Aplicabilidade de algoritmos genéticos para calibração de redes viárias urbanas microssimuladas

Abstract

Traffic simulation modeling of mid-sized and big cities is a fundamental tool to analyze the performance of strategies idealized to maximize efficiency of urban transport systems. Traffic microsimulation models allow for foreseeing the impacts of different alternatives of solution, such as creation or duplication of lanes, reversion of flow direction, and implementation of priority lanes for public transport. However, the main microsimulation software packages available were developed in other countries, what means that they usually do not suit traffic behavior found in Brazilian urban centers, so parameter calibration is necessary to better adjust the models to local reality. Advances in computational capabilities have allowed traffic microsimulation models incorporate greater amount of parameters to be calibrated, so the process of calibration has become relatively more complex. Therefore, the use of computational methods to optimize the parameters’ values simultaneously is indispensable. This Master’s thesis aims to verify the applicability of genetic algorithms to calibrate microsimulation traffic models in urban Brazilian networks. At the end of this study, it was concluded that the calibration method used proved to be efficient. It is believed that the proposed methodology to calibrate the microsimulation software and the genetic algorithm, as well as the final configuration of the genetic algorithm selected, may be applied to other urban networks, provided that they present similar characteristics to the studied network.