An agent-based approach to customer crowd-shipping

Date
2022-04
Journal Title
Journal ISSN
Volume Title
Publisher
Stellenbosch : Stellenbosch University
Abstract
ENGLISH SUMMARY: The challenge of effective last-mile deliveries is progressively becoming more important with the acceleration in the e-commerce industry that is accompanied by a growing number of doorstep deliveries. Crowd logistics provides innovative solutions whereby ordinary people become in- volved in the execution of logistics operations. A particular crowd logistics initiative, referred to as customer crowd-shipping, recently gained interest from researchers after initial implemen- tations thereof had been performed by companies such as Walmart and Amazon. The approach involves the use of a retailer’s in-store customers, in addition to regular delivery vehicles, for delivering orders to online customers. Such in-store customers, referred to as occasional drivers, are offered incentives to deliver orders on their way home after visiting the retailer. In this thesis, an agent-based simulation model is proposed for studying the highly dynamic working of the customer crowd-shipping initiative. The model encompasses a traditional last- mile delivery system, complemented by the ability to utilise autonomous occasional drivers. The modelled traditional last-mile delivery system consists of a dedicated fleet of delivery vehicles serving online customers from a single depot. The execution of deliveries is formulated as a vehicle routing problem and subsequently solved by means of well-known vehicle routing heuristics. In addition, the occasional drivers are modelled as autonomous agents who have the ability to act outside of the control of the retailer. Rather than being assigned to particular orders, occasional drivers are presented with potential orders from which they may select an order suitable for them to deliver. Their decision to participate is modelled based on self- interest, where an occasional driver agent aims to maximise the difference between the incentive offered and his or her perceived value of the additional time required to deliver the order. An integrated approach to customer crowd-shipping is developed in order to consider the benefits for both the retailer and occasional drivers. This includes an incentive scheme and a method for identifying online customers as candidates for crowd-shipping. The latter involves the dynamic calculation of the company’s cost to serve an individual customer, which is determined for all online customers. Finally, user-friendly access to the agent-based simulation model is facilitated by a graphical user interface. The proposed model is subjected to systematic verification, ensuring the correct functioning and integration of its subcomponents. Moreover, the model is evaluated under various operating conditions to gain a deeper understanding of the crowd-shipping initiative, while simultaneously validating the model as adequate. In particular, parameter variation, sensitivity analyses, and scenario analyses are conducted, followed by face validation by subject matter experts. The results of the various analyses indicate that customer crowd-shipping may successfully function as an extension to an existing last-mile delivery system, with the potential of reducing both the total delivery cost and customer waiting time. These benefits are, however, shown to be influenced by the incentive scheme and the strategy by which online customers are se- lected as crowd-shipping candidates. Finally, it is deduced that the maturity of the customer crowd-shipping system and the occasional population’s perceived value of time influence the performance of the customer crowd-shipping model.
AFRIKAANS OPSOMMING: Die uitdaging van doeltreffende laaste-myl aflewerings word geleidelik belangriker met die versnelling in die e-handelsbedryf wat gepaard gaan met ’n groeiende aantal voorstoepaflewerings. Skare-logistiek bied innoverende oplossings waardeur gewone mense betrokke raak by die uitvoering van logistieke bedrywighede. ’n Sekere skare-logistieke inisiatief, waarna verwys word as kli¨ente-skareversending, het onlangs belangstelling by navorsers ontlok nadat aanvanklike implementering daarvan deur maatskappye soos Walmart en Amazon plaasgevind het. Die benadering behels die gebruik van ’n kleinhandelaar se in-winkel kli¨ente, benewens normale afleweringsvoertuie, om bestellings by aanlynkli¨ente af te lewer. Sulke in-winkel kli¨ente, na wie daar ook verwys word as geleentheidsbestuurders, word aansporings gebied om bestellings op pad huis toe af te lewer nadat hulle die kleinhandelaar besoek het. In hierdie tesis word ’n agent-gebaseerde simulasiemodel voorgestel vir die bestudering van die hoogs-dinamiese werking van die kli¨ente-skareversendingsinisiatief. Die model sluit ’n tradisionele laaste-myl afleweringstelsel in, aangevul deur die mootlikheid om outonome geleentheidsbestuurders te gebruik. Die gemodelleerde tradisionele laaste-myl afleweringstelsel bestaan uit ’n toegewyde vloot afleweringsvoertuie wat aanlynkli¨ente vanaf ’n enkele depot bedien. Die uitvoering van aflewerings word as ’n voertuig-roeteringsprobleem geformuleer en vervolgens deur middel van bekende voertuig-roeteringsheuristieke opgelos. Daarbenewens word die geleentheidsbestuurders as outonome agente gemodelleer wat oor die vermo¨e beskik om buite die beheer van die kleinhandelaar op te tree. Eerder as om aan spesifieke bestellings toegewys te word, word geleentheidsbestuurders potensi¨ele bestellings aangebied waaruit hulle een kan kies wat geskik is om deur hulle afgelewer te word. Hul besluit om deel te neem berus op eiebelang, waar ’n geleentheidsbestuurder-agent poog om die verskil tussen die aansporing wat aangebied word en sy of haar waargenome waarde van die bykomende tyd wat benodig word om die bestelling af te lewer, te maksimeer. ’n Ge¨ıntegreerde benadering tot kli¨ente-skareversending word ontwikkel om die voordele vir beide die kleinhandelaar en geleentheidsbestuurders te oorweeg. Dit sluit ’n aansporingskema in sowel as ’n metode om aanlynkli¨ente as kandidate vir skareversending te identifiseer. Laasgenoemde behels die dinamiese berekening van die maatskappy se koste om ’n individuele kli¨ent te bedien, wat vir alle aanlynkli¨ente bepaal word. Laastens word gebruikersvriendelike toegang tot die agent-gebaseerde simulasiemodel deur ’n grafiese gebruikerskoppelvlak moontlik gemaak. Die voorgestelde model word aan sistematiese verifikasie onderwerp, wat die korrekte funksionering en integrasie van die deelkomponente daarvan verseker. Boonop word die model onder verskeie bedryfstoestande ge¨evalueer om ’n dieper begrip van die kli¨ente-skareversendingsinisiatief te verkry, terwyl die model terselfdertyd as voldoende bekragtig word. In die besonder word parametervariasie, sensitiwiteitsanalises en scenario-ontledings uitgevoer, gevolg deur sigvalidering deur vakkundiges. Die resultate van die verskillende ontledings dui daarop dat kli¨ente-skareversending suksesvol as ’n uitbreiding van ’n bestaande laaste-myl afleweringstelsel kan funksioneer, met die potensiaal om beide die totale afleweringskoste en kli¨entewagtyd te verminder. Daar word egter getoon dat hierdie voordele be¨ınvloed word deur die aansporingskema en die strategie waardeur aanlynkli ¨ente as skareversendingkandidate gekies word. Laastens word afgelei dat die volwassenheid van die kli¨ent-skareversendingstelsel en die bevolking geleentheidsbestuurders se waargenome waarde van tyd die prestasie van die kli¨ente-skareversendingsmodel be¨ınvloed.
Description
Thesis (MEng)--Stellenbosch University, 2022.
Keywords
Crowdsourcing, Microeconomics, Collaborative economy, Vehicle routing problems, Agent-based modelling, UCTD
Citation