Slashdot Mirror
Traffic Sim Predicts Jams Before They Happen
Via_Patrino writes "The New Scientist reports that: A traffic simulation system is helping drivers by predicting jams up to an hour before they happen. Traffic flow can be divided into three categories: freely flowing, jammed, and an intermediate state called synchronised flow in which dense traffic moves in unison. Physicists at University of Duisburg-Essen have developed 'the first model to reproduce all known traffic states.' Predicted conditions are displayed on the official website, and more than 90 per cent of the time, traffic density is predicted correctly."
What if people start using it? They avoid the traffic jams, thus no traffic jams. Wrong prediction!
I remember a few years ago being told by a teacher that traffic flowing under 28mph on Motorways obeys hooks laws for compression waves travelling through a spring almost perfectly...
The compression waves travelling through the traffic are the reason that everything goes stop/start once traffic slows below a certain speed...
I hate seeing people speeding to red lights, when if they'd just follow the flow of traffic and the lights they'd get where they're going just as fast, and without causing traffic jams. I'm a terrible driver, but I figured this simple thing out pretty early on. Why can't the rest of drivers?
Hi! I make Firefox Plug-ins. Check 'em out @ https://addons.mozilla.org/en-US/firefox/addon/youtube-mp3-podcaster/
An idea borrowed from quantum mechanics...
Measuring the system changes the system.
What if everyone takes the same alternate route to avoid the "busy" route?
As the story points out:
"But the website has already become a victim of its own success, admits Schreckenberg. Some of the 300,000 people a day who are visiting the site are replanning their journeys on the basis of its forecasts, and this is beginning to make the forecasts themselves less accurate."
There were and are several projects in Germany that develope traffic models in conjunction with models simulating the route choices of people depending on their lifestyles. The problem is that there are people that do not believe that people can be simulated by cellular automata and reject those simulation. But there are also city planers who use those models when planning new structures like stadiums, airports and others that need road connections.
There is a sourceforge project thatsimulates multimodal traffic with cellular automata. Also have a look at this link for more information about traffic simulation.
***Quis custodiet ipsos custodes***
Yeah, it's pretty common practice nowadays to use computer models to assist in urban planning, not just for designing new roads, but also to see where best to widen lanes, where to add additional lanes, and even how to time traffic lights. One problem with prediction, over the long term, is that it turns out no matter how much road you build, it isn't enough. So building in 5% additional capacity staves off traffic problems in the short term, but within a decade or two you've got 10% more drivers, the additional traffic coming from the fact that your original (planned) increase in capacity made it that much easier to drive. Of course, this is why governments are more and more turning to congestion tolls as a way to thwart this cycle. More here.
Some segments of highway are only two very narrow lanes as the highways haven't been re-done since the 1930s. In fact they are huge slabs of concreted put end to end with sometimes gaps in them. Exit turns aren't inclined cause these were made for cars that didn't exceed 30 Mph back in those days. These are, funny enough, the ones where you can speed without limit :)
On top of that, in southern germany, drivers never leave more that 2 inches in front of them as if they won a price if they can push you. I have never seen more unsafe driving practices. The whole German driving idea is to overtake you as soon as possible and then driving at the same or even lower speed than before.
So when cars collide, you get 40 or 50 smashed up together. And a traffic jam for miles and miles. It always happen every year, check the news evey winter.
Now, I don't think these correspond to most higways in the world, most other European countries have things like wider lanes, continuous asphalt, and drivers who have heard of the word "safety cushion" and respect the concept. Also for a highway system to work you need cars and cops. In Germany, all drivers think they are the cops and that the laws apply to the others.
So before you adopt this algorithm (90%!! That's VERY VERY Goog, what are you all complaining about?) make sure they model other than German traffics on it.
Also: US traffic is particular. If a cop car is there doing the speed limit, you get something like 80 cars behind in a herd and in most states people don't use the fast/slow lane systems to overtake or let faster cars have their way. It's a less dynamic system in a way and thus easier to model.
"in which dense traffic moves in unison, like marchers moving in step"
Some important facts, which the system designers seem to have forgotten: Real systems veer towards chaos and orderly behavior is anathema to all nationals except Germans. That apart, any traffic simulation design could be considered a success if it can be put to use in all conditions. I am quite sure that this system will come a cropper on Indian roads where cars have to jostle for space with cattle, cycles, pedestrians, three wheelers, two wheelers and so on. (Digression: I always find the sight of a stray dog crossing the busy road a lesson in traffic manners. The dog looks either side, ensures that there are no vehicles in vicinity and then crosses: A case of animal adapting his behavior to real time situations.) Human beings are not so adaptable and would do the best they could to defeat even the most beautifully designed systems.
