Experiment Shows Traffic 'Shock Waves' Cause Jams 642
Galactic_grub writes "Japanese researchers recently performed the first experimental demonstration of a phenomenon that causes a busy freeway to inexplicably grind to a halt. A team from Nagoya University in Japan had volunteers drive cars around a small circular track and monitored the way 'shockwaves' — caused when one driver brakes — are sent back to other cars, caused jams to occur. Drivers were asked to travel at 30 kmph but small fluctuations soon appeared, eventually causing several vehicles to stop completely. Understanding the phenomenon could help devise ways to avoid the problem. As one researcher comments: 'If they had set up an experiment with robots driving in a perfect circle, flow breakdown would not have occurred.'"
stability (Score:4, Interesting)
Is that true? If the robots had been fixed to a set driving speed (open loop), maybe. But if the robots had some sort of collision avoidance, it could still happen. It's instability in the control algorithm, no?
Re:Brakes. Not breaks. (Score:5, Interesting)
I wonder how much aggressive driving (someone speeding up to 90, and then cutting in front of you for seemingly no reason), contributes to breaking shock waves. I've seen it happen often enough where someone will make an unnecessary maneuver to get 30 feet ahead of traffic.
Wow, big news. (Score:3, Interesting)
Then again, I remember seeing stuff like that back at the university, where they were trying to combine traffic models with a Kalman filter to achieve better traffic jam prediction. That was, uh, over five years ago.
robots (Score:3, Interesting)
I, for one, welcome our new japanese robot driver overlords.
but seriously, I take this as a hint as to what is to come in the future for japan.
Re:Interesting, but not a solution (Score:5, Interesting)
Breaking or other external factors (an accident or flashing lights by the side of the road) can certainly precipitate a change from a swiftly moving flow to a slow moving flow. However, they only cause a transition when the density is high enough. If there's an accident during a low traffic time, you whiz by it. If they close two lanes out of four, and it's low traffic, you get a little backup, but it reaches a modest steady state size in low traffic. In high traffic you get a "wave" - the backup moves steadily backwards from the scene of an accident, and remains after the accident clears.
I often tell my wife that I can tell if a slowdown is just due to high volume or an accident by the abruptness of the slow down. An abrupt slowdown, I think, means heavy traffic "precipitated" into a jam by an external event.
So braking as described may be a precipitating event, but it's the sensitivity of the traffic flow to it that is the fundamental issue. I'd guess that even if people didn't brake so much, in those sensitive conditions a fender bender by the side of the road could cause a major backup.
(Clearly, I've thought about this WAY too much.)
Speed of light trumps wave speed (Score:5, Interesting)
But I could see a half a mile of cars all with little green lights, I could see (at the speed of light) the wave of yellow lights approaching and ease off the gas. The wave would be absorbed by this 'viscosity'. Traffic would flow near the speed limit or average flow rate, whichever the LED's were keyed to.
And don't even get me started on those GPS nav screens. Don't show me were I am. Show me where everyone else is. Let me see the compression 2 miles ahead and I'll chill (heh heh kinetic gas pun).
"Those who forget the past", etc, etc, etc (Score:4, Interesting)
Guys, there really is a benefit to hitting the library and thumbing through back issues of ld technical journals.
Modern Marvels (Score:4, Interesting)
Re:Brakes. Not breaks. (Score:2, Interesting)
Even older than that (Score:3, Interesting)
Re:That's why I never use my brakes (Score:4, Interesting)
Re:Brakes. Not breaks. (Score:3, Interesting)
I remember getting this as a result from a simple discrete model written in Turbo Pascal as far back as early 90-es. No need to make volunteers drive cars. Once the traffic exceeds a certain density waves and fluctuations in it will show up straight away. There is even some math proof of the instabilities in mass service theory. It's been a while so I cannot remember.
Anyway, this is Japanese science. Anyone who has had to suffer from reading a Japanese publication knows what I am talking about. Phenomenal engineering, insane experimental precision and with all due respect lousy science (unless it is done by an imported foreign devil).
Re:stability (Score:3, Interesting)
Absolutely. Last semester in my graduate robotics class, I had robots follow each other through loops (eventually meant to simulate an intersection control technique). I used collision avoidance on each robot.
I first tested each loop in simulation. The robots would all start at the same acceleration. At a certain saturation of robots, the whole system would break down due to the "waves" of traffic congestion caused by collision avoidance.
Mind you, this was with simulation that was nearly perfect (the only imperfections would have been in thread timings).
Re:Brakes. Not breaks. (Score:5, Interesting)
Maybe not, but there's something your highway authority can do about it: Adopt German rules. Passing on the right gets you a ticket; driving on the left without passing gets you a ticket.
Flipping someone off gets you a ticket too, but that's another issue.
rj
Very cool traffic sim (Score:5, Interesting)
http://vwisb7.vkw.tu-dresden.de/~treiber/MicroApplet/ [tu-dresden.de]
shame this post is buried down deep
Re:Brakes. Not breaks. (Score:2, Interesting)
Sadly, automated traffic lights train people not to do this. If traffic is flowing with a lesser interval, the first time an interval of that size passes over the sensor, it triggers a light change. Result: the car leaving the large gap just makes it through the light, and everyone behind him gets screwed.
Therefore, drivers that have learned this by observation make every effort NOT to be behind the guy leaving a gap.
