Traffic flows in San Francisco Bay, generated at UC Berkeley from AT&T mobile data. Alexei Pozdnukhov, UC Berkeley

How AT&T is using drivers’ cellular data to help fix California traffic

Updated: Oct 16, 2015 5:47 PM UTC

Los Angeles’ snarled, rage-inducing roads have been infamous for decades. And now, thanks to a tech industry-fueled population explosion, San Francisco is right behind L.A. in the title race for Worst Traffic in America.

AT&T (t), UC Berkeley, and California’s state transportation authority are testing a new way to get a grip on the situation—by collecting and analyzing drivers’ cell phone location data. The study leads insist that users’ privacy is protected, and the information could revolutionize how we plan and manage highways and transit.

“The idea of using cellular data for mobility is not very new,” admits Alexei Pozdnukhov, assistant professor in UC Berkeley’s Smart Cities program. “What is new . . . is that our approach is much more detailed modeling. We can simulate very detailed scenarios, and answer questions.”

Really understanding commuters’ behavior has been a challenge up to now, in part because the available data, such as that from induction loops embedded in asphalt, has been mostly limited to counting the number of cars that pass over a single point.

In contrast, “The data that you get out of third parties can tell us more immediately what’s happening on the road,” says Dr. Nick Compin, who manages the Connected Corridors pilot project for Caltrans, which operates California’s highways.

The new California projects —Connected Corridors in Los Angeles, and SmartBay in San Francisco — are something like Google Maps on steroids. They compile region-wide cell data into big portraits, not just of where traffic is most congested, but of overall daily patterns.

“[It shows] where people . . . work, where they go for shopping, where they go for leisure, and how they choose to get there,” says Pozdnukhov. Dr. Compin says that’s “the holy grail” of transit planning.

MORE: Smart cars, meet smart signals

The data will help planners develop detailed responses to congestion events — Compin says there are a stunning 5,000 to 6,000 events per year on the I-210 corridor, making up about 50% of traffic delays. By working closely with local authorities and public transit providers, Caltrans hopes to make better decisions about how to re-route traffic onto parallel corridors and local roads, and communicate changes to commuters more smoothly. The San Francisco pilot is centered on Interstate 80, and among other things, says Pozdnukhov, hopes to determine the potential impact of increased development on the Treasure Island neighborhood the highway passes through.

Compin says the basic idea is to avoid building new highways by making better use of the capacity that’s already built. It’s all part of a much larger trend to make data a bigger part of transit management — what some are calling “the internet of moving things.”

Concerns about security and privacy are inevitable, but, much like Google, AT&T and Berkeley say there are several layers of protection between any individual and the aggregated traffic data. First, AT&T doesn’t give Pozdnukhov’s team GPS information — just users’ movements between cell tower coverage zones. Then that data goes through another layer of anonymization, when the Berkeley team plugs it into a traffic model.

“The models work at a certain level of [abstraction],” says Pozdnukhov. Location data is randomized within neighborhood blocks of about 3,000 residents. The resulting overview summarizes the group’s habits, but not those of individuals. Finally, only the output of any simulation is sent on to planning authorities — never the original data.

Whether data from a single cellular provider can be used to make decisions for all travelers is another important question. Chris Volinsky, head of big data research at AT&T Labs, says the carrier has between 25 and 50% of market share, depending on the region. The team extrapolates from that, and adjusts for factors like income, since it under-represents groups using prepaid or feature phones.

Caltrans hopes to expand this kind of next-generation coordination across California, while Volinsky says AT&T (which is pushing hard on the broader Internet of things) plans to eventually offer similar services to other transportation planners.

“If you look at a crystal ball, you’re going to see this across every major urban corridor in the nation,” says Compin. “They’re not going to build more. They want to manage what they already have and make it as efficient as possible. So this is just the tip of the iceberg.”

Sign up for Data Sheet, Fortune’s daily newsletter about the business of technology.

For more Fortune coverage of data-driven traffic solutions, watch this video: