Stories
Slash Boxes
Comments
Search

News for nerds, stuff that matters

Slashdot Log In

Log In

Create Account  |  Retrieve Password

Supercomputers Used to Study Urban Traffic

Posted by Roblimo on Thu Aug 05, 1999 12:30 PM
from the welcome-to-the-world-of-Road-Physics dept.
Science
itachi writes "This is a great article in the [Washington] Post about using supercomputers at Los Alamos and physics modeling to study traffic jams. The basic notion is that light traffic is a fluid state, with cars instead of particles, and traffic jams are sort of equivalent to a change of state to a solid. There is even talk of trying to simulate traffic along the east coast from DC to Boston, using a computer along the lines of Blue Mountain. "
+ -
story
This discussion has been archived. No new comments can be posted.
The Fine Print: The following comments are owned by whoever posted them. We are not responsible for them in any way.

Supercomputers Used to Study Urban Traffic 50 Comments More /

 Full
 Abbreviated
 Hidden
More
Loading... please wait.
  • In addition to the article on Traffic Waves [amasci.com] that someone posted up here a few moments ago, here's another one from the same author on another site, discussing practical applications: Curing lane-marge traffic jams [eskimo.com].

    To give you an idea of the scale of the modelling problem itself, there are commercial companies selling software in the $500-1500 range (and up, no doubt) for analyzing these problems. Here's an example: http://www.trafficware.com [trafficware.com]. In addition to demos (sadly, only for Windoze) it also contains many links and information on the mathematics behind traffic modelling.

    All I can say is that I modified my driving habits after reading these sites, not out of any altruistic desire to improve traffic flow, but because it was fascinating to experiment with the theory that even a single car in a large traffic jam can act as an "antiparticle" and singlehandedly improve flow in two or three lanes. The improvement in traffic flow behind me (and my reduced blood pressure as a driver) was just a happy side effect.


  • I agree that the bad driver factor is an important consideration. Take I-280 between San Francisco and San Jose. Running primarly though the 'green belt', it is usually lightly travelled.

    The average speed in the left lane is usually 75-80 MPH, that is except for the people who just feel like driving 60-65 MPH. Nobody really wants to slow down for these people, so they swerve right a couple lanes to get around the back up.

    To make matters worse, there seems to be a large portion of the population that can't turn on their turn signal without hitting their brakes, leading tailgating drivers to slam on their brakes harder. People start cutting other people off and acting generally stupid just to try to get past the one or two retards blocking traffic.

    Pretty soon, traffic speed is down to about 45 MPH - same amount of traffic, just one or two blockages thrown in.

    (Note, most Americans might think it wierd to bitch that you can't drive 75 MPH all the time, but that's California.)
    --
  • People used to know how to do this; now, they enter the freeway ~20-30 mph below the flow of traffic, causing a chain reaction slowdown.

    Another big problem, especially here in Portland, Oregon, has to do both with driver skill/courtesy and street/highway design. There are a lot of places where a right lane begins for no apparent reason, and then ends abruptly, usually on the far side of a signaled intersection. A certain percentage of rather discourteous people will attempt to drive in this lane as far as they physically can, and then swerve into traffic at the last second, causing everyone else to hit their brakes.

    Most cases of this could be eliminated by only terminating right lanes into right turn only lanes, preferably with a cement barrier to prevent people from illegally going straight through the intersection.

    --

  • (Note, most Americans might think it wierd to bitch that you can't drive 75 MPH all the time, but that's California.) Not really, from what I drove in chicago people go 85 or so until they hit traffic and then sit still for 30 minutes and repeat. It sucks, and there is no reason for it. I hope these guys get around to doing chi town soon, we sure as hell could use it.

  • I occassionally drive north on US101 to San Francisco during afternoon rush hour. Traffic is usually very heavy between the San Mateo bridge and I380 a few miles north. At that point, enough people get off the 101 that traffic opens up until you to San Francisco.

    The traffic bufferers are fine in the heavy traffic. Less stop and go, and as has been noted, they 'plug' up the lanes so people stay put.

    However, when traffic opens up and people should be driving the speed limit, the "bufferer" guy is always the one who sits in the left lane and forgets to get up to 65. By the time his buffering algorythm has informed his brain that he can speed up, several people have already made dangerous lane changes to get around him.

