The trouble with traffic reports is they’re real-time at best: By the time you hear about the mess, you may well be sitting in it. It’s often too late for you to change your route, much less decide to take the train or stay at home.
But transportation researchers at MIT and the University of Texas at Austin think they can predict where congestion will strike 30 minutes in advance. Two traffic forecasting programs (DynaMIT and DYNASMART-X) are being tested in Irvine, Calif., and neighboring Anaheim, where highways and secondary roads are equipped with embedded magnetic sensors that detect passing vehicles and their speed. Historical and real-time data from the sensors fuel the prediction programs.
DynaMIT tries to predict the behavior of each vehicle. If traffic is light, the computer assumes cars will accelerate; if it’s heavy, the computer predicts lane changes and heavy braking. The software simulates thousands of driver choices every few seconds, and predicts when these choices will converge to create a traffic jam, says Moshe Ben-Akiva, head of MIT’s Intelligent Transportation Systems lab, which created DynaMIT. The University of Texas’ DYNASMART-X is similar, but can make traffic predictions in smaller regions within a larger traffic system.
If testing goes well over the next two years, researchers may begin announcing the forecasts to the public. In anticipation of the system’s catch-22-drivers who hear the forecast may render it moot by changing routes–researchers equipped the programs with a loop that predicts this effect and continually adjusts the forecast.