| Name: | Description: | Size: | Format: | |
|---|---|---|---|---|
| 2.9 MB | Adobe PDF |
Authors
Abstract(s)
São inúmeras as aplicações dos modelos de microssimulação de tráfego, desde a otimização de
operações com o objetivo de maximizar a capacidade da infraestrutura existente à procura de
soluções que conduzam à redução da sinistralidade rodoviária.
Porém, a utilização dessas análises requer alto nível de especialização, são dispendiosas e o
processo de obtenção dos parâmetros pode não ser simples, nem replicável para locais
diferentes, uma vez que ainda não há metodologias consolidadas e os parâmetros variam de
população para população.
O modelo de car-following é um dos principais componentes dos modelos de microssimulação.
A calibração dos respetivos parâmetros é essencial para atingir resultados precisos, mas sendo
uma tarefa complexa e dispendiosa, os analistas recorrem frequentemente a valores pré-
definidos ou a técnicas de calibração simples que oferecem reduzida transferibilidade.
A presente dissertação teve como objeto a calibração microscópica do modelo de car-following
Intelligent Driver Model (IDM). Para isso, foi realizada uma coleta de dados através de um
veículo instrumentado, que permitiu analisar o comportamento de cinco condutores, alunos e
professores, do Instituto Politécnico de Viseu, Portugal. O procedimento é relativamente
simples e de baixo custo, necessitando apenas de um datalloger e uma pistola LIDAR.
O processo de calibração se deu através de duas metodologias. A primeira por estimativa
individual dos parâmetros, na qual foram separadas seções da coleta de dados onde cada
parâmetro tinha maior influência que os demais, e a segunda, por otimização simultânea dos
parâmetros.
Os distintos processos de calibração executados nesse estudo mostraram desempenhos
satisfatórios, e passiveis de serem replicados. Um desempenho ligeiramente melhor foi notado
na calibração por otimização simultânea.
ABSTRACT: There are numerous applications of traffic microsimulation models, from the optimization of operations with the objective of maximizing the capacity of the existing infrastructure to the search for solutions that lead to the reduction of road accidents. However, the use of these analyzes requires a high level of specialization, they are expensive and the process of obtaining the parameters may not be simple, nor replicable for different locations, since there are still no consolidated methodologies and the parameters vary from population to population. The car-following model is one of the main components of microsimulation models. The calibration of the respective parameters is essential to achieve accurate results, but being a complex and expensive task, analysts often resort to pre-defined values or simple calibration techniques that offer little transferability. The present dissertation had as its object the microscopic calibration of the car-following Intelligent Driver Model (IDM). For this, a data collection was carried out through an instrumented vehicle, which allowed the analysis of the behavior of five drivers, students and teachers, from the Instituto Politécnico de Viseu, Portugal. The procedure is relatively simple and inexpensive, requiring only a datalloger and a LIDAR rangefinder. The calibration process took place through two methodologies. The first by individual parameter estimation, in which sections of the data collection were separated where each parameter had greater influence than the others, and the second, by simultaneous optimization of the parameters. The different calibration processes performed in this study showed satisfactory performance, and likely to be replicated. Slightly better performance was noted in the simultaneous optimization calibration.
ABSTRACT: There are numerous applications of traffic microsimulation models, from the optimization of operations with the objective of maximizing the capacity of the existing infrastructure to the search for solutions that lead to the reduction of road accidents. However, the use of these analyzes requires a high level of specialization, they are expensive and the process of obtaining the parameters may not be simple, nor replicable for different locations, since there are still no consolidated methodologies and the parameters vary from population to population. The car-following model is one of the main components of microsimulation models. The calibration of the respective parameters is essential to achieve accurate results, but being a complex and expensive task, analysts often resort to pre-defined values or simple calibration techniques that offer little transferability. The present dissertation had as its object the microscopic calibration of the car-following Intelligent Driver Model (IDM). For this, a data collection was carried out through an instrumented vehicle, which allowed the analysis of the behavior of five drivers, students and teachers, from the Instituto Politécnico de Viseu, Portugal. The procedure is relatively simple and inexpensive, requiring only a datalloger and a LIDAR rangefinder. The calibration process took place through two methodologies. The first by individual parameter estimation, in which sections of the data collection were separated where each parameter had greater influence than the others, and the second, by simultaneous optimization of the parameters. The different calibration processes performed in this study showed satisfactory performance, and likely to be replicated. Slightly better performance was noted in the simultaneous optimization calibration.
Description
Keywords
Calibração microscópica Intelligent Driver Model Microssimulação de tráfego Veículo instrumentado