Sampling for real-time traffic info

To get real-time traffic information, we need to know how well traffic is moving at key points — usually, though major intersections and interchanges, toll plazas, and other points of congestion. Comparing what we know now with “normal” situations, we can figure out anomalies and report them to alert systems or navigation systems.

People have set up the monitoring needed for this in a number of ways. Intersections that already have traffic sensors in the ground can feed their information into the system. Cameras can be added to collect more. But these and other methods that add hardware to the infrastructure are very expensive.

Here’s a way that costs much less, reported in New Scientist:

Researchers at Nokia and the University of California, Berkeley, have developed a system that collects GPS data from mobile phones in moving vehicles and uses it to create traffic maps. The maps are available on the internet or sent to your cellphone to provide local traffic analysis.

Alex Bayen at UC Berkeley says that if enough people download the free software (from http://traffic.berkeley.edu), the system should help relieve congestion, even on small intercity roads. And in case you’re worried that the neighbours will now be able to follow your every move, he says that the system anonymises GPS data so that it will be impossible to track individual cars.

(Here’s an earlier article about the system from the San Francisco Chronicle.)

The mechanism is similar to other synergistic uses of GPS information from mobile phones. Google Maps, for example, collects GPS and cell data when it can from phones that make queries... and uses that information to build a database that maps between cell data and actual location. When it gets a query from a phone that doesn’t have GPS or won’t provide the information, it can use the phone’s cell data to get an approximate location from the mapping database.

Clearly, the general technique of anonymizing location information and using it to synthesize new information — like the location database and the traffic flow — is a very powerful one with which we can create new applications that can help everyone.

The real coup here is that this can give us information all over the roads, without the need for new infrastructure and without limiting us to certain predetermined points that get monitored. As long as enough cars — actually, enough mobile phones with GPS information — pass a given point that the sampling is useful, we have a good measure of the traffic volume at that point compared with other days and times.

In fact, this could be used for early dispatch of assistance to collision sites. When the system suddenly sees that traffic is not moving past a certain point, that point can be flagged as a probable traffic accident, and help can be sent without waiting for a manual report.

What a great idea! I hope to see more of these sorts of applications (with appropriate anonymization, of course; that’s a critically important piece of it).