Department of Economics and Business Economics

TY - JOUR

T1 - Estimation of the arrival time of deliveries by occasional drivers in a crowd-shipping setting

AU - Zehtabian, Shohre

AU - Larsen, Christian

AU - Wøhlk, Sanne

PY - 2022/12/1

Y1 - 2022/12/1

N2 - The success of e-commerce business offering same-day delivery depends on customer satisfaction. To speed up deliveries and lower costs, some companies have been using private individuals as non-dedicated drivers to perform pickup and delivery tasks for online customers. Such delivery systems are known as crowd-shipping. Customers have come to expect an accurate estimate for the delivery times of their online orders. The coordination of online deliveries with private individuals is done by a crowd-shipping platform. In this paper, we focus on the estimation of pickup and delivery times. This is a challenging job because not only are the requests unknown and submitted dynamically, but so is the pool of drivers, i.e. delivery capacity. We model the problem as a Markov decision process and integrate it into a simulation study. To improve the estimates that can be done by a naive policy, we propose two policies that use lookahead: one with a fixed lookahead horizon and one with a dynamic. Our numerical experiments demonstrate that a lookahead policy with dynamically adjusted horizon outperforms the other two policies in terms of estimation accuracy, which is up to 19% higher in some instances.

AB - The success of e-commerce business offering same-day delivery depends on customer satisfaction. To speed up deliveries and lower costs, some companies have been using private individuals as non-dedicated drivers to perform pickup and delivery tasks for online customers. Such delivery systems are known as crowd-shipping. Customers have come to expect an accurate estimate for the delivery times of their online orders. The coordination of online deliveries with private individuals is done by a crowd-shipping platform. In this paper, we focus on the estimation of pickup and delivery times. This is a challenging job because not only are the requests unknown and submitted dynamically, but so is the pool of drivers, i.e. delivery capacity. We model the problem as a Markov decision process and integrate it into a simulation study. To improve the estimates that can be done by a naive policy, we propose two policies that use lookahead: one with a fixed lookahead horizon and one with a dynamic. Our numerical experiments demonstrate that a lookahead policy with dynamically adjusted horizon outperforms the other two policies in terms of estimation accuracy, which is up to 19% higher in some instances.

KW - Transportation

KW - Crowd-shipping

KW - Lookahead policy

KW - Simulation

U2 - 10.1016/j.ejor.2022.02.050

DO - 10.1016/j.ejor.2022.02.050

M3 - Journal article

VL - 303

SP - 616

EP - 632

JO - European Journal of Operational Research

JF - European Journal of Operational Research

SN - 0377-2217

IS - 2

ER -