The Physics Behind Traffic Jams | Hacker News Reader

The Physics Behind Traffic Jams

smartmotorist.com

256 pointsnetvarun102 comments

>"It was dusk, the headlights were on, and I was going down a long hill to the bridges. I had a view of miles of highway behind me. In the other lane I could see maybe five of the traffic stop-waves. But in the lane behind me, for miles, TOTALLY UNIFORM DISTRIBUTION. I hadn't realized it, but by driving at the average speed, my car had been "eating" traffic waves. Everyone ahead of me was caught in the stop/go cycle, while everyone behind me was forced to go at a nice smooth 35MPH or so. My single tiny car had erased miles and miles of stop-and-go traffic. Just one single "lubricant atom" had a profound effect on the turbulent particle flow within the "tube.""

When I try this, most other people have no idea what I'm doing and/or are stupidly greedy and go "OMG A GAP" and rush into it. I'll also typically have people behind me honking / flashing their headlights at me, as they're equally clueless. I've had one dude and his wife, in a Harley Davidson edition F150 actually roll down their window and scream at me for "not knowing how to drive".

I drive a 6spd stick too, so the stop/go is even more painful. I try to be that "lubricating atom" both for my sanity and to ease the jam, but the collective inability of other drivers to think beyond "OMG MAKE AS MUCH PROGRESS AS SOON AS I CAN" greatly inhibits my efforts.

TL;DR traffic jams are caused because people are greedy and short-sighted in their greed. It's not really different from all of the other big problems with humanity.

I experienced this as well but you know what, my driving experience is much less stressful on the whole and it's kind of amusing seeing people try to fill that gap by speeding past you only to slam on the brakes as second later.

I have a different theory on why people jump into gaps: they're bored. Traffic jams are really dull, and the illusion of progress is much more entertaining. So not really greed.

And it's more ignorance/self-centeredness than short-sightedness; they haven't thought of the jam as a systemic problem to which they're contributing.

I beg to differ, and that's a criticism I also have for the OP:

From my humble observations, traffic jams are caused because too much traffic (x cars/sec) flows into a bottleneck, i.e. something that can't allow x cars/sec to pass through. Whenever I get into a traffic jam, I play "spot the bottleneck" to ease my boredom, and so far I haven't found a case where there was none.

Transient traffic jams like the OP described seem to get solved by people accelerating as much as they can once they're past the jam. If there's no bottleneck ahead, they're indeed transient.

I agree with this observation. I cringe when I hear other comments talking about people driving through traffic trying to smooth out traffic waves. I am very suspicious about the claim that this speeds things up for anyone.

People accelerating hard and braking hard is not good for traffic, but if there is some guy moving slower than possible on a highway with a lot of commuters, many people will need to change lanes to avoid that obstacle. That is not good for traffic either.

Bottlenecks caused by an accident or merging lanes on a motorway are one thing, but I've seen traffic jams where the bottleneck was a disabled car that was well clear of the road, or the police have pulled someone over. In those instances there's no good reason to have a jam.

Well, there's a reason, it doesn't matter if it isn't good: other curious drivers. Here in Israel, traffic reporters on the radio routinely implore the listeners not to slow down and take a look, since it causes jams.

My point is, whatever causes a segment of the road to only allow x cars/sec will cause a bottleneck if there are more than x cars/sec incoming flow, regardless of whether it is inherent or not.

Of course there's a reason for the traffic jam, but people make it far far worse by tailgating.

I saw some fantastic traffic jams in China. There'd be some incident on a highway (not the toll highways, but the badly supervised free ones between nearby cities). People would drive on the wrong side of the road to get around it; on both sides. Eventually, there'd be thousands of cars backed up, on both sides of the road, unable to move more than a few feet. It was good news for the nearby shopkeepers, who'd have people leaving their cars to get something for dinner (since they couldn't move anywhere) - this did cause some issues when the traffic started moving again though.

But yes, originally a truck had broken down under an overpass or something, so arguably that was why it all started.

