PolySciFi Blog

Wednesday, June 29, 2005


Predicting traffic

An interesting piece of software.

A traffic forecasting system capable of predicting traffic conditions seven days in advance will go live to the public in California on Wednesday.

Alongside the weather forecast, viewers of KXTV News 10 in Sacramento can now get 3D animations of their local road network, showing not only where the gridlock is but also where it is likely to be.

The system, called Beat-the-Traffic, is the first public traffic forecasting system that combines real-time traffic density and speed with historical trends on major routes.

I don't expect that the predictions will work too well though.

Unlike when you predict the weather (whose predictions are kinda of spotty anyways), making a public prediction of a traffic pattern is likely to alter the traffic pattern. Specifically, if you predict heavy traffic on road X, numerous drivers will choose to drive on road Y instead. Thus X will not be as congested as predicted while Y will be more congested.

Thus for the prediction to be successful, this tendency to adapt based on the predictions would have to be accommodated. Furthermore there is an infinite recursion where the drivers know the predictor knows the driver knows the predictor knows.... that's kind of messy to resolve particularly when the drivers have 7 days to mull over and change their driving plans.

However, that being said, having detailed and timely information about the current state of traffic and the typical impact of accidents (the other task of the software) can't hurt and should be quite useful.


This page is powered by Blogger. Isn't yours?