Hybrid Recurrent Traffic Flow Model with Artificial Neural Networks For Approximation of Unknown Urban Road Subnetworks

摘要

The article considers a hybrid recurrent mathematical model of traffic flow control, that consists of two types of models: a universal recurrent finite-difference model for describing subnetworks with known parameters, and a model based on the artificial neural networks to describe subnetworks with partially unknown parameters. The choice of Elman recurrent neural network is substantiated and a method for its training based on a variational genetic algorithm is described.