Quantifying the Potential for Dynamic Ride-Sharing of New York City's Taxicabs

Abstract

In recent years, mobile application-based transportation businesses like Uber, Lyft, and Via have redefined the notion of ride-sharing, scrapping the traditional impression of simply arranging neighborhood carpools to work and fashioning a wildly lucrative peer-to-peer business model. Specifically, new services like Uber Pool and Lyft Line offer dynamic ridesharing, matching passengers with geographically suitable origins and destinations in realtime. These matches have the potential to unlock great efficiencies, saving crucial vehicle, road, and fuel resources. Indeed, we could imagine a future without vehicle ownership, in which a fleet of autonomous taxis adequately and efficiently serve the needs of an entire metropolis. In this thesis, armed with the rich New York City Taxi & Limousine Commission trip record data as well as a comprehensive street-level map network, I take a simulation-based approach to evaluating the scale of this potential. I first outline the motivations and specifications of two simulators|one meant to emulate current taxi behavior, and the other designed to implement a dynamic ride-sharing scheme. I then examine the temporal and spatial distributions of the sampled records used in the study to gain insight into the patterns they exhibit. Finally, I carry out the simulation using the trip records to ultimately quantify the ride-share potential of the system. This study confirms the existing potential for ride-sharing in the New York City Area, with ride-sharing simulations exhibiting a significantly higher average vehicle occupancy and requiring a lower fleet size than "direct" or non-ride-sharing schemes.