Robo-Taxi Fleet Coordination with Accelerated High-Capacity Ridepooling

TL;DR

This paper introduces two acceleration algorithms for large-scale high-capacity ridepooling in robo-taxi fleets, demonstrating improved real-time performance using real-world data from Manhattan, NYC.

Contribution

It proposes novel acceleration algorithms for high-capacity ridepooling, enhancing the efficiency of centralized robo-taxi fleet control.

Findings

Improved real-time algorithm performance demonstrated

Effective control for high-capacity ridepooling achieved

Validated with real-world NYC data

Abstract

Rapid urbanization has led to a surge of customizable mobility demand in urban areas, which makes on-demand services increasingly popular. On-demand services are flexible while reducing the need for private cars, thus mitigating congestion and parking issues in limited urban space. While the coordination of high-capacity ridepooling on-demand service requires effective control to ensure efficiency, the emergence of the paradigm of robo-taxi opens the opportunity for centralized fleet control for an improved service quality. In this work, we propose two acceleration algorithms for the most advanced large-scale high-capacity algorithm proposed in [1]. We prove the improvement in the real-time performance of the algorithm by using real-world on-demand data from Manhattan, NYC.