EURO 2024 Copenhagen
Abstract Submission

EURO-Online login

1545. An Approximate Dynamic Programming Approach for a Crowdshipping Platform

Invited abstract in session TB-56: Crowdsourcing Logistics, stream Transportation.

Tuesday, 10:30-12:00
Room: S04 (building: 101)

Authors (first author is the speaker)

1. Emma Innocente
Operations and Information Systems (OIS), Université Catholique de Louvain
2. Jean-Sébastien Tancrez
CORE - Louvain School of Management, Université catholique de Louvain

Abstract

Spurred by the growth of online shopping, parcel delivery has grown rapidly, driving a need for cost-effective last-mile delivery. Amidst this growth, sustainability concerns are gaining prominence. Crowdshipping leverages crowdsourcing for the personalized delivery of freight, turning ordinary citizens into couriers for the distribution of small items. In this collaborative delivery system, individuals already traveling from an origin to a destination take charge of all or part of the delivery, taking a package along with them and making a stop along the way to drop it off, potentially reducing freight trucks and enhancing sustainability. To fully leverage the potential of crowdshipping, real-time matching of crowdshippers to parcels is crucial, considering spatial and temporal uncertainties in crowdshippers’ and parcels’ availability. Due to its size, the dynamic assignment problem considered in our article cannot be solved with conventional dynamic programming methods. Using approximate dynamic programming with value function approximation, our algorithm learns value functions offline through a training horizon, enabling efficient decision-making. This adaptive learning algorithm provides nonmyopic behavior yet requires only solving sequences of assignment problems no larger than would be required with a myopic algorithm. Initial results demonstrate its superiority over myopic solutions in reducing delivery distances.

Keywords

Status: accepted


Back to the list of papers