If there is just a small change in velocity of one driver, the next guy is going to respond to it by hitting the breaks. The next guy is going to panic and hit it harder, and so it goes. I've seen this happen in real life many times: Just a small riple can make a jam, three or four cars involved is sufficient.
I fact, I think I saw an article about this too, it could have been long ago in Europhysics News or something. They like publishing stuff like that.
I'm trying to keep a lot of distance when I'm driving: Three seconds in normal situations (just count), and up to five seconds if I'm in synchronized flow. That way, I can absorb many ripples if the three or fours cars in front of me is slowing down. I think it helps, but surely it doesn't help a lot if it happens further ahead.
It is of course important not to lag too much when the cars starts moving again, so I generally speed up to follow in the start, but then try to build up some good distance, when the flow gets going.
But then, I'm just speculating...
Employee of Inrupt, Project Release Manager and Community Manager for Solid
The ring road around Eindhoven has recommended speed indicators that show what speed you should be travelling at to hit the next light on green. It seems to work quite well.
The Hanshin Expressway network around Osaka has video processing equipment that can automatically recognise traffic congestion, including the characteristics of traffic accidents. It then alters roadside information boards to route drivers around the congestion. Other areas do the same thing.
adio bReacons update vehicle navigation systems in real time to show time to destination, congested roads, and if you're following a planned route, will re-route you as traffic conditions change.
Unfortunately most signals in Japan aren't load-sensitive, but set to 1 minute green in each direction irrespective of time of day, day of the week or class of road. So circular routes around Tokyo, for example, become major barriers to traffic entering in the morning and leaving in the evening.
On Japanese freeways, the major cause of congestion is the humble tollbooth.
Good call.
I'm sure only a small pecentage of people would be checking the website. Heck, in my town they publish where the sobriety checkpoints are going to be set up for the weekends and holidays, and the cops STILL have a bumper crop of DUI/OMVI arrests.
Oh well, I guess it pays to pay attention...and drink at home.
--
The Marines: The few, the proud, the not very bright. - Slashdot tagline 04/21/05
Just some ideas.
1. Reduce the number of redlights needed and make the ones left more intelligent
2. Use more public transportation, especially for commuting
3. Get the idiots off the road
4. All exits should be to the right, because exits to the left mix fast and slow traffic
5. Restrict 18-wheelers to the right lane only and make them use by-passes when available
6. Get the idiots off the road.
7. Move wrecked vehicles off the highway as soon as possible, even if it means coming back to tow them later. Why shut down a whole interstate because of a vehicle in the median waiting for a tow truck?
8. Make not using turn indicators a MAJOR fine, at least $500.
9. Extend onramp and acceleration lanes
10. Get the freaking idiots off the road!
Idiots include - people talking on the phone, reading the paper, putting on makeup, sight-seeing, watching DVDs or TV, . . . you get the idea.
Why has nobody made a "Rubbernecker screen"? After an accident or incident, clear the roadway and put a screen up. Nobody can see the accident, and therefore they start moving again once the road is clear.
You could even pay for it by selling ad space!! (I can't beleive I suggested that!!)
Well, yeah, call me captain obvious. I saw this show on the history channel a little while ago that did a computer simulation of traffic on a highway. It actually showed that up to a certain point, a lot of cars on the road is fine. However, once it hits a certain point, the flow dramatically drops.
Living in New Jersey, traffic is a part of my daily life. I've noticed that even when there are a lot of cars on the road, the traffic can still move along above the speed limit in rush hour. But there is still that point when all the traffic just seems to stop and you're crawling along at 10 miles an hour.
The show also mentioned a study the government funded. They spend several million dollars to try and figure this problem out. Their conclusions: "there are too many cars on the road." Now, for the low low price of $20,000, I could have told them that.
Just like the old "gridlock alert" days in NYC. They stopped doing them back in the 80's, because of that very effect. They announced a gridlock alert day for one of the games in the '86 world series, and my uncle said the traffic was so nonexistant that he didn't do under 55 until he got to Shea stadium.
The previous has been a secret message to my comrades.
In 1998 William Beasley posted a paper on Traffic Waves in Seattle, subtitled "SOMETIMES ONE DRIVER CAN VASTLY IMPROVE TRAFFIC".
The site has great animations and excellent explanations of the impact of different drivers actions on the overall flow.
Worth a look: Traffic Waves
Right.
Note that this system is infinitely better than the radio "traffic reports," like those in Boston on 1030 WBZ. The announcers already know from experience where all the slowdowns are likely to be, so they just repeat the same B.S. every morning, true or false, until they get lucky and their helicopter spots something, or the State Police radio in an accident.
