Quantcast
3000 Ubers Could Replace NYC’s Entire Taxi Fleet

There’s just one catch.

A new study released Monday by MIT's Computer Science and Artificial Intelligence Lab (CSAIL) has found that just 3000 ride-share vehicles could do the job of New York City's roughly 13,500 licensed cabs. While this is a remarkable finding about the potential in ride-sharing, there is a catch—it only works if all the passengers are using the carpool option offered by Uber and Lyft.

The study used an algorithm developed by the CSAIL researchers that used publically available data from 3 million taxi rides in New York City over the course of one week in 2013. In a model that simulates taxi-requests in NYC, the algorithm was able to work in real-time to reroute ride-share vehicles based on incoming requests and send non-occupied cars to high-demand areas.

Using this model, the researchers found that four-person cars could serve 98 percent of the demand in NYC with an average waiting time of just 2.7 minutes and an average delay of only 2.3 minutes—which is about 20 percent faster than current Uber wait times. When the team adjusted the model for different vehicle types, it found that 3000 two-person vehicles could serve 94 percent of NYC's demand and 2000 ten-person vehicles could cover 95 percent of NYC's demand.

"To our knowledge, this is the first time that scientists have been able to experimentally quantify the trade-off between fleet size, capacity, waiting time, travel delay, and operational costs for a range of vehicles, from taxis to vans and shuttles," said Daniela Rus, a professor of computer science at CSAIL who led the study. "What's more, the system is particularly suited to autonomous cars, since it can continuously reroute vehicles based on real-time requests."

This is bad news for NYC's already embattled taxi drivers, who have spent the last few years fighting Uber's rapid expansion in the city. But it might also be bad news for the Uber drivers themselves, who are facing obsolescence as Uber rolls out its autonomous vehicles, a program that could be aided by CSAIL's algorithm. On the flipside, such algorithms could be a boost to the US economy as a whole, which loses an estimated $121 billion annually (about 1 percent of the US GDP) as a result of the 5.5 billion hours people send sitting in traffic (to say nothing of the 2.9 billion gallons of fuel that is also wasted).

While the concept of carpooling is not at all new to ride-share apps, the CSAIL algorithm vastly improves on Uber and Lyft's proprietary algorithms for organizing carpooling. Normally, ride-sharing systems require that different users be on the same route in order to match them for carpooling and need all the requests to be finalized before creating a carpool route. CSAIL's algorithm, by contrast, rematches users to different vehicles in real-time to optimize the system while also analyzing where high occupancy vehicles can be of the most benefit based on metrics like cost, time and the inconvenience to the passenger.

"Ride-sharing services have enormous potential for positive societal impact with respect to congestion, pollution and energy consumption," says Rus. "I think it's important that we as researchers do everything we can to explore ways to make these transportation systems as efficient and reliable as possible."