Large-Scale Dynamic Ridesharing with Iterative Assignment

Abstract: Transportation network companies (TNCs) have become a highly utilized transportation mode over the past years. At their emergence, TNCs were serving ride requests one by one. However, the economic and environmental benefits of ridesharing encourages them to dynamically pool multiple ride requests to enable people to share vehicles. In a dynamic ridesharing (DRS) system, a fleet operator seeks to minimize the overall travel cost while a rider desires to experience a faster (and cheaper) service. While the DRS may provide relatively cheaper trips by pooling requests, the service speed is contingent on the objective of the vehicle-to-rider assignments. Moreover, the operator must quickly assign a vehicle to requests to prevent customer loss. In this study we develop an iterative assignment (IA) algorithm with a balanced objective to conduct assignments quickly. A greedy algorithm from the literature is also tailored to further reduce the computational time. The IA was used to measure the impact on service quality of fleet size; assignment frequency; the weight control parameter of the two objectives on vehicle occupancy -- rider wait time and vehicle hours traveled. A case study in Austin, TX, reveals that the key performance metrics are the most sensitive to the weight parameter in the objective function.