December 2005

Traffic Avoidance

A new service from Kirkland, WA-based Inrix predicts traffic slowdowns by crunching road sensor data, weather, history, and local events.

By Erika Jonietz

In the interminable battle against traffic, a growing number of government and private initiatives offer U.S. drivers high-quality real-time traffic data and even short-term predictions of travel time from, say, one freeway intersection to the next.

But most of the forecasts don't extend beyond 15 or 20 minutes. Though a veritable traffic jam of companies has sprung up to offer data, they generally inform commuters of snarls as they occur, which is often too late for drivers to change their plans.

Now, actual traffic prediction -- forecasts of congestion levels hours and even days in advance -- is on the horizon. It's coming from Kirkland, WA-based Inrix, founded in 2004 by former Microsoft executives Bryan Mistele and Craig Chapman and former Expedia executive Seth Eisner.

The company uses algorithms that originated in the labs of Microsoft Research; its technology is the first fruit of Microsoft's initiative to license intellectual property to venture capitalists and startups.

The Inrix software starts with a mass of data obtained from government agencies -- real-time traffic flow and incident information from gadgets installed on highways, including toll-tag readers, cameras, radar units, and magnetic sensors embedded in the pavement. Inrix then adds speed and location data from computers and Global Positioning System (GPS) units in vehicles owned by trucking and delivery companies. These vehicles effectively act as mobile sensors, and Inrix buys the data they collect. Finally, Inrix adds up to two years of historical traffic flow data, weather forecasts and conditions, and even local road construction schedules, school calendars, and dates of events like concerts and athletic contests.

The company's proprietary statistical models combine all this data to provide not only a snapshot of current traffic flow but also predictions about expected congestion and road conditions over the next several hours and even days. Each city requires its own unique model; the model for San Francisco alone contains about half a terabyte (500 gigabytes) of data, says Oliver Downs, Inrix's chief scientist.

Inrix plans to have models for the 30 largest U.S. cities available by the end of 2005 and to provide traffic predictions to drivers through partnerships of various kinds. It announced its first partnership, with digital-mapping company Tele Atlas, in September.

Tele Atlas will offer Inrix services to all of its customers, which include companies such as MapQuest and T-Mobile Traffic. Inrix plans additional partnerships, with companies such as cell-phone operators, traditional and satellite broadcasters, and in-car navigation services.