This repository contains the source code for my undergraduate thesis, which was presented to receive the degree of Bachelor of Civil Engineering at the Escola Politécnica da Universidade de São Paulo in December, 2022.
The algorithmic part of this work is based on the lmr_analyzer Python module, which is also available on another repository. Changes to the lmr_analyzer are allowed at that repository, but not here. This current repository is only for archiving the thesis itself, therefore it will not be updated.
The thesis was written in Portuguese, so it is only available in that language. However, questions about the work can be asked in English.
You can easily navigate through the repository by clicking on the links below:
- Title: Impacto da geometria de malhas viárias sobre o desempenho da distribuição de entregas de última milha
- Authors: Felipe Novaes Fernandes and Guilherme Fernandes Alves
- Supervisor: Claudio Barbieri da Cunha
- Institution: Escola Politécnica da Universidade de São Paulo
- Department: Transportation Engineering
- Date of approval: 16th December, 2022
If you use this work in your research, please cite it as follows:
@booklet{fernandes2022impacto,
title = {Impacto da geometria de malhas viárias sobre o desempenho da distribuição de entregas de última milha},
author = {Alves, Guilherme Fernandes and Fernandes, Felipe Novaes},
howpublished = "Final paper for the degree of Bachelor of Civil Engineering at the Escola Politécnica da Universidade de São Paulo",
month = dec,
year = 2022,
}The last mile distribution plays a key role in supply chain management nowadays. Characterized by covering the last stage of the large-scale delivery process, the last mile segment presents large dynamics on a daily basis, with a high impact on the overall performance of deliveries. Therefore, in order to either improve or optimize last mile deliveries, it is crucial to have a good tactical planning of your operation. This paper has elaborated on top of the phenomena known as ``non-adherence between planned and executed routes'' on last mile distribution.
The non-adherence can be evaluated based on three key performance indicators: Failed delivery attempts, Sequential Non-Adherence and Rejections. These phenomena represent unexpected behaviors in the delivery route and imply in several economic and environmental impacts. With that in mind, to provide a greater control of last mile operations, it is important understand which indicators may influence the occurrence of the non-adherence. To accomplish with that, this paper presents the first steps towards identifying the correlation between the indicators and different potential causality factors Therefore, two different case studies with actual data on last mile deliveries were performed. One of the studies is related to a large Brazilian beverage company with a database of deliveries in the São Paulo Metropolitan Region (Brazil), while the other is about the delivery database of Amazon.com in the United States. Our contributions has allowed for an extensive set of correlation analyzes that address measures of characterizing deliveries routes, spatial autocorrelation, the relative positioning of drop-off points, the circuity factor and a set of variables for measuring different aspects of the street network topology.
Furthermore, the results suggests the existence of correlated effects between elements related to deliveries, such as the time of delivery and the delivery team that performs it, as well as characteristics associated with the street network. After series of both simple and multi-variable regressions, it was possible to establish that three indicators associated with the size of the route, the circuity factor and the orientation may directly affect 18% of the occurrence of failed delivery attempts. These contributions may be useful to different study areas, including education, health, traffic management. The major part of this work has been based on top of open-source tools provided by previous works and also by the authors. For instance, the Python module lmr_analyzer, created during this work, is now totally available on open-source platforms, allowing for further investigations under different contexts, as well as the reproduction of this work's results.
Keywords : Logistics, Last-mile distribution, Vehicle routing problem, Street Network Analysis, Circuity.
The repository is organized as follows:
├── build
│ ├── build_vX.pdf # Compiled PDF file, where X is the version number
├── source # Source code for the thesis
│ ├── images
│ │ ├── ... # Images used in the thesis
│ ├── 0_resumo.tex
│ ├── 1_introducao.tex
│ ├── 2_revisao.tex
│ ├── 3_caracterizacao.tex
│ ├── 4_materiais_e_metodos.tex
│ ├── 4-2_lmr_analyzer.tex
│ ├── 5_empresa_bebidas.tex
│ ├── 6_amazon.tex
│ ├── 7_conclusoes.tex
│ ├── 8_appendix.tex
│ ├── main.tex # Main file for the thesis
│ ├── main.bib # Bibliography file
├── LICENSE.md
├── README.md
To easily read the thesis, you can download the latest version of the PDF file from the build directory. Unfortunately, the thesis was written in Portuguese, so it is only available in that language.
To compile the thesis, you will need to have a LaTeX distribution installed on your machine. Then, you can compile the main.tex file using your favorite LaTeX editor. Images were referenced using relative paths, so you will need to compile the file from the source directory.
The only contribution that we accept is bug reports. No new features will be added to this project. Pull requests will be majorly rejected, and, when the repository is archived, no changes will be allowed at all (archiving repositories).
If you have any questions, please contact at least one of the following by e-mail:
- Guilherme Fernandes Alves:
gf10.alves@gmail.com - Prof. Dr. Claudio Barbieri da Cunha:
cbcunha@usp.br