    So, while a theoretical approach might make sense, without practical observations on traffic condition in particular locations, it just makes you look like a bad driver and gets others frustrated enough to do stupid things which could endanger you.
    --
  • Sound argument, except for the car costs. Bus pass is X numbers of dollars.

    Cars cost more then Gas. They cost insurance and maintanence. Which does add up to a LOT more then $50 a month in the long run, especially if you do something dumb like lease a car.

    Sadly, the dollar value isn't enough to get ppl off the roads and on the busses. Because there's more to life then just commuting.
  • Anyone know of any good books on this subject?

    I've culled all the URLs I can from the posts above, but I'm looking for some good texts, too.
    Please, suggest a few!
  • It's SO cathartic!
    Yeah! Fsck you you slowpoke! SLOWER TRAFFIC KEEP RIGHT! The PASSING lane is for PASSING you moron!

    Then there's tolls.
    It tends to be in states (like Florida) where there's no state income tax - DAMN, $1.50, for one lousy toll? (of course this road is like a ghost town, because nobody can afford to drive on it - it's the hwy just east of Orlando).
    Then there are states like California, where there is an obnoxious state income tax, and NO toll roads (yea! I'm for that - on some of these roads the booth slows you down so much that you don't really save any time over the sidestreets).
    Then there are states like Illinois, which just SUCK, because the state income tax is high, and they still bend you over at toll booths.

    "The number of suckers born each minute doubles every 18 months."
    -jafac's law
  • Transport modelling has been boing on for many years. I've been doing it professionally for the last ten.

    This sort of micro modelling was starting to come about over five years ago in Scotland.

    dave
  • What they *should* be trying to do is use the traffic as a computer. Manipulate the lights in such a way as to have the traffic break RSA...
  • Good point. The least crowded way out of Chicago is the Skyway--a toll road. The free roads are always clogged.

    Part of the problem is that if you reduce the congestion, by building a new road or whatever, you make it less painful to drive--therefore more people will, bringing you back to the same problem.
  • Atlanta's got a worse problem.
    Everybody's LOST, because every frickin street is named Peachtree.

    "The number of suckers born each minute doubles every 18 months."
    -jafac's law
  • I'll post tomorrow re: the 15 mph 'limit'. Somebody I work with has FASTLANE. I was wondering that myself. I have seen neither transponder up close, but from outside a car they appear to look fairly similar. A friend of mine who has been to Europe tells me that they have similar systems there but they are much smaller and drivers don't have to slow down at all.
  • "Finally, there are many modeling software systems used by the transportation planning community.
    Here are the names of some: TranPlan, TransCAD, EMME/2, TModel/2, TranNetSim, and QRS/2.
    Their prices range from (US) $70 - $25,000 per computer license."

    Obviously, these aren't doing any good.

    "The number of suckers born each minute doubles every 18 months."
    -jafac's law
  • Sounds like you should get what the motorists got: a $140/month raise.
  • Sounds like a study done by a company that sells traffic lights.

    "The number of suckers born each minute doubles every 18 months."
    -jafac's law
  • I used to have a car and it did this averaging automatically, simply because it couldn't accelerate like most other cars.

    I remember it being more relaxing to drive this way, but this works in the "fluid" state, not in the "jello" state, 'cause that's when people jumping in front of you causes a problem.

  • Well, It isn't an exact science. If you have to brake, do so, and try again, going a mile or two per hour slower. No calculus is required :)

    As to your second point, check out the web sites that other people have posted. One of them has the observation that even though some people will cut into your buffer, they cars that don't cut in front of you form a "plug" of patient drivers who don't change lanes, and that "plug" holds the lane-changers behind you.

