive play leading to more injuries instead of the expected less.

AKA "The Peltzman Effect" - https://en.wikipedia.org/wiki/Risk_compensation#Peltzman_eff...

Was too busy to cite sources at the time, but for posterity:

https://www.nytimes.com/2017/01/06/opinion/the-dangers-of-sa... https://www.usatoday.com/story/sports/ncaaf/2013/07/30/concu... https://sports.cbslocal.com/2015/07/03/lemco-why-football-ne...

> This suggests to me that to make roads safer, we might want to design them so that drivers feel they need to drive slower and more safely. Narrow the lanes where they are too wide

anecdote about the opposite happening:

There is the I90 tunnel in Seattle which I use very often. Some passages have 2 lanes, super narrow, no shoulder, nothing. Looks like this: https://goo.gl/maps/DzvhWZAoD5AkRVwA9

You won't believe how fast people are going through that. Also, almost every time I get tailgated and I'm frankly amazed there hasn't been any major accidents in that tunnel considering how close people are driving (no following distance). Just yesterday I was tailgated by somebody on their phone (I could see the screen reflected in his face while he was texting). I don't want to imagine how an accident would look like in that tunnel. Honestly, a fender bender can become catastrophic before you realize what happened.

Now if you mean 'narrowing' roads then probably yeah. Take a wide-ass lane and suddenly make it narrow (so that people using it every day are 'surprised' and they're more careful) and I could agree with you.

More like unintuitive rather than counterintuitive, because the prediction doesn't fail because wider lanes are more dangerous per se, but because of the behavioral response of the drivers. Which can actually be considered "intuitive" to someone who is used to trying to model such things.

Where I live in the south of Germany nearly every residential area has these artificially narrowed roads and it works a lot better than putting up a incongruent signs ie 30 on a 50 looking road.

It also can look pretty nice when they use flower beds and trees for the narrowing:

http://www.ibfunk.de/ibfunk-wAssets/img/leistungen/verkehrsa...

I enjoyed the survey that the book _Traffic_ by Tom Vanderbilt provided in the traffic engineering space.

Many of the topics discussed in the comments I see here are discussed in the book.

I don't know how many time roundabouts and gps have caused near dangerous split attention that just does not happen with normal stoplights. They may be better for local traffic but for anybody from out of town they are worse.

>Roundabouts reduced injury crashes by 75 percent at intersections where stop signs or signals were previously used for traffic control, according to a study by the Insurance Institute for Highway Safety (IIHS)

https://www.wsdot.wa.gov/Safety/roundabouts/benefits.htm

Accidents happen at intersections because many drivers are not good at making themselves aware of traffic approaching from multiple directions. At a roundabout drivers only need to look in one direction. They are strictly better than traffic lights for all drivers except those who are not experienced with roundabouts. Those drivers should familiarize themselves with roundabouts.

Of course it's always possible to screw up a good thing: the intersection next to our house has been replaced by a roundabout with a 2-way stop sign. To make it worse, you can really only see the traffic coming from one of the other 2 directions when you are stopped.

Result (people not being familiar with roundabouts doesn't help): people stop in the middle of going around the roundabout, and/or get honked at by the people coming from the 2 directions without stop signs. I'm sure an accident is coming soon :(

this sounds like some traffic calming measures that were recently installed near me just last year

they are islands in the middle of the intersection, but not roundabouts, forcing cars to "swerve" to pass through

they haven't worked as intended and are being re-evaluated

Roundabouts are better for car-car crashes: there are fewer of them and they tend to be less serious— rear-enders where there's lots of crumple protection instead of T-bones like a conventional intersection.

However, they're much, much worse for people walking and cycling.

I agree that roundabouts and sidewalks do not interact well. Cyclists should take the entire lane in a roundabout, because otherwise cars will not notice them. Pedestrians should be very careful.

Depends on the design. See https://bicycledutch.wordpress.com/2015/10/13/explaining-the...

There's definitely a design component there, but I do wonder how much of it is in culture and enforcement (which can change the culture over time).

For example, the roundabouts in my city (Kitchener ON) all have a sidewalk or multi-use trail crossing each vehicle entrance and exit, and without fail, every single crossing is marked with a paint ladder, stopping teeth, and a "Yield to Pedestrians" sign. I've also confirmed with city engineers and law enforcement that people walking are supposed to have right of way at these crossings— like, you're waiting to cross, and drivers see you and stop, and remain stopped until you're fully off the roadway.

But it's not like that at all. I have dozens of helmet cam videos of me walking my bike through these crossings and being buzzed by cars who don't see me at all, or don't see me until it's too late, or see me but don't think they need to stop for me. I send these videos (many with clearly visible license plate numbers) to the local media and contacts I have at regional police, and I get a collective shrug back.

I met with city engineers in the winter to ask whether they thought it was safe, and they said that yield compliance to roundabout crosswalks was 70-90%. They said this like it was a good thing. Literally one in ten cars doesn't stop for a person walking in a marked crosswalk, and that's somehow acceptable.

I'm not sure how to change a culture, but camping some cruisers there for a week and handing out a few hundred tickets for $1000 each [1] to non-yielding drivers sure would be a great start.

[1]: http://www.mto.gov.on.ca/english/safety/pedestrian-safety.sh...

Lived in the UK most of my life, roundabouts and walking/cycling was never a problem more than crossroads. Don't know what you're on about.

One of the fundamental advantages about roundabouts is that even when crashes do occur (and all the studies show they happen less frequently), they're less likely to be fatal. Instead of being t-boned, with at least one of the vehicles driving at speed, you're getting either side swiped or rear-ended, which is safer for all drivers involved.

I miss roundabouts whenever I drive in North America. I'm not used to intersections (which could be the problem) but it seems like you have to look everywhere as opposed to just looking in the direction traffic is coming from.

Could put a big spike on the steering wheel