Time spent accommodating multiple riders is said to be offset by quicker pickup times.
Algorithms developed by researchers at MIT and Cornell could help revolutionize urban mobility by eliminating a large portion of taxi traffic.
A paper published in the Proceedings of National Academy of Sciences journal and spotted by Ars Technica used an actual database of New York City's taxi rides to refine an algorithm that could be used to manage ride-sharing across the entire fleet in real-time.
Such algorithms already exist, but the researchers aimed to improve the technology by minimizing wait time and enabling quick decisions. A so-called 'greedy' algorithm kicks off the first trip assignments by focusing on the longest routes, while further optimizations are simplified to reduce computing requirements and hasten further decisions.
The authors attempted to balance overall system efficiency with considerations that could make or break a service from a customer satisfaction standpoint, such as the amount of time one rider can be delayed while the system picks up other passengers.
The study found that only 3,000 four-seat taxis would be required to serve 98 percent of the real-world ride requests served during a one-week period in New York City, far below the 13,586 taxis that operate in the city. Riders would experience an average in-ride delay of 2.3 minutes to pick up and drop off other passengers, but the time is partially offset by a short 2.3-minute average wait to hail a cab.
Many major automakers are attempting to restructure their businesses for 'mobility,' envisioning a future where ride-sharing and ride-hailing services outpace personal car ownership. Advancements in algorithms that manage such services will be critical in carrying the technology beyond the experimental stage or small-scale niche.