    - John
  • Sounds to me like an extremly detailed version of SimCity 3000. But on a 6200 processor machine. (And probably no 3d accelerator)
  • This is exactly what I have experienced. I used to drive to school, and there was one place where the traffic slowed down *just* before I had to exit. By doing the buffer thing (leaving space before me), and helping the merging traffic, I at least always thought I got through faster.
  • I don't think that the "three-phase" model is a good model. Stop & go traffic, at least the freeway variety, appears to be more a combination of people not leaving intervals between cars, and people waiting too long to reduce speed.
    There are at least three other factors at work in most urban traffic jams:
    • People waiting too long to move when traffic in front moves (with slow-moving trucks failing to keep right greatly exacerbating the problem);
    • People not merging before lanes close, forcing traffic to stop to let them in;
    • Opaque vehicles, such as semis, panel trucks and SUV's, not allowing any visibility past them and increasing the necessary following distance (this reduces the carrying capacity of the road). And every time one of them moves between two cars, the usable pavement decreases again.
    Instead, the typical action seems to be to coast as long as possible, then brake to avoid collision.
    I have never understood the people who drive that way. I watch the traffic several vehicles ahead, and I can often manage just fine by "driving" the cruise control while the driver right in front of me has to hit the brakes. It's not hard to guess how I have 82,000 miles on the original brake pads.
    If a dozen or so people do this in a row, it will naturally cause the traffic to stop and go in a ripple pattern.
    It only takes one. A single driver who reacts to a near-miss by stopping to recover from the scare will do it.

    We could get rid of a lot of these problems by having tighter qualifications for drivers, and eliminating things like cell-phone use while in motion (vehicle-integrated cellphones could easily be made to refuse to dial or answer if the driver was alone and the car was moving). The problem is, we lack the will.

  • Was told of this perplexing oddity:

    A traffic light can speed up traffic throughput (!).

    Reasoning:

    Drivers don't monitor the distance between themselves and the car infront. They monitor speed. Front car slows down, they slow down, etc. This leads to: if one car brakes, the next will brake, etc. etc., leading to braking "waves". This most often happens near narrowing entrances (eg bridges, tunnels).

    An arbitary stoplight placed a distance back from this narrowing can be used to stop the multiple brake waves into one larger brake-and-go (more fuel efficient, too, I guess) that gets stopped at other already present traffic lights before.

    Oddly enough, throughput increased from 8000 cars/hr to 12,000 cars/hr.

    This was a few years back when i heard it, so I can be completely off, too
  • Once upon a time, I had a job working for a small company located maybe 10 miles from my home in Silicon Valley. I didn't have a car, so most of the time I used some combination of walking, busses, and possibly the local light rail system. This took me an hour and a half. Each way. Every day. If it wasn't too hot, I could bike, which only took an hour. Or, if I was lucky, I could borrow a car -- and make the drive to work in only 15 minutes.

    It's easy to say, "people would ride public transit if it worked". But the simple fact is that "modern" urban developments, like what is now San Jose, are incredibly hostile towards working transit systems. Consider:

    • Residents don't want busses going down residential streets, but major streets are far enough apart that it can be at least a 15-20 minute walk to the nearest bus stop.
    • San Jose can be reasonably neatly divided into two pieces along route 280. South of the freeway, the city is primarily residential, but all of the jobs are north of it. This means that, in a north-south transit system, everybody will get on at the south end and get off at the north end, meaning you need higher capacity than if people were continuously getting on and getting off.
    • People are already quite attached to their cars, and I'd suspect the automotive and petroleum industries have quite a bit more political clout than the nonexistent mass transit industry. So funding is more likely to go to big road projects than to big transit projects.
    • And without funding, public transit is stuck at the same level it is now: busses that run on extremely sparse routes every half hour at most, and sometimes just don't show up.
    All of this means that public transit systems have to fight a hard uphill battle to keep the minimal level of service they have now. And this level of service isn't encouraging; when it takes six times as long to take the bus as it is to drive, anybody with a car will prefer to use it at almost any cost.
  • Edinburgh Parallel Computing Centre [ed.ac.uk] had a group several years ago working on this problem using big iron hardware - they have now downsized their software from MPP's to workstations and PC's, and it's being flogged by a commercial company called Quadstone [quadstone.com] Last time I saw a demo of Paramics (c. 1996) it was doing the whole of central and suburban Edinburgh (inside the A720) in close to realtime on a then state of the art SGI workstation, and they had just put in modelling for different nationalities' driver tempraments.
  • I'm sure the Russians thought about using supercomputers to model bread lines before they gave up and decided to let market forces take over. Here's my prediction, based on empirical observation, which can be a remarkable tool for predicting the output of computer models: if you tax people to build something, and then give it away free, there will always be shortages.
  • A friend of mine who has been to Europe tells me that they have similar systems there but they are much smaller and drivers don't have to slow down at all.

    I don't know what size the american transponders are, the kind we use in Norway is about the size of two long fingers and is usually hidden behind the rearview mirror. The system has been tested up to at least 250 km/h (156 miles/h) which is way above the speed limits.

