Analysis of dynamical systems based on information theory
Résumé
Knowledge based control and supervision of unknown or uncertain systems needs the development of learning procedures which can be run on line together with the control, and which allow the identification of the actual rules governing the system evolution, in order to adapt its control, or to detect its deviations from nominal behaviour.
The proposed paper is devoted to on line analysis of dynamical order, for systems whose description variables take on qualitative (or symbolic) values. The proposed approach uses information theory based indexes, which also provide a means to consider systems in which symbolic as well as numerical variables are used for the system description.