>When I try this, most other people have no idea what I'm doing and/or are stupidly greedy and go "OMG A GAP" and rush into it. I'll also typically have people behind me honking / flashing their headlights at me, as they're equally clueless. I've had one dude and his wife, in a Harley Davidson edition F150 actually roll down their window and scream at me for "not knowing how to drive".

I understand your plight. I've had this happen to me on a number of occasions, it sucks :(

Some years ago (2008) research (1) demonstrated that traffic jams were chaotic dynamical systems and given a certain traffic density would be inevitable based on the variability of drivers. Here is a video demonstrating as such:

http://www.youtube.com/watch?v=7wm-pZp_mi0

I always found this fascinating. My hope is that the driverless car idea will take off in earnest and normalize the variability between drivers, making for a much more rational/logical commute.

1: http://iopscience.iop.org/1367-2630/10/3/033001/fulltext/

The problem with driverless cars is that not everyone will switch to driverless cars overnight, so the driverless cars will have to deal with drivers. Ever had your GPS tell you to get in the wrong lane? I'd love to see how well a Google self-driving car gets on trying to get over 4 lanes in a mile in gridlock traffic. I live in the bay area, and you only ever see the self-driving cars out when traffic is very light. It'll probably be a "better cruise control" long before we have cars that we can get into, program a destination, and sit back while the car takes us there. It's a nice dream, but it seems 10+ years from being viable.

For the "better cruise control" aspect, I can envision the user interface giving the drive the illusion of driving in a rut in the road. So that when the car drifts to one of the lane edges, there would be a slight tug on the steering wheel similar to what you would feel if the road was cupped -- but something you can easily overcome if you intend to.

Contrary to one of the recommendations in the article, the Minnesota Department of Transportation recommends against merging early, and instead has made a large public-education effort (including signs at merge points) to encourage zipper merging:

http://www.dot.state.mn.us/zippermerge/

More discussion:

http://www.drmomentum.com/aces/archives/003800.html

Their research is fundamentally flawed...and it is rooted in an fruitless desire to optimize something that is physically unoptimizable. There is no sand-arranging possibility that can cause the falling-rate of sand in an hourglass to change significantly.

Merging later encourages better utilization of road capacity and eliminates bottlenecks further away from the overloaded resource. It's quite simple really, in this case the polite behavior is wrong, and if everyone would just be greedy, the behavior would no longer be impolite and would be more optimal (source, I live in china, this really works better than the alternative, we have plenty of overloaded resources).

Every advocate always has their pet theory as to why their method works better than others. They also have "evidence" which is either completely anecdotal or does not control for even the most obvious of factors (such as rate of queue arrival or average speed in the constrained resource).

Simply put, they are all wrong. A bottleneck is a bottleneck and there is no fairy dust magic queueing technique that works better than another. At best, some techniques might have a behavioral advantage...but I stress again, they are all inconsequential. We are talking tens of seconds of time savings at best.

Late merging makes better use of available road bandwidth. That's not anecdotal, that's simple math.

Late merging makes more effective use of the car waiting area, at the expense of a faster flow.

So you mean the flow through a straw is better than through a funnel?

Bandwidth is probably the wrong term here. It keeps traffic from backing up as far, true, but this really only matters to people who are turning off before the merge.

Ignoring early exits, at steady state, the only thing that really matters is flow _after_ the merge - that's what determines the rate at which traffic passes the restriction.

I would never try this myself mind you; I hate pissing people off in the road in the states and I'm happy to outsource my driving to a taxi driver in china.

However, it's a simple matter of utilization, if everyone waited properly in line, traffic would seize up in Beijing as many intersections would become blocked by single lines of cars next to empty unused lanes.

We are talking about saving minutes, 10s of minutes, or even days [1]. Funny how Americans haven't seen real traffic before and think they know the right global solutions.

[1] http://www.npr.org/blogs/thetwo-way/2010/08/23/129376194/tra...

Assuming this is true, zipper merge is still a vastly preferable system.