I used to commute the Mass Pike eastbound from 495 into 128/95, and I can count on the fingers of one hand the times that the "traffic report" was actually accurate.
As far as people paying attention to the web site and changing their route based on 30 minute or 60 minute advanced prediction, thus screwing up the "prediction," that's only an issue if you need advance warning before you leave the house. What's really needed is a real-time decision when you come up against a junction point. Which way? Tell me now. Should be easy enough to do with a real-time feed from your car to the model.
Not that there's that many obvious alternate route choices around Boston anyway, but it sure would be helpful to have precise information on the jam-ups for those of us who know the back roads.
SimCity did this ages ago!
The view can be seen here in the most recent version.
Beware: In C++, your friends can see your privates!
On the M25, the road that circles London, grid lock is avoided with variable speed limits and discouraging drivers from excessive changing of lanes. The road is constantly monitored by CCT and the speed limit set accordingly. The speed limit is enforced by speed cameras. In the rush hour the speed limit is reduced to 50 mph as apposed to the usual speed limit of 70mph and this, some how, reduces the risk of grid lock. Also if there's congestion at say junction 15 the speed limit will be reduced prier to the congestion, say at J13-J14. This help to reduce more trafic build up at J15 and gives it time to clear.
In a former job many years ago, I did a lot of highway driving (deliveries and field service) and spent a good deal of that time thinking about traffic patterns, both as a mathematical system as well as a design process. My final conclusion after 2 years was that there are too many cases of poor road/intersection design and WAY too many cases of pathetic traffic light design, and even more cases where traffic lights hinder traffic flow in a very severe way, while not providing any of the originally-intended safety of said traffic light. The prevailing "wisdom" seems to be: if there's a traffic problem, put up a stoplight to control it. Not.
Traffic dynamics is fascinating, and certainly deserving of some studious focus, but none of this means a single, blessed thing to me unless people will:
I'd make the common observation that older folks tend to drive slower and tend to do so, totally oblivious, in the left lane. But then there is at least an equivalent problem in the soccer moms and high-strung business suits on cell phones driving SUVs and mega-SUVs who pay even less attention to what they're doing. Heck, I've seen people driving down the interstates here (I-40, I-95, I-85 -- pick one) during morning traffic, travelling at over 80MPH and reading the morning paper!
That said, I think traffic problems tend to be less a mathematical/system problem than a brainless person problem, in many cases. Not all, but very many. Sadly, you can't "in-idiot" a driver, or a person for that matter.
I've always wished for a traffic law that gave every driver a dart gun. When someone does something obscenely stupid or hazardous (e.g. driving in reverse on a 65MPH intersate because they missed their exit), you shoot them with the dart. Three darts means your vehicle is incapacitated for 30 days. (Hmmm... I think RFID tags would be perfect for this!! :) If your vehicle is tagged more than 6 times in 2 years, you lose your license for 1 year.
Harsh? Definitely. But consider the *costs* of traffic in lost time, productivity loss, frustration/rage, increased fuel consumption, vehicle wear and tear, air pollution, etc. Pulling one person or one thousand people out of the traffic system to improve the flow for the masses sounds like good planning to me.
Oh, and please direct any comments about my tendency to drive well above the speed limit to /dev/get_out_of_my_way...
See also this site.
Have you ever been driving on an interstate highway when traffic suddenly slows to a crawl? You inch along for many minutes while waiting to see the accident which must have caused the jam. At the same time you also curse the "rubberneckers" who are causing the whole problem. But then all the cars ahead of you take off at high speed. The jam is over, but no accident, no police cars, nothing. WHAT THE HECK WAS THAT! A traffic jam with no cause? In the rear-view mirror you see all the poor saps behind you still stuck in the jam. But why? If all those people could just speed up at the same time, the whole traffic jam would evaporate. Why don't they ever do that? What caused the mysterious slowdown in the first place?
City planners don't need access to this sort of technology in places with traffic jams. I am one... in a city with really bad traffic. Like a few astute individuals have pointed out below - the traffic jams happen at the same times everyday, on the same routes! I saw a nationally-recognized traffic engineer speak once - he was a real renaissance engineer, for his age. In his presentation he showed us data that suggested that regardless of the capacity of a road, the dynamics of vehicles and drivers in the US results in a average maximum throughput of about 35 mles per hour of any typical arterial or highway. Basically, we get to choose how many cars we want stuck in traffic by choosing the number of lanes (and the capacity at the ending points) in our arterials and freeways. Adding lanes is only a very temporarily successful measure. Another interesting point he made (I've GOT to remember this guy's name... senior moment) is that at 30 mph or less, it is much more effficient to distribute traffic over a grid system, even at lower speeds, than it is to try to puch them all through a big pipe (freeway). Sounds familiar, eh?