Comparative assessment of fairness in on-demand fleet management algorithms
Résumé
On-demand mobility systems in which a fleet of shared vehicles are increasingly tested and deployed. Their efficiency gains are partly due to central algorithms that control the movements and actions of vehicles and drivers. Existing assessments of the performance of such algorithms in large-scale simulation environments assume homogeneous users and vastly ignore special needs of vulnerable users. In this paper, we perform an assessment of two frequently used fleet management algorithms and compare their behaviour when working with heterogeneous customer demand. We show that requests for which higher interaction times at pick-up are anticipated are rejected with higher probability, propose measures to increase the fairness of these algorithms, and propose pathways for future research.
Domaines
Modélisation et simulationOrigine | Fichiers produits par l'(les) auteur(s) |
---|