Hacker News new | past | comments | ask | show | jobs | submit login
Why does traffic bottleneck on freeways for no apparent reason? (2009) (engineering.mit.edu)
14 points by fzliu 8 months ago | hide | past | favorite | 26 comments



I've studied this problem for 30 years. The short answer is not enough following distance to the car in front of you. That distance represents a time-space buffer. The time-space buffer allows for local variations in speed without collisions AND without reducing speed to zero.


Couldn’t we require manufacturers to add something that annoys the driver if they are too close to the driver ahead of them? An anti-tailgating alarm that can’t be disabled or something along those lines?

I am not big on regulation, but I can think of fewer things that could be more impactful to the average person than the elimination of traffic jams.


you want to solve this you give a monetary benefit to people who traffic break, a methodology of driving to create buffers in "time-space" that remove brake lights from areas and allow the flow to speed up.

Merely having an annoying alarm won't do it, if anything, i figure it would cause more road rage, because it's something that feels like its out of your control. My 2023 subaru has adaptive cruise control, and you can set the follow distance, and even the furthest follow distance feels too close in traffic. When i would do traffic breaking while driving rush hour in Los Angeles county, i'd maintain 3 car lengths whenever possible between me and the car in front, for example.

The best place to traffic break is in the "slow lanes", if you have enough people doing it the ingress/egress traffic can merge without causing issues and more efficiently, which eases traffic. The second place traffic breaking helps immensely is when there's interchanges to other freeways, you traffic break in the lane that "ends" for the other direction, so that people who wait until the last minute for whatever reason don't have to cut anyone off to merge which reduces the brake lights in that area.

i don't have any videos of me doing this, as "dash cams" weren't really a thing when i was doing it in L.A.; i suppose i could record some video the next time i'm in a populated area like DFW or NOLA


> Couldn’t we require manufacturers to add something that annoys the driver if they are too close to the driver ahead of them?

I'm in a state where it's a violation to not exit the passing lane if a vehicle behind is overtaking you. Lets add alarms for non and hostile compliance.


Just a proximity sensor that cuts off the music and plays a repetitive tone if you are too close while traveling above 20mph should be enough.


That may be the reason for sudden traffic jams but it isn't the reason for traffic in general. Leaving more distance reduces throughput and necessarily creates congestion further back.


> The short answer is not enough following distance to the car in front of you.

I don't doubt your expertise or qualifications.

I do however wonder if your conclusion factors in realty. Specifically where the buffer endless gets filled by one vehicle after another merging into it. A high % of the new arrivals barely pass (or don't) the vehicle they wanted to get by, converting the passing lane into not-that.


People would have to change their behavior from a selfish mindset to a more collectivist mindset. As someone else said, they will need an incentive to do this - either a carrot or a stick. The current "stick" - wasting time and fuel - is not working.


> People would have to change their behavior from a selfish mindset to a more collectivist mindset.

My private fantasy is that self-driving would introduce this. However, auto manufacturers wield selfish interests like a superpower. I doubt anything like benevolence could escape into their product lines.


I think it's more the failure to start driving soon enough and fast enough. People wait and wait and then go and then go slowly, instead of moving as soon as possible and faster. That adds up over the course of many cars.

I see it every time I get to the area of slowdown (like one lane into two lanes). People just go slooooooow there.


As explained by William Beaty (of amasci fame): https://www.youtube.com/watch?v=iGFqfTCL2fs


this is incomplete, and merely describes a problem set. Researchers showed that it only takes one set of brake lights to mess up the flow of traffic, and the resulting obstruction will move backward from the point it started at about ~20MPH (iirc, may have been 10mph).

when you're driving before traffic starts and you see someone merge and step on their brakes, making the person behind them step on their brakes, you've just witnessed the beginning of "rush hour traffic" - that will propagate backward and cause rippling slowdowns for over an hour.

Hit "save recording" on your dashcam, because that person is 100% today's proximal cause.


George Carlin said it was because of the idiots in front of you and the maniacs trying to pass you.


So was the submitter sitting in i95 holiday traffic yesterday, lamenting the delays lingering after a long-cleared accident? I know I was and I remembered this simulation.


A good explanation is that freeways behave like superheated water in that they are somewhat bistable. Just like water can be above 100C without boiling when there are no impurities, freeways can transport more traffic throughput than saturation when there are no disturbances. As soon as there are disturbances the freeway can only support a maximum throughput, unfortunately that maximum throughput occurs at low speed so the traffic jam occurs. To get back to the free flowing regime the throughput has to be lower than the maximum for speed to increase. Like with water you can't get back to the superheated regime without stopping the boiling. So yes you can improve throughput and reduce traffic by removing disturbances but it doesn't always work, better to reduce traffic volume in other ways or reduce bottlenecks by widening roads consistently.


I thought it was pretty much proven that widening roads doesn't decrease the frequency of congestion.

I think the only way to get rid of those disturbances is to force cars to move at a set speed and at set positions. For example, by pulling them on hooks along rails at fixed distances.


>reduce bottlenecks by widening roads consistently.

I too like trillion dollar infrastructure debts.


that have been repeatedly shown to not solve the problem? count me in!


Epiphenomenons are so delightful. Once they occured they are often easy to explain, e.g. by Invisible Hand explanations [1] like in TFA, but it is still hard to predict when they happen.

[1] short summary https://academic.oup.com/book/26496/chapter-abstract/1949678...


In Germany there's the concept of "Stau aus dem Nichts" based on the Nagel-Schneckenberg-Model

https://de.wikipedia.org/wiki/Stau_aus_dem_Nichts

https://de.wikipedia.org/wiki/Nagel-Schreckenberg-Modell


There is this one area of highway in my commute that always slows to around 60mph, down from 75mph. No idea why.

My best guess is that the trucks have run out of momentum at that point of the incline, though I don't notice this behavior at other incline crests.

I run on cruise control the entire way so this is very noticable and annoying.


I really like this experiment: https://youtu.be/7wm-pZp_mi0?feature=shared


It is nice, but it looks like it was filmed with a flip phone camera. Try this one: https://www.youtube.com/watch?v=Rryu85BtALM


CGP Grey: The simple solution to traffic

https://www.youtube.com/watch?v=iHzzSao6ypE

aka: traffic snakes


Yeah I was saying that above. If you watch the cars after the breaking, they take too long to start going and they don't go fast enough.


(2009)




Consider applying for YC's Spring batch! Applications are open till Feb 11.

Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: