Citation

Li Y, Zhou C, Yuan P, Ngo TTA. Transp. Res. E Logist. Transp. Rev. 2023; 171: e103036.

Copyright

(Copyright © 2023, Elsevier Publishing)

DOI

10.1016/j.tre.2023.103036

PMID

unavailable

Abstract

The parcel delivery industry has enjoyed rapid growth with the rise of the e-commerce business. To survive the highly competitive market, service providers have introduced various methods to improve the customer experience, for example, providing faster response or wider delivery coverage. One way is to adopt the territory-based delivery system, in which each courier serves a fixed group of customers. In this study, we propose a novel territory design method allowing the territory plan to be adjusted while guaranteeing service consistency. The territory planning problem (TPP) can be formulated as a Markov decision process (MDP), and we develop a two-stage Rolling Horizon (TSRH) method to compute the optimal territory plan. In the first stage, the algorithm assigns certain cells to the drivers based on the predicted demands. In the second stage, the remaining cells are assigned to the drivers with the actual demands while taking driver experience into consideration. The computational studies reveal that the proposed TSRH method is able to resolve the TPP efficiently, and it is robust under different situations. The TSRH method outperforms the classical method with fixed core areas. We also find that the learning potential and learning efficiency can largely impact the optimal territory plan, which eventually leads to significant improvement on driver's performance.

Language: en

Keywords

Driver assignment; Learning effect; Markov decision process; Rolling horizon method; Territory planning