Trip sequencing algorithm development for centralized, prescheduled taxi systems

Abstract

This dissertation presents a reservation-based model designed to improve the taxi system performance in New York City. By using offline parallel machine sequencing, the research aims to enhance the allocation of ride requests, reducing the number of taxis required while maintaining a high level of service. Using data from the New York City Taxi and Limousine Commission, this research addresses operational challenges such as deadheading and balancing the supply of available taxis to match demand. ☐ The key contributions include the development of a novel approach to the taxi assignment problem, using predictive offline models rather than traditional real-time algorithms, significantly lowering computational demands. Two allocation strategies—time-based and spatial-based splits—were evaluated experimentally to assess their impact on fleet management. The time-based split consistently showed better results in terms of minimizing the fleet size compared to spatial-based allocation and existing conditions. ☐ The research also employs a parallel processing strategy, which further enhances the taxi assignment process by minimizing unnecessary cross-zone travel and increasing fleet utilization. The results indicate that the proposed model is a scalable and effective solution for urban taxi dispatch, providing practical insights for improving fleet operations in high-density areas like New York City.