How carpool apps could take 10,000 taxis off NY streets

If we were all willing to use the carpool option in apps like Uber and Lyft, MIT researchers say it would transform city streets.

Research shows exactly how threatening ride-hailing services could be to taxis.

Declan McCullagh/CNET

The uneasy relationship between taxi companies and ride-hailing startups got a little more tense today thanks to research from the Massachusetts Institute of Technology.

It's no secret that carpooling can help reduce traffic. But what if everyone in New York turned on the carpool option in ride-hailing apps like Uber and Lyft? Researchers at MIT found that just 3,000 four-person cars could serve 98 percent of the demand currently being met by almost 14,000 taxis.

Researchers from MIT's Computer Science and Artificial Intelligence Laboratory (CSAIL) fed data from 3 million New York City taxi rides into an algorithm that imagined those rides as carpool requests via a ride-hailing app. The computer model routed and re-routed all those rides and found it would only take 3,000 cars to meet most of New York's demand.

What's more, this boost in ride efficiency comes at a minimal cost to convenience. In the MIT researchers' models, no more than 5 minutes would be added to your trip in order to wait for a car along your desired route and pick up and drop off other passengers.

"High-capacity ride sharing has the potential to reduce the number of vehicles, the distance traveled by vehicle and the waiting times," said lead researcher Dr. Javier Alonso-Mora, who is now at the Delft University of Technology. "Therefore, improving the quality of life, the earnings per vehicle and reducing the cost for the passengers. All at the cost of just a few minutes delay, equivalent to parking a car."

The team also looked at different vehicle options. They found that 3,000 two-person cars could serve 94 percent of demand and just 2,000 ten-person vans could handle 95 percent of demand. Their findings were published Monday in the Proceedings of the National Academy of Sciences.

The MIT algorithm running a simulation of New York taxi requests being filled by carpools.

Video screenshot by Eric Mack/CNET

The takeaway: if we were willing to add a few minutes to our trips and rub elbows with other passengers on the regular, we could take thousands of cars off crowded streets.

In case you're wondering, the research was funded by grants from the Office of Naval Research and an MIT-Singapore partnership to explore the future of urban mobility. Not by Uber or Lyft.

It should also be mentioned that thousands of driving jobs would be eliminated if this computer model became reality. Even worse for professional drivers, the algorithm developed by the MIT researchers would work best when used with a fleet of autonomous vehicles that can easily be re-routed on the fly.

"The algorithm finds the best assignments and routes. It is indeed true, therefore, that it can have vehicles running non-stop for many hours (as a self-driving car could do)," Alonso-Mora explained. "Additional constraints could be added to the algorithm to account for the preferences of human drivers and dispatchers, reducing the quality of service."

The MIT algorithm presumes ride-hailing apps that can handle a little more carpool complexity than they currently do, such as being able to modify a route or add ride requests continuously.

If the idea of carpools replacing taxis sounds silly to you, consider that in many cities half of the Uber rides taken are via the UberPool option. We could soon be getting much more serious about all those times we've been told to share the road.

Close
Drag