System identification of discrete event systems

Abstract

System identification (SI) for discrete event systems (DES) addresses the issue of identifying system dynamics from externally observed sample paths. To that end, we propose a new mathematical framework in the context of a Mealy machine and language theory. Without any a priori modeling information regarding the identification target, the objective of SI is to derive minimal valid automata that duplicate the input‐output relation in the observed sample paths, while ensuring minimal realization. An algorithm to compute minimal valid automata and special properties regarding SI are explained with illustrative examples.

Key words:

• system identification
• discrete event dynamic systems
• minimization for Mealy machines

Notes

Corresponding author. (Tel: [phone]; Fax: [phone]; Email: [email protected])