1) It's a more effective use of the roadway. Why have 1 lane of traffic backed up for 2 miles when you can have 2 lanes of traffic backed up for 1 mile? 2) It's equitable. Everyone moves at the same speed, there's no way to game the system for personal advantage.

It's not evident that this is a more effective use of roadway, as the article hints at.

The problem is the bottleneck point, where people are obligated to move into one lane. This is what causes delay - the faster you can get cars moving through the bottleneck point, the less delays there will be behind. So it doesn't matter so much where the cars behind are positioned to wait.

Zipper merge likely causes more delay - in practice - as there has to be a late negotiation over whose 'turn' it is to merge. If this negotiation is planned earlier, it doesn't need to have such an effect (in fact, I imagine it could eliminate all the delay) at the bottleneck point.

I think the point of zipper merging is that the negotiation has, in fact, been planned a couple miles before the actual merge point, so that the cars are alternating like

  | _   _   _
  |   _   _   _

Then it's easy to condense to

  | _ _ _ _ _ _

Minnesota DOT probably recommends zipper merging because they are taking into account that it's likely a safer (and easier) method for drivers in denser traffic.

That might be true, but it certainly sounds like a behavioral, not absolute, position. If it is behavioral, you have no guarantee that what works in Minnesota will work in Illinois, let alone a different country.

Well, that's anti math. Granted their are behavioral positions, but this not one of them.

How is my comment in any way related to math, let alone taking an antithetical position to it?

Zipper merging works far better in practice than the early merging proposed in the article. To begin with you usually can't see where the land ends well enough to merge early, and once things slow down you're screwed. Zipper merging works great if everyone expects everyone else to do it. The problem in the US is that people trying to zipper merge are punished by drivers who think early merging is polite.

Most drivers (at least in Australia) will be courteous in this way. Where it breaks down is that the point of merging tends to move backwards, eventually to a point where drivers are merging quite early, sometimes to the point where arriving vehicles don't even know a merge is occuring. What then happens is drivers push ahead and attempt to start a new merge close to the barrier/traffic cones again. However no one wants to merge twice and see these restarts as skipping the line.

The OPs main point of leaving larger gaps might solve this, but generally gaps will fill and you are then back to gridlock.

This accords with my own experience driving in Sydney, where the need to merge is frequent, but not Canberra, which has more more US-like freeway system.

Brisbane had a few bad years of building insane combined on/off ramps that tend to block frequently as everyone needs to be in that lane but with three or four lanes of traffic. Hale Street as you pass Suncorp Stadium is a particularly bad one.

You also get many US drivers who refuse to let anyone in and just tailgate the car in front of them, so they can be "first" (in their minds).

The article does not argue about merging late vs early. It's about leaving a larger gap to allow merging in high speed, which still can be done late.

The article shows what looks like a time-continuous microscopic traffic flow model ("microscopic" means that the model accounts for individual cars as opposed to fluid-like "traffic streams"). There also exist cellular automaton-based models like the two-dimensional Biham–Middleton–Levine model [1] or the one-dimensional rule 184 [2] that can also show fascinating behavior but are much easier to program a computer simulation of and to alter (say, introduce different kinds of vehicles to).

Check out the videos on the Wikipedia page to which I've linked, especially [3], to see what I mean. In those videos the blue dots are trying to get from the top to the bottom while the red ones are trying to go from the left edge to the right edge of an "intersection".

[1] https://en.wikipedia.org/wiki/Biham%E2%80%93Middleton%E2%80%...

[2] https://www.wolframalpha.com/input/?i=rule+184

[3] https://upload.wikimedia.org/wikipedia/commons/1/18/Biham-Mi...

Original copy at http://trafficwaves.org/

A video from there, omitted from OP: https://www.youtube.com/watch?v=iGFqfTCL2fs

OP is a cut-down version of this site, which has been up for a decade or two. I'm curious if it's a legitimate copy.

