How to Hail a Passenger

A new study uses GPS data to analyze taxi-driver performance

Photo: Huang Zongzhi/Xinhua/Landov

5 April 2011—Sure, your smartphone can help you find a cab, and it probably uses GPS to do it. But now that most taxi drivers are also equipped with GPS, what kinds of smart applications are helping them find you? The millions of taxi GPS devices are creating huge stores of data that researchers are mining to solve routing problems, improve dispatch operations, and shake up the economics of transportation.

Research reported in March at the IEEE Computer Society’s 2011 PerCom conference gives some new insight into a little-studied corner of taxi systems—the decision-making and behavioral tendencies of drivers who are looking for the next passenger. Seven researchers, hailing from France’s Institut Télécom SudParis, Zhejiang University, and Hong Kong University of Science and Technology, used a year’s worth of GPS data from taxis in Hangzhou, China, hoping to determine whether hunting for passengers or waiting for them generates more business for a driver. The research is a step toward GPS-based software that can recommend more profitable behavior for cabbies, but whether that behavior will be good for both drivers and riders is an open question.

The study aimed to answer two questions: Does the time of day or location affect the outcome of hunting or waiting? And is a driver better off staying in the location of the last passenger drop-off, or returning to a busy location, even if it’s far away?

Science aside, many taxi drivers seem to already intuitively know the answers to those questions. Undercover video released in March shows that New York City taxi drivers will go to great lengths to avoid taking passengers to locations in distant parts of the city. Some drivers claim that they don’t have a GPS, refusing rides to passengers, even though such refusal is illegal. Drivers who don’t have a choice regarding destination might welcome insights that help them find the most profitable route back to a busy spot.

The Hangzhou study confirms that cabbies avoid certain destinations for good reason: A driver behavior that consistently led to low pickup performance was lingering too long in a sparsely populated place after a drop-off.

But further data mining produced more nuanced insights: The researchers found that hunting in the early morning, before 7 a.m., resulted in more passenger pickups, while waiting in busy areas could result in more pickups during the height of rush hour, from 6 p.m. to 7 p.m. (Important to note, however, is that some of the data the researchers used was hard to parse: An immobile cab at rush hour might be stuck in traffic, rather than deliberately waiting for a customer, especially in the evening hours, the researchers say.) The top 10 performing drivers—those who had the highest number of passenger pickups overall—consistently chose to hunt for passengers or drive back to a busy area after drop-offs. Eight out of 10 of the lowest-performing drivers chose to wait or stay in the location where they dropped off their last passengers. Waiting, then, is associated with fewer pickups on average.

Daqing Zhang, who led the study, says that the novelty of his group’s research lies in "how to guide the taxi drivers to choose the right strategy and optimal route to find the next passengers." Zhang’s future research will focus on the route the driver takes when the cab is empty. If it’s generally true that hunting and driving back to busy locations results in more passenger pickups, then cabbies would do even better by choosing the best paths while on the hunt or while driving back to a busy location. An application that could suggest such optimized routes might, in theory, make individual taxi drivers richer. But if every driver starts hunting on optimized routes, the distribution of cabs might change in ways that aren’t optimal for the overall balance of a city’s taxi system.

Several years ago, joint research conducted by a group from Kyoto University and the University of California, Davis, showed just that. They developed a model to better understand the behavioral tendencies and economics at play in a taxi driver’s decision either to cruise for passengers or wait in a taxi bay. But in the Kyoto-Davis study, the tendencies of the individual were considered in relation to taxi availability citywide. The examples, derived from taxis in the Nagoya metropolitan area of Japan, showed that while cruising might increase an individual driver’s overall performance, it could have an adverse effect on "socially motivated system optimization." Taxi drivers have a propensity for cruising—and for good reason—but too many cruising taxis leads to increased competition and inefficient system performance.

The performance of a city’s taxi system will also be increasingly affected by a growing base of smartphone applications that are geared to the ride-seeking strategies of the passenger rather than the driver. CabCorner and Weeels are examples of new applications that encourage taxi sharing among users with similar origins and destinations, with the promise of conserving both energy and money. These start-ups haven’t yet found a critical mass of users, which critics speculate is because of passenger attitudes about privacy and an aversion to riding with strangers. However, if these start-ups succeed, they may affect cabby profits.

In cabs, the capabilities of GPS are bound to become increasingly coupled to intelligent recommendation systems. Will arming both riders and cabbies with GPS-based recommender systems shift the balance of supply and demand toward more efficient taxi systems? We’ll have to wait and see.

Related Stories

Advertisement
Advertisement