    It is meant to work at any speed, but athorities tend to set a 60 km/h (37 miles/h) limit anyway because there are always some who have to stop and pay with coins and they don't want a mix of high and low speed traffic. :-(
  • It is well known that slow traffic has higher throughput. Sure, cars going at half speed moves slower and take twice the time to get there. But you can fit more than twice as many cars on the same length of road that way, hence the increase in throughput. (Distance between cars is roughly proportional to braking distances, wich goes up as the square of the speed)

    This makes a road largely self-regulating. You can put in more and more cars, traffic slows down but throughput increase enough to accomodate them. Unfortunately this breaks down as distance between cars approach zero. Cars aren't zero-length.
  • Hopefully my employers will never find out about this, but I think the only real solution is to put fewer cars on the roads. Mass transit will help do that, but here in the US you're considered some sort of second class citizen if you ride the bus to work

    Everybody on motorcycles! They use less space.
  • Oh yeah, I forgot the mention, the carpool only lane is completely empty.

    You answered your own question, Alan.

  • In order to keep traffic moving on the M25 (London orbital motorway) they have introduced variable speed limits. As the traffic get heavier, the speed limit gets lower. Supposedly this stops the traffic flow from becoming "turbulent", and so everything keep moving.

    Unfortunately, the way it is enforced is rather scary. They have electronic cameras placed at intervals along the road. The cameras do OCR on each car's numberplate as it passes, and if your average speed exceeds the limit, you get a fine+points through the post a few days later. Being digital, the cameras never run out of film, and as they track your average speed, the normal trick of slowing down when you see a camera and speeding up inbetween does not work.

    Apparantly the caught 4300 speeders on the first day the system went live. And of course, tracking cars all over London via their numberplate isn't an infringment of civil liberties. Honest.
  • There are not a bunch of good books that directly address traffic simulation. This is in part because there is a massive mis-match between the research community and the practitioners doing the work. It is also because there is no general acceptance as to how to do this and how to measure it.

    Search Amazon or Barnsandnoble with the keywords "Transportation Planning" or "Intelligent Transportation Systems" for texts.

    Finally, because this is a simulation, the lessons taught by: Simulation Modeling and Analysis by Law & Kelton (ISBN 0070366985) must be observed. TranSIMs perhaps needs a refresher in the basic statistics that control all simulations.
  • This is a real (no pun intended) problem. When a major arterial runs through many jurisdictions, there is no coordination between them, even if there is coordination within the jurisdiction. There are lots of reasons for this and the industry is attempting to address some of them. One is that two adjacent cities may not be using the same vendor's equipment so therefore they can't communicate with one another. I am involved in an effort to standardize the comm. protocols used by all manufacturers. This new protocol suite is called NTCIP and is based on HDLC, TCP|UDP/IP, and SNMP. If you're interested, there's a website at http://www.ntcip.org

    btw, some terminology - the diamonds you refer to are called 'loops'. They are literally loops of wire in the road surface that senses when a large piece of metal passes over them. When a traffic controller is using detectors (loops or otherwise) it is said to be actuated.

    aj
  • Just the important ones. ;)
    Seriously, I wonder if these folks have taken into account the possibilities of sporadically placed street signs.
  • That is called a "pressure wave" or "shock wave". One slowdown ripples backwards in the flow. There can be positive feedback to make things wors, such as 3 lanes to 2 lanes; the funnel effect causes one wave slowing traffic in the 2 lanes but the lane which ends is empty because everyone merged...a car hops out of line, runs up to the front of the line, jumps back in where the lane ends, causes another wave...
  • If cars moving freely along the freeway equals a fluid, and cars stuck in a jam on 42nd street equals a solid - just heat the cars up to get them to change state to a fluid? Any idea how much energy we would have to put in to get this to work?

  • ...especially the processing power available at Blue Mountain. Did I read that right? 6200 processors?

    That they can put this kind of effort in is really interesting, and the practical applications like variably-timed onramp traffic signals and wireless traffic co-ordinaters sound useful, is it really worth all the effort? Why not just telecommute or encourage people to use mass-transit?


