Improved fittings for distribution of vehicles' velocity in intelligent transportation systems

To find out the approximations for the probability distribution functions (PDF) of vehicles' speeds during different periods of time respectively is one of the most important targets to analyze a transportation system. Nevertheless, the distribution of the velocity generally does not consist with the normal (Gaussian) distribution. In this article, two statistic models for fitting are investigated, which are Generalized T (GT) distribution and the Gaussian Mixture Model (GMM). Maximum likelihood Estimation (MLE) rule is used in the optimization of the models' parameters through Genetic Algorithm. For simulation, data from NGSIM is utilized, and the result manifests that both of the two proposed models are better than Gaussian model. GT performs better in maximum likelihood, while GMM has better performance in MSE.