The solution to traffic jams is to have a dynamically calculated speed limit based on the density of the traffic flowing through it. A car going 60mph down the highway needs 50 feet in front and 50 feet in back between it and the other cars (for braking distance). When that same car is going 20 mph, you only need 10 feet in front and back. Therefore, to support more cars, speed needs to slow down. You could have a device that counts the amount of cars entering the highway, and then an algorithm could assign a speed limit for a section of the highway. In the middle of the nigh, theres no one else on the road, so the speed limit will be very high. During rush hour the speed limit could be as low as 15mph. This would make merging easier since you're going slower.

You wouldn't think that lowering speed limits would easy traffic congestion, but it would. Its a counter-intuitive, so it will never be implemented anywhere.

It's already implemented on the M25, the orbital motorway around London. It gets used quite frequently to slow commuter traffic down from a limit of 70mph to as low as 40.

There's a particular incentive there because being a loop a bad traffic jam could (theoretically - I'm not sure it's ever happened) stretch so far that its head and tail collide. Plus, of course, the fact that the M25 was demonically planned in the shape of the mystic sigil Odegra, from the alphabet of the Black Priesthood of Ancient Mu.

It's done in central Auckland too.

Many German motorways use that system and have dynamic speed limits based on the current traffic situation. The downside is that most people think »Why the hell should I drive 100 now when I can still drive 120?«, so depending on the mindset of the drivers this may or may not work. I rarely see those speed limits followed or enforced.

Variable speed limits are well implemented on UK motorways.

One thing I find interesting is that, as a traveller, it's completely impossible to fathom what's going on. Because the signs are on in apparently free-flowing traffic, it's really not clear whether the signs are providing any use or are just still turned on from a earlier occasion.

They put up dynamic (digital) speed limit signs in Seattle (where this article is based) in the last year or do, and I've been wondering how effective they are. After reading this article I feel like this is the reason.

There is one speed limit over each lane when it's slower than the regular posted limit, which to me seems like if you see your lane slower than the one over from it, you'll merge, causing next lane to go slower.

Reading the article, and the comment above,I can now see the purpose of the dynamic signs. But unless drivers adapt a mentality that's best for the traffic and not themselves, this doesn't seem like it would work.

I've yet to see someone follow them.

This Happens in the Netherlands too, we get cut down from 120/130 to 100, 80, 70 or even 50 kph.

It's the same as when approaching a red light or any other intersection. If you drive all the way up and make a full stop you will have to accelerate from 0 once you get green. If you slow down when you see the red you can be able to roll into the intersection with speed just as the light switches to green effectively giving you a 30km/h headstart or more.

You leave the intersection faster and you save gas by not accelerating. I don't understand why more people don't do this very simple move. The tricky part is timing the switch perfectly which is almost impossible, instead you should just slow down more and more and more and try to never ever fully stop the car fully.

I honestly try this every time and slow down before nearing the intersection. But often there are assholes who overtake and come in front and fill up the road ahead, making me brake and come to a stop as well.

> just as the light switches to green effectively giving you a 30km/h headstart or more

Please be careful with this. If you immediately occupy an intersection after a light turns green there's a chance you'll get blindsided when another car decides to run a red.

A shame "rubberneckers" were mentioned (even if in passing); it always bothers me when "rubberneckers" are blamed for a slowdown, despite that it may simply be the long-term wave slow-down after the incident (as also discussed in the article). The reason it bothers me, especially when it's mentioned in a traffic report, is it gives everyone a little easy road rage "blame the bad driver" excuse when everyone just needs to chill out and realize this will only cause a few minutes delay at worst.

It's also futile to try to "train" everyone to overcome these little wave effects. All you're going to do is give some people an excuse to rage at others about their "inefficient" driving. Just keep your distance, follow the rules of the road, and put on some good music/audiobook/podcast/etc. You will get there eventually, and one day better self-driving cars or mass transit will make this frustration moot.

(Note that even if self-driving cars aren't specifically programmed to solve the wave effect, it won't matter because you'll be too busy watching TV or reading to give a damn. I'm all for efficiency, but having commuted by car for years, it's not about these little things anyways, it's the sheer time spent that is the worst thing (and damaging on a personal level), and traffic accidents truly do block traffic anyways.)

