Research and Development of an Intelligent System For Rapid Train Schedule Adjustment Based on Step-by-Step Neural Control

This paper describes the realising and analysis of the applicability of artificial intelligence technology and game theory to resolve conflict situations that occur during railway traffic. In this technology, decision-making on the train traffic control on the railway section is based on the using of an intelligent management approach. The proposed approach includes a model of dispatcher operating in conflict situations and a model of railway traffic. The dispatcher operating model is implemented using a fully connected artificial neural network with several hidden layers. The neural network is trained using a genetic algorithm. The main idea of this approach is to approximate the problem of train traffic control to one of the problems of game theory, which involves a single player-dispatcher interacting with a group of agents-trains in a dynamic environment.