Calibration and Validation of a Volume-Delay Function Using a Genetic Algorithm

Abstract

In microscopic traffic networks, searching and optimization algorithms are commonly used to calibrate behavioral and performance models of simulators, such as genetic algorithms (GAs). GA applications in macroscopic simulations have not been reported. The aim of this study was to develop a GA to calibrate and validate a volume-delay function (Akçelik) to macroscopically simulate a Brazilian city, Belo Horizonte, in AIMSUN software. Traffic data were used to estimate the capacity, free-flow speed, and parameter J of the Akçelik function for arterial segments. For other segment types (collectors, locals, and expressways), these data were obtained by the GA. The algorithm application provided higher values of the GA’s fitness function (that compares simulated and observed traffic volumes) than those obtained with the parameters suggested in the literature.