Simulation of stochastic elements in railway systems using self learning processes.

The railway traffic follows deterministic rules, whose selection and application depend on the choices of human operators. These choices may be different in similar situations and produce different effects on the circulation. The difficulty to code, in a general and comprehensive way, these behaviours suggested to test the use of systems capable to reproduce events without requiring a previous definition of the operating rules but acting by means of self-learning processes. The present research deals with: an analysis of the critical behavioural parameters, difficult to be effectively modelled by means of analytical simulation tools; the selection of the self-learning process for the application to the reliability of a railway network capable to work as a part of a wider simulation model of railway traffic; the development of a preliminary version of the model simulating the stochastic failure events and its application to a case study.