I think it's easy to call everybody rubber-neckers, but in reality when you pass a big mess of police lights or anybody stopped on the side of the road, I consider it a bit of a courtesy and safety to slow it down a little and, if possible, merge into a lane so you don't do a "fly by" at full speed with somebody standing at the side of the road.

It's entirely too human to have spent 20 minutes in a jam to finally get to the bottleneck, then spend all of 5 seconds seeing what all the fuss was. It'd be nice if folks didn't do this, but they're not going to stop en masse.

"ON THE RIGHT >>>>> DRIVERS WITH UNUSUAL BEHAVIOR: THEY ENCOURAGE OTHERS TO MERGE AHEAD OF THEM, AND THEY TEND TO MAINTAIN LARGE SPACES AHEAD, EVEN IF TRAFFIC SLOWS TO A CRAWL. MERGING IS EASY. SEE HOW MUCH FASTER THEY GO?"

This is misleading 1)The actual throughput of cars through the constriction even in the animation is only slightly faster despite 'apparent' better speeds.

2)The animation does not account for drivers re-assuming their previous following distance. Regardless of whether your 'comfort' distance is half car-length or 5 car lengths, you will eventually resume it. Leading to a slowdown. The advice, and animation seems to assume both that we've merged (onto the highway in the first place) at a rate that can absorb those driving distances without additional slow-down and that we won't resume our previous driving distance (after merging past the constriction). In short, its moved the problem and then shown a window into the part that was improved.

*edit: added parenthetics to improve clarity

This appears to be a prettier version of William Beaty's work at http://trafficwaves.org/ which is well worth the read. I send this link out multiple times a year to friends, because it advocates (and shows some proof for) some simple driving changes that can substantially reduce traffic: Leave space ahead of you, and let folks cut in if they want. Given the chaotic pattern of traffic and drivers, these simple habits appear to reduce (but never eliminate) traffic jams.

Just to be clear, both are by the same person.

Indeed, the article is credited to him, thanks for pointing that out (missed that!). But the original posted site ("smartmotorist") didn't seem to have any credits, so unclear if it's a rip-and-post or if Mr. Beaty was part of "smartmotorist". Regardless, I think Trafficwaves.org gives more info, so wanted to share.

I love Beaty's stuff. I've tracked his amasci.com for a couple years now- although his updates are few and far between, they are often gems. He's a damn good modern day mad scientist.

I have been experimenting a bit whenever I am stuck in the traffic. The goal is not to come to a complete stop. So, in order to do that, I keep long enough gap in front of me while the traffic is crawling, and keep a low speed such that I do not get too close to the car in front (in which case, I have to come to a complete stop).

Just by doing that, the car behind me can also just follow me at the same speed without coming to a complete halt.

I just hate sitting still on the highway. This game keeps me entertained while also helping a few cars (if not all) behind me.

The best place to see this effect is in a tunnel (say, Holland tunnel, where switching lane is not allowed). Just one car maintaining a low speed with enough space in the front can make the whole lane flowing so easily.

I just don't understand people who would accelerate and the brake, and repeat this until they are out of the traffic. Not only you are causing yourself a lot of anguish, you are also not helping others, nor your car.

>I just don't understand people who would accelerate and the brake, and repeat this until they are out of the traffic. Not only you are causing yourself a lot of anguish, you are also not helping others, nor your car.

I don't either. I developed the same strategy as you describe accidently in an effort to conserve fuel. Braking is effectively converting your precious gasoline into heat/sound/wear-on-parts. It's best to never have to break. I realize in practice that's not achievable, but there are many cases where a little foresight can mean just not accelerating and expecting stops earlier, and can give you a ~3-5mpg increase.

>>I just don't understand people who would accelerate and the brake, and repeat this until they are out of the traffic. Not only you are causing yourself a lot of anguish, you are also not helping others, nor your car.

Accelerating and braking is also a fantastic way to waste gas.

