Abstract
In last-mile parcel delivery, consolidation refers to the process of grouping parcels with common destinations, from multiple suppliers, together for transportation. In recent times urban areas have seen a large increase in supplier delivery vans on their streets, due to the expansion of e-commerce, resulting in increased congestion and air pollution. Consolidated last-mile delivery aims to reduce supplier vehicle movements in urban areas by taking responsibility of last-mile delivery starting from consolidation centres. This approach allows for the use of smaller more efficient delivery vehicles, such as electric-cargo bikes (e-cargo bikes), and a better coordinated overall delivery process. Consolidation centre location selection and consolidated last-mile delivery routing can be modelled as a two-echelon location routing problem (2E-LRP). This work proposes a novel metaheuristic-machine learning algorithm for solving 2E-LRPs. An accurate neural network architecture is introduced for predicting first-echelon routing costs to satellite locations, which is incorporated in the cost evaluation function of a second-echelon multi-depot multi-modal vehicle routing metaheuristic, allowing 2E-LRPs to be solved efficiently. Our solution approach is also expressed as a general framework for solving integrated combinatorial optimisation problems with relatively low time complexity. The generality of the 2E-LRP is exploited to model a variety of consolidated last-mile delivery strategies, including consolidation centre-based delivery and mobile-hub-based consolidated delivery. This work identifies social cost modelling for last-mile vehicle choice as a research gap and proposes a sensitivity analysis methodology to identify social cost thresholds that transport authorities should be aware of when recommending vehicle types for consolidated delivery initiatives. Our experiment results demonstrate the high accuracy and explainability of the proposed neural network architecture. Further results reveal threshold minimum demand levels for consolidated delivery to be cost effective. In high demand scenarios a multi-modal fleet composed of mostly e-cargo bikes and electric-vans (e-vans) can be most cost effective. For high demand scenarios of 500 deliveries or more the inclusion of e-cargo bikes in last-mile delivery fleets remains attractive even if their daily fixed costs (tax, maintenance and insurance) are as high as those of e-vans. Mobile hub-based consolidation avoids the need to travel back to a specific start depot, reducing transport costs, and avoids rent costs of consolidation centres. For a real-world transport network with 1000 customers, we demonstrate potential cost savings of over £100,000 and CO2 emission reductions of over 50,000 kg annually compared to business-as-usual operations.
| Original language | English |
|---|---|
| Article number | 112374 |
| Number of pages | 26 |
| Journal | Computers and Industrial Engineering |
| Volume | 222 |
| Early online date | 20 Sept 2026 |
| DOIs | |
| Publication status | Early online - 20 Sept 2026 |
UN SDGs
This output contributes to the following UN Sustainable Development Goals (SDGs)
-
SDG 11 Sustainable Cities and Communities
-
SDG 13 Climate Action
Keywords
- Consolidation strategies
- Last-mile delivery
- Machine learning
- Metaheuristics
- Optimisation
- Two-echelon location routing
Fingerprint
Dive into the research topics of 'Metaheuristic-machine learning algorithm for evaluating the sustainability impacts of consolidated last-mile delivery'. Together they form a unique fingerprint.Cite this
- APA
- Author
- BIBTEX
- Harvard
- Standard
- RIS
- Vancouver