    Who am I?
    Why am here?
    Where is the chocolate?
  • If you think about efficiency in the roads
    around where you live, it's pretty clear that
    most of them arn't particularly designed well.
    An ideal road system would minimize the number
    of stops neeeded to go somewhere, as this would
    both reduce waste of fuel (F=MA), wear on
    the brakes, and reduce time spent behind
    stoplights (or otherwise waiting for other cars
    to act). IMO, tossing a supercomputer at the
    problem is a bit silly when the problems are
    obvious.
  • The folks at Maxis (creators of the SimCity family of "games") could very likely be of some help here. From the reading that I've done, their simulations of traffic flows and city modeling behavior is actually pretty sophisticated, though, obviously on a much smaller scale than these people are talking about. Regardless...their experiences with the SimCity "games" might be useful.

    Jeff

  • Heating isn't exactly what makes a solid into a liquid. More precisely, it is the _exciting_ of the molecules. So instead of heating the cars in a jam, we just need to get them excited! I don't know about cars, but what works for me is a little fender bumping, some hood rubbing, flashing headlights, etc.

  • The Scientific American Frontiers program on PBS here in the US did a show awhile about answering some of life's little questions and one of the segments was about this topic. They talked to some guys in Atlanta (I think) where they modeled a huge traffic tunnel to find out the best ways to get traffic jams under control and how to take care of accidents. They also did a talk with the guys from LANL who modeled their entire city. They even gave the city people with tasks to perform and places to go in certain time frames. It was really neat and if the show reruns in your area, try to catch it. As a semi-side note, the Scientific American Frontiers program is usually really good, it's one of my favorites.
  • More Tourists who don't know where they are, clogging up roads more often then during the off-peak months.

    Tucson is the opposide, Winter time is the worst (all the vacationers come home.)
    *Carlos: Exit Stage Right*

    "Geeks, Where would you be without them?"

  • The only thing that will ever "encourage" people, at least in the states, to be more efficient when driving is $10/gallon gas


    Oo!oo!I know this one! There has to be a $6/gallon tax (putting gas at about $6.75, I suppose) for drivers to take into account all the consequences of driving: pollution, increased death rates, The Gulf War, etc. This according to some study I read for work in July. Can't remember the source, though.
  • The only thing that will ever "encourage" people, at least in the states, to be more efficient when driving is $10/gallon gas

    Oo!oo!I know this one! There has to be a $6/gallon tax (putting gas at about $6.75, I suppose) for drivers to take into account all the consequences of driving: pollution, increased death rates, The Gulf War, etc. This according to some study I read for work in July. Can't remember the source, though.
  • I'd prefer a PRT system like Taxi2000 [taxi2000.com]. Walk less than 4 blocks, push the button to call a vehicle, while waiting tap the touchscreen with my destination, get in the car. The vehicle goes automatically to the stop nearest my destination. The vehicles are for 3-4 riders and do not stop to pick up/drop off more riders.
  • The traffic sensors can fail so the traffic light designers may make the light change every few minutes to prevent trapping vehicles.
    Or the traffic sensors are overly sensitive.
    Or the traffic designers put in some light changes to keep that light synchronized with the rest of the traffic flow.
    Or you simply did not see the car which hit the sensor, turned right on red, and vanished from sight before you came along and got the red light.
  • Feel free to improve LinCity traffic...
  • When the pipe narrows, the water molecules rub together and the pressure increases. I guarantee that when cars rub together the pressure increases.

    The advantage of computer-controlled cars is they could chat and agree to do things such as alter spacing and balance between lanes, and cars would not jump ahead in the line to the detriment of everyone behind them.

  • More exciting car options to increase traffic speed:

    6. Lower center of gravity or lift kit (pick one.)
    7. Electronic defensive and offensive counter measures (radar guns are more useful than detectors!)
    8. Full bottle of N2O (gas masks not included.)
    9. Air horns.
    10.Dual 3" exhaust.
  • Russians Develop New Tires to Combat US Threat

    CNN today reported that a massive increase in tire production by the Russians has strained diplomatic relations. The situation has been made worse by the prospect that China may be retooling to produce even smaller and more efficient cars. The Whitehouse had no official comment today, but rumor has it that a new SUV will be released to combat the communist threat. "Citizens need to be aware that driving is no laughing matter! The future of democracy may well be at stake!" However, a random American was quoted as saying "I don't care, as long as I get my super-sized cup holders on time!"

    --

All trademarks and copyrights on this page are owned by their respective owners. Comments are owned by the Poster. The Rest © 1997-2010 Geeknet, Inc.