Most people drive using the least amount of attention as possible to the road. The typical person behind the wheel is thinking more about their life's problems than they are thinking about the flow of traffic around them. When I'm driving to a place I've never been before. My attention is mainly on my GPS screen, not the flow of traffic ahead of me. When I'm on my daily commute, I notice the traffic patterns because that drive is boring and there is nothing else to think about when driving.

Yeah, Traffic Waves. There's no cure. They've been studied, documented, they have many causes but there's no way to get rid of them unless all the cars on the highway are autonomous and self correcting.

Once a highway increases past 2 lanes they are inevitable as people from the far left lanes have to merge right to get to their exit and the 2 right lanes than can normally hold 20 cars with a 1 car space in between all of them now have to cope with 35 cars. The cars in the left lane can't merge in time to get to their exit so the far left lanes start slowing down as well.

There are a LOT of causes, from bunching (cars driving too close together) to pattern breaking (cars slowing down to look a pulled over cop), to merging, to accidents, to bottlenecks, to over-crowded exits, etc...

The only way to provide some relief is for ALL drivers to travel in groups/packs with large gaps in between the groups/packs. The gaps can obsorb slow downs.

>The only way to provide some relief is for ALL drivers to travel in groups/packs with large gaps in between the groups/packs. The gaps can obsorb slow downs.

Whenever I'm stuck in Philadelphia traffic I try to do this and it seems to make a difference. I've even thought about getting a bunch of friends to all get on at different exits and do it simultaneously. The end goal being official "pace cars" coordinated by a central command center to alleviate stop & go traffic and keep everyone cruising at a constant speed. Needless to say, my mind wanders a lot when I'm stuck in traffic!

I've actually done something similar. When in bad traffic i purposely drive so that i leave a large gap between me and the car ahead of me. I basically drive so that i have to do minimal accelerating/braking. Likely just a coincidnce, but it does seem to improve flow.

Are you located in/around Philly? If so, I'd be interested in coordinating something on 476 or 422 one afternoon.

Nope! Bay area (which btw, does not have bad traffic relatively speaking; I've been in traffic jams in the NE and they are way worse, same with LA).

I do this when stuck in a jam with a manual transmission car. I'll just leave it in first and idle along at ~5 mph... this is slower than the momentary maximum ~15 mph "rubber band" speed but certainly feels like it smooths things out in the long term, and is a lot easier on my clutch foot.

I've done this, too. Frankly I just feel better going a steady rate - even if it's 0.5 mph - rather than continuously stopping and starting. The cars behind me tend to disagree, and sometimes even honk ... even though the space ahead of me clearly only leads to a continued jam, and not open road.

>"but there's no way to get rid of them unless all the cars on the highway are autonomous and self correcting." Sure there is; teach everyone how to do race starts. I always get on it pretty hard coming away from a jam and hope that the person behind me does the same. It rarely ever happens, but I've had some people follow my lead. If people accelerated faster off the front, it'd move the wave back faster. How this might collide with a second wave is not something I've really considered.

> The only way to provide some relief is for ALL drivers to travel in groups/packs with large gaps in between the groups/packs. The gaps can obsorb slow downs.

You see this being done with traffic lights where the green lights are timed to move each group through subsequent sections. Cars that get too far ahead or behind the group get reformed into other groups when they come to a red.

This is awesome. The problem is that no individual agent is incentivized to drive slower. I wonder if it'd be worth it to hire "stir-the-pot cars" whose job it is to drive slowly during rush hour and ease jams. Kind of like dietary fiber for the road.

Hey, doesn't Uber have a whole fleet of cars in cities around the world? It would be a cool $&@! project for them to be the all-wondrous curers of traffic.

One could argue that there is an incentive. If you drive a little slower and leave a large gap in front, you get to avoid always being on watch for the guy in front of you slamming on their brakes. This can be a whole lot less stressful, which is incentive itself.

I think the problem is less one of incentive than it is one of awareness. If more people knew about — and actually understood — anti-traffic, more people would employ it. I don't know where the inflection point is, but I'd bet you don't even need all that many people engaging in anti-traffic behavior before it has an appreciable impact, either.

