How Taxis Can Double as City-Wide Traffic Sensors

As part of a city's transit infrastructure, taxi data can monitor the pulse of traffic.

Jordan Pearson

Jordan Pearson

Image: Kenny Louie/Wikimedia Commons

Monitoring traffic in a busy city is a tricky and costly business. Traditional sensors like cameras and in-ground sensors are expensive to implement and thus not able to infer wider trends. Taxis outfitted with GPS trackers, on the other hand, could form a powerful and ready-made traffic detection network as they continually speed down every main road and side street in a city.

Previous studies have shown that GPS data from smartphones and vehicles can be used to accurately model traffic flows. Conveniently, many cities—like New York—have already equipped taxis with GPS trackers. Analyzing their data could provide city officials with a low-cost alternative to monitoring traffic that provides enough information to be useful in predicting wider trends and managing flows during emergencies.

Researchers from the University of Illinois at Urbana-Champaign tested the ability of taxis to infer traffic patterns by analyzing the GPS data from 700 million taxi trips in New York City over a four year period.

The immediate goal was to see how taxis responded to the impact of Hurricane Sandy, which caused extensive flooding and resulted in massive evacuations throughout the city. If trip lengths became significantly longer, for example, that would indicate that all traffic—not just taxis—had slowed to a crawl. This information could then be used by officials to identify traffic choke points and better manage roads and transit routes.

According to their results, the greatest point of congestion wasn't during the storm itself, or even when the flooding was at its worst. Surprisingly, taxi trips were shorter in Lower Manhattan at the height of the flooding, likely because of the lowered demand for fares. Instead, traffic in the city was at its worst several days after the storm had passed due to the influx of returning evacuees.

"This is unfortunate, because this is the time when first responders need to be quickly getting around the city to clear debris and aid people, and they can't because the roads are so congested," Brian Donovan, one of the paper's authors, told me.

For several years, the US Department of Transportation's Federal Highway Administration has openly discussed the need for better, and cheaper, methods for monitoring traffic. According to the FHA's Traffic Detector Handbook—a fun read if if you're bored on a Sunday afternoon—inductive loop sensors are the most widely used method of traffic detection, but the data they provide is not rich enough to make large-scale inferences about speed and traffic density. Collecting the data from GPS-enabled taxis could remedy these issues with current traffic monitoring techniques.

"If all your taxis are stuck in a traffic jam, that means that there's a traffic jam potentially affecting people who aren't taxi drivers. It really gives you a measure of how well the infrastructure is working," Donovan said. "Most taxis already have GPS sensors in them. It's good if you can just use that data instead of installing more loop detectors, because you have to tear up the roads, it causes disruptions, and it's very costly. GPS is a way to monitor infrastructure without buying anything expensive or disrupting anything."

Speaking of disruption, Donovan and his colleagues' findings show there may be unplanned consequences to the recent trend of startups like Uber and Lyft displacing city taxis. These services have been criticized by governments at home and abroad (they're banned in South Korea, Germany, and Spain) for their reluctance to conform to the same standards as cab companies regulated by a municipal body. A lack of driver training and in-car cameras are current examples of such complaints, and the inability to contribute to low-cost city traffic monitoring could be one of the future.

While the GPS data of Uber cars could conceivably be gleaned from drivers' phones, accessing that data set may not be as easy as asking the city was for the University of Illinois researchers—and it wasn't all that easy in the first place. They had to file a Freedom of Information Law request.

"When you want to measure something and you need data, you always need to have people who are willing to cooperate with you. It can be difficult to get that from private companies," Donovan said. "And governments, too," he added.

Like so many other devices—including our smartphones, which can track cosmic rays en masse—data collection could turn taxis into a mobile information network with practical applications. If those familiar yellow cars are still around in a few years, anyway.