Calibração de redes viárias urbanas microssimuladas com o uso de algoritmos genéticos

Abstract

The main traffic microsimulation packages require the calibration of various parameters. This article aims to evaluate the applicability of genetic algorithms to the calibration of an urban reticular, high-density network located in the city of Fortaleza, similar to networks found in other Brazilian cities. The effects of the parameters and operators of genetic algorithms in the calibration of 10 parameters of VISSIM microscopic simulator were investigated. The results indicate that variations in the configuration of the genetic algorithms did not affect the calibration results significantly. The simulation with all VISSIM parameters under default values resulted in a MAPE of 29%. After inserting the values of two parameters collected in the field, as well as the inclusion of public transport, the error was reduced to 13%. After calibration of the parameters, the error decreased to 8%. The validation error of the network was 13%.