The Colorado Department of Transportation does essentially this with "pace cars" at high traffic times coming back from the ski areas.

If you're interested in the physics of it, there's a PDE that explains these backwards propagating density shockwaves (aka traffic jams).

http://en.wikipedia.org/wiki/Burgers'_equation

If you enjoy this sort of stuff, the book Traffic: Why We Drive the Way We Do (and What It Says About Us) by Tom Vanderbilt[1] is an enjoyable read. He covers traffic waves and a lot more.

[1] http://www.amazon.com/Traffic-Drive-What-Says-About/dp/03072...

> My single tiny car had erased miles and miles of stop-and-go traffic. Just one single "lubricant atom" had a profound effect on the turbulent particle flow within the "tube".

I did this in the 1970's. It gets rediscovered constantly.

The problem with leaving a gap is that, as soon as it's a few car-lengths long, people in the adjacent lane(s) notice that there's a gap and quickly cut into your lane to fill it, because it moves them a car-length or two forward. In order to do this, they have to cut in front of you, speed up quickly, and brake quickly. And you're left in the same position that you were in before you made the gap -- plus one extra car.

I try to do this all the time and it's frustrating because it's completely impossible. As soon as a vacuum exists, drivers rush to fill it.

So its interesting to watch the traffic around when this happens. It actually smooths out the flow "behind" you because the folks who cut in front eased up pressure behind them. The tail gater issue is real though so if you're going to do this it is best to do it in the right lane rather than one of the inner lanes.

My experience has been that the "slow lane" is usually significantly faster through a jam, probably because of the number of lorries practicing smooth driving.

Can't do this for traffic waves, but it would appear the ideal solution to merge slowdowns is to have different speed limits before and after the merge. Regardless of anything else, the flow rate before and after a merge must be the same. Ideally you want the flow rate after the merge to be approximately one car every 2.5 seconds, with those cars traveling at the speed limit. (As opposed to the usual situation where post-merge traffic accelerates from a near stop, causing large, unnecessary spaces to form.) One way to achieve this is to have pre-merge traffic leave twice the space, but it's unlikely there are enough traffic-flow vigilantes out there to achieve that, if such a thing would even be desirable. Much easier is to keep the same time spacing between cars, but to have pre-merge traffic travel at half the speed. Traffic approaches the merge in two lanes, traveling at 50 km/h. The speed limit increases to 100 km/h, then immediately afterward the traffic merges. Speeding up from 50 to 100 naturally opens spaces, and traffic merges painlessly. There is no reason for a slowdown to occur, so no self-perpetuating gridlock.

Of course, the slower speed limit is only useful when the road is filled to capacity; otherwise it would be inefficient. So for most roads you would want this limit to vary based on time of day (or even dynamically based on traffic volume, but obviously that would only be useful/necessary in extreme cases).

This is the prisoner's dilemma. If you slow down and drive smoothly, and everyone else does too, then everyone benefits greatly. But if you cut in front of one of the smooth drivers, which you can since he's leaving lots of space, you benefit personally by gaining a car length while inconveniencing everyone else in that particular lane. Well, at least until the guy behind you has smoothed out the difference. Human nature makes smoothing out traffic an improbable solution.

This is really interesting. I wonder if the existing California Highway Patrol practice of swerving around the highway[1] to slow down traffic can be used to create these "anti-traffic" gaps during rush hour?

[1] http://articles.latimes.com/1996-08-12/local/me-33464_1_traf...

Near where I live in England we have managed motorways which employ electronic displays and enforce the speed limit with speed cameras.

http://en.m.wikipedia.org/wiki/Managed_motorways_in_the_Unit...

Even better solution to traffic: develop real high-speed rail in the US, live closer to where you work / need to be (stop suburbs / sprawl), ride bikes or even just motorcycles / mopeds in California where filtering is legal, or walk.

This has a lot of other nice first-order benefits like real communities, decreased pollution, less time wasted commuting, and tons of second-order benefits.

