Abstract
One of the fundamental issues in dynamic vanpool services is to schedule a fleet of vans to serve passengers efficiently, which is usually modeled as the dynamic and stochastic dial-a-ride problem (DARP) in recent studies. However, these studies do not consider prepositioning when dispatching the vans. Prepositioning means positioning the vans in advance to satisfy potential future requests, instead of only considering sending the vans to requests received. This study developed a metaheuristic scheduling algorithm for the dynamic and stochastic DARP. The algorithm uses multiple scenarios which include future requests and traffic conditions to generate and evaluate potential decisions considering prepositioning. The study uses a real dataset, which includes requests from a vanpool services provider and traffic conditions achieved from an online map service, to test the algorithm. The results show that incorporating stochastic requests without considering prepositioning can improve the average profit by 18.6% and that prepositioning can improve average profit by 23.8% and reduce average waiting time by 74.7%. The influence of different algorithm parameters is tested to provide guidance for the practical usage of the proposed algorithm.
Get full access to this article
View all access options for this article.