Universal recurrent traffic flow model as microscopic model

Abstract

The work is devoted to the study of a universal recurrent model of traffic flows and its application for simulation and control in the urban road network. It is shown that this model, based on the theory of controlled networks, is a discrete graph model, that takes into account uncertainties and allows solving the optimal control problems for working phases of traffic lights in the classical statement. The model effectively describes the subnetworks in the presence of data on possible ma-neuvers, their capacity, data on restrictions on road sections, on the distribution of flows between different directions, etc. If the network contains areas with partially unknown parameters, it is pro-posed to approximate such areas using artificial neural networks and use a hybrid model consisting of a universal recurrent model and neural networks. The structure of the hybrid model is described. The hybrid model allows to obtain detailed simulation for the most important sections of the net-work, significantly reduce the amount of data collected for simulation and at the same time cope with the incompleteness of information about the parameters of some subnetworks. As a step to val-idation of the entire hybrid model, as well as to search for other types of hybridization, the article provides a methodological example of calculation for several control steps using a universal recur-rent model. The universal recurrent model is compared to other most popular microscopic traffic flow models, such as cellular automata, following-the-leader, and predicate models. A comparative analysis of the considered models by the type of network, space, and time representation, etc. is pre-sented.