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."
Besides, as far as I understand, the system currently only considers highways, which (at least in Germany) have at least dual carriage ways. If the traffic is not dense, most accidents or break downs do not cause jams, as the traffic simply passes by.
Sebastian
Bill Beaty investigated how to avoid traffic jams by recognizing the intermediate state of synchronized flow and undoing the damage six years ago. Apparently, a traffic jam can be stopped, even once started, by a single car.
Traffic Waves
When building software like this, you do a great service to take us R/G colorblind people into account. We're not as rare as you think! BWHAHAHA! ;)
It only applies if a large portion of the potential traffic was aware of the prediction
Not necessarily. Depending on the situation, the threshold number of people whose behavior can change the situation will vary; could be large or small.
Where I live (Seattle) the state DOT puts up a GREAT online traffic map, which I check religiously before hitting the road. I hope it someday incorporates technology like this.
As I read your comment, it struck me that the flow of cars past a given point resembles the flow of water from a drippy faucet, and that both can be modeled as a 2D strange attractor. Just as you can map the time differences between Drips [ D1, D2, D3, etc. are absolute time of drips; T1=D2-D1; T2=D3-D2; etc.], you can map the space differences between cars. Then map it on cartesian coordinates thusly:
Point1: T1, T2
Point2: T2, T3
Point3: T3, T2; etc.
When traffic is flowing smoothly, the difference in time between cars is essentially random, so the map will just be a bunch of randomly scattered dots. But when jams began to occur, the average length between cars shortens, and the dots began to describe a spiral inward. At total jam, the length between cars is, essentially, zero, leading to a dot centered at 0,0.
It's the point when the average length between cars begans to shorten that is when the graph becomes most interesting. Also, this is where intervention might occur "on the ground" so to speak.
One drawback of this graph is that it describes only a single location along the roadway. You'd need many of these graphs to simulate a large traffic situation.
And, as to the poster's note that People are just flat out unpredictable., that's only true in the *individual* sense. In the aggregate sense, people are *remarkably* predictable. This traffic problem happens to be an interesting problem because it involves *both* aggregate (so it's fairly predictable) AND individual (ONE driver makes an error...) behavior.
DNA is a Turing machine. You, however, being dynamic and emergent, are not.
Look at the clock. I predict that every weekday at 7:00am-8:00am and 5:00pm-6:00pm traffic will be at its heaviest. Avoid driving at these times and you will be set. Also, I predict that road construction will cause traffic congestion as well. In the foreseeable future, I also predict unknown random traffic accidents to cause congestion. Use your radio tuner device in your car to listen for reports of these strange incidents and avoid those routes!