Railway traffic management: simulation and heuristic optimization
The railroad operations often require planning on the routing of the trains in order to comply with physical restrictions (like single-track operations) while handling priorities on crossings and overtakes, among others. In order to facilitate the route design, some auxiliary tools were made. This p...
| Autor principal: | Silva, Fernando Augusto Constantino da |
|---|---|
| Formato: | Dissertação |
| Idioma: | Inglês |
| Publicado em: |
Universidade Tecnológica Federal do Paraná
2021
|
| Assuntos: | |
| Acesso em linha: |
http://repositorio.utfpr.edu.br/jspui/handle/1/25058 |
| Tags: |
Adicionar Tag
Sem tags, seja o primeiro a adicionar uma tag!
|
| Resumo: |
The railroad operations often require planning on the routing of the trains in order to comply with physical restrictions (like single-track operations) while handling priorities on crossings and overtakes, among others. In order to facilitate the route design, some auxiliary tools were made. This project aims to create an open-source simulation tool for railroad routing and perform an optimization based on two bio-inspired metaheuristics (Genetic Algorithm - GA and Particle Swarm Optimization - PSO) and another randomaction controller (RND). A literature review about the historical context of railroads over the world and mainly in Brazil is made, from where the routes used for comparisons are based. The controllers’ results are later compared over the best solution cost, the total number of successful solutions, total execution time, and the cost evolution per epoch. A Wilcoxon signed-rank test is executed for each possible pair of controllers in order to determine the statistical difference of the resulting data-sets. The obtained results suggests that the RND controller performs better in the evaluated scenarios, having the faster execution time on both scenarios and achieving the best global solution cost in the harder one. The tool also outputs a video presenting the synoptic panel with the entire execution of the simulation, allowing an easy audition and debugging of the solution. It was built using Python in a Docker container so it can run under different platforms and architectures, being hosted on GitHub and available for further public contributions after the registry process. |
|---|