Re:Brakes. Not breaks. (Score:3, Interesting)
as for why people go all the way to the left when they get on the high way. it's just a zombie driver mentality thing. They _THINK_ they're a "fast driver" and that lane is for them, or they think that by being in that lane they're allowed to go faster.
Some drivers just go one speed no matter what road they're on. For example: last year I needed to follow my aunt somewhere, she warned me that she was a "fast driver" and that I should do my best to keep up. On the back 30MPH roads she barreled down at about 50MPH and tail gated any slower moving traffic we encountered, we got to the highway... and she still drove at about 50MPH in the left lane holding up all the traffic behind her.
In NH we have some other problem drivers too, we have long stretches of single lane highways with lots of passing zones. but all too often people are too chicken sh*t to pass and just slow down when they encounter slow moving traffic. What ends up happening is you'll get 2 or 3 of these people stacked up and now there are too many cars to safely pass with average on coming traffic. On more than one occasion I've passed 12+ cars in one shot that were putting along at 35MPH behind a tractor or dump truck in a 55MPH zone, though unless there is no on coming traffic at all it can be a rather dangerous maneuver. Completely avoidable if the slow vehicle would just pull over, or other drivers would grow a pair and pass. I know lots of people who claim they never pass because they're afraid to pull into the oncoming traffic lane even when it's completely clear.
Re:Not that simple (Score:3, Interesting)
I've talked with the local head government transportation guy, and he said the term that they called this phenominon is the "accordian effect".
I think main motorways should be allowed to use the shoulders when traffic is backed up (w/o rumble strips). But you can get in some legal trouble if you do that today.
I also just think that cars don't scale well, and we just need a better way to get from point A to B.
Traffic flow analagous to supersonic flow (Score:3, Interesting)
Of course the analogy breaks down because... well it isn't the same phenomena. Each molecule of traffic has an emotional person controlling it.
Some of my observations:
When traffic reaches what I call critical flow (a combination of high density and high speed) then it doesn't take much disturbance at all to cause shocks (a shock being a rapid decrease in speed combined with an increase in traffic density, they are characterized by lots of brake lights.) When traffic is at critical flow, cause and effect can have a very non-linear relationship. Even a polite lane change, or a pothole might tip the traffic from critical to supercritical (traffic jam.) Sometimes shocks are standing shocks. This tends to happen approaching a constriction in the road when traffic sufficiently far back is sub critical. Sometimes shocks propagate backwards through traffic like waves at the beach. An excellent example of this is 880 Southbound approaching the turnoff to the San Mateo Bridge around commute time. This one is interesting because there is drag introduced by people exiting 880 but the main disturbance comes, I believe, from cars in the carpool lane cutting across all the lanes to exit to the San Mateo Bridge. As anyone who has driven this stretch of freeway can tell you, traffic blasts along at 70 then everyone is standing on their brakes, then 70, then brakes, then 70, then...
That was the best of my macro-traffic observations, here is one funny micro traffic observation. I call it the "sticky effect" or the "stupid effect" depending on my mood. If one car is overtaking another car but only going slightly faster the slower car will speed up at least temporarily to match speeds. If the passing car is going sufficiently faster then it won't happen. Two cars on a two lane road will frequently end up right next to each other as a result of the sticky effect. Causing a "moron roadblock," which is just a line of cars going too slow blocking every lane. I also call it a rolling roadblock, and when I lived in Utah I called it a mormon roadblock.
Drive safe!
Re:Brakes. Not breaks. (Score:3, Interesting)
A potential solution (Score:-1, Interesting)
The goal is throughput, not individual speed- get 80,000 morning commuters from point A to point B as fast as possible, not allow 1 jerk that is going too fast or too slow screw it up for everyone else. One way to accomplish this is to have "pace cars" that travel at a steady rate (something like 55 or 60mph). Cars queue up behind them and travel at a relatively constant rate. The pace cars can control their speed and keep the individual "waves" seperate. If a car travels too slowly and a pace car catches up to it, it gets a ticket. In practice, this system would probably require longer on/off ramps to allow cars to get in and out of the queues, but other than that I think it is a pretty workable system.
Re:Brakes. Not breaks. (Score:3, Interesting)
Standing Wave Theory of Traffic (Score:3, Interesting)
This article is finding many of the same conclusions I had back then. Is there a fix? I don't know but traffic on a large scale is fluid.
God help us when we have flying cars and we have to deal with idiot drivers above us and below us!
Re:Brakes. Not breaks. (Score:1, Interesting)
Re:Not that simple (Score:5, Interesting)
What happens if there is a break down? You are back to three lanes of traffic, just like there would have been if you had never allowed driving on the hard shoulder in the first place. The rest of the time, you get an extra lane.
Re:Not that simple (Score:3, Interesting)
Re:Not that simple (Score:3, Interesting)
Right, I first learned about this from my GCSE Chemistry teacher in about 1986. He'd been in North Africa in the war, where supply convoys could easily be a couple of miles of a single line of traffic. Each jeep set off when it saw the one in front move, which meant that the ones at the back had to go hell-for-leather to catch up, only to realise that the ones at the front weren't going all that fast. There were regular ~100-jeep pile-ups until they instituted a system where everyone starts their engine and raises their hand when the driver in front does, then drops their hand and moves off when he does that -- which can be seen from a much greater distance along the line.
Peter