Choice-based crowdshipping: A dynamic task display problem

This paper studies the integration of the crowd workforce into a generic last-mile delivery setting in which a set of known delivery requests should be fulfilled at a minimum cost. In this setting, the crowd drivers are able to choose to perform a parcel delivery among the available and displayed re...

Full description

Saved in:
Bibliographic Details
Main Authors: ARSLAN, Alp, KILCI, Firat, CHENG, Shih-Fen, MISRA, Archan
Format: text
Language:English
Published: Institutional Knowledge at Singapore Management University 2022
Subjects:
Online Access:https://ink.library.smu.edu.sg/sis_research/7356
https://ink.library.smu.edu.sg/context/sis_research/article/8359/viewcontent/Choice_basedCrowdshipping_wp_202209.pdf
Tags: Add Tag
No Tags, Be the first to tag this record!
Institution: Singapore Management University
Language: English
Description
Summary:This paper studies the integration of the crowd workforce into a generic last-mile delivery setting in which a set of known delivery requests should be fulfilled at a minimum cost. In this setting, the crowd drivers are able to choose to perform a parcel delivery among the available and displayed requests. We specifically investigate the question: what tasks should be displayed to an individual driver, so as to minimize the overall delivery expenses? In contrast to past approaches, where drivers are either (a) given the choice of a single task chosen so as to optimize the platform’s profit, or (b) allowed full autonomy in choosing from the entire set of available tasks. We propose a dynamic, customized display model, where the platform intelligently limits each driver's choice to only a subset of the available tasks. We formulate this problem as a finite-horizon Sequential Decision Problem, which captures (a) the individual driver’s utility-driven task choice preferences, (b) the platform’s total task fulfilment cost, consisting of both the payouts to the crowd-drivers as well as additional payouts to deliver the residual tasks. We devise a stochastic look-ahead strategy that tackles the curse dimensionality issues arising in action and state spaces and a non-linear (problem specifically concave) boundary condition. We demonstrate how this customized display model effectively balances the twin objectives of platform efficiency and driver autonomy. In particular, using computational experiments of representative situations, we exhibit that the dynamic and customize display strategy significantly reduces the platform’s total task fulfilment cost.