Someone at Waze should get on this. Send a small notification to Waze users that there's a jam ahead, and to slow down by 10mph to help alleviate it! It's a simple solution and might even work since Waze users are already a "community" of drivers in the sense that they're looking out for each other and not necessarily just themselves.

I usually try to avoid hitting the brake as much as possible in order to maintain the most uniform speed I can. This is due to a selfish, lazy motivation, because I don't want to rapidly alternate between gas and brake.

I think there is also a psychological component to seeing bright red brake lights. I want to say it creates a reactive, and often overly reactive response. I imagine lifting a foot off the gas pedal often creates sufficient deceleration without the alarm caused by hitting the brake pedal but many people don't drive this way. Maybe we should have two shades, or colors, or a full gradient to indicate level of braking? What if a brake light could indicate current speed?

Interesting blog article.

I've always found it strange that when school is out, traffic is like 30-50% better. This effect is the best I can think of, because I wouldn't think there's a lot of people directly impacted by the school schedule. Elementary, middle and high schools all start and end at different times, and there's people directly employed by all of those schools, and certainly some parents who pick up and drop their kids off, but still it doesn't seem like that many people compared to how much better the traffic gets.

So maybe this article is right, you take a few people off the road and the merging madness, and everything becomes better.

I notice the same thing, Even in one week holidays where only the elementary schools are closed the traffic is much lighter. I expect this because a lot of working parents take this week off too to be with the kids, probably a 10-15% drop of traffic means that traffic is a lot easier.

I remember some sort of experiment in Colorado in which state police formed a line of cars at intervals forcing everyone to drive the same speed, and thus was able to eliminate the daily blockages in that experiment.

A related Japanese experiment on the Human factor of traffic jam - https://www.youtube.com/watch?v=Suugn-p5C1M

The merging solution is oversimplified. Leaving enough space for another car in front of you sounds easy. But in reality, if 1,000 cars each add 2 car lengths in front of them you've added 40,000 feet to the length of the slow down. Yep, 7.5 miles.

If it was feasible, we could also stop at red lights with 3 or 4 car lengths between each car and when the light turned green everyone could go at the same time. There isn't space on the road for this type of behavior.

I found the idea that one person can make a huge difference really interesting, I will have to start experimenting with this myself. All the roads OP wrote about are where I drive, but the Lynnwood ramp he talked about recently got expanded.

This also makes me want to make some sort of simple traffic simulator to do experiments like this, does anyone know if any already exist?

My heuristic when trying to help get rid of a traffic jam is this:

> Maintain a higher minimum speed (and the same or higher average speed) than the car in front of you.

The only reason why traffic jams ever disappear is that people are doing this. Most of the time I guess the true root cause is that incoming traffic is dissipating, though.

I am reminded of this:

http://www.technologyreview.com/view/425435/in-car-algorithm...

On the freeway system in LA, a lot of the jams are caused by slow drivers in the passing lanes. Most the time, when there's a clump of cars and empty space ahead of them, it's one or two slow drivers blocking the passing lanes.

There's a nice simulation of these physics here: http://www.traffic-simulation.de

The jams seem pretty robust, unless you turn down the input flow -- a bit like real life!

There's a whole chapter on traffic jams in the book Critical Mass. From what I remember, innocuous things like a single driver slowing down can cause massive jams due to the natural 'physics' of traffic.

reminds me of an experiment by some college kids in atlanta - http://www.youtube.com/watch?v=OoETMCosULQ

Jams spontaneously occur after a certain road density is reached.

The sadistic side of me wishes they'd just close entrances to the highway whenever it's dangerously full.

I saw William Beaty's zen traffic warrior video years ago, and I don't know if it works, but it sure makes me more relaxed when I'm driving.

Does this mean it's actually possible to improve traffic by introducing extra tempo setter cars whose only job is to smooth traffic?

interesting!

This reminds me of an interesting traffic simulation applet, where you can control things like aggressiveness of the drivers, traffic density etc. and observe the traffic jam that happens.

http://www.carbibles.com/trafficmicrosimulation.html

Reply on news.ycombinator.com