Off-line identification of concurrent Discrete Event Systems exhibiting cyclic behaviour

This paper presents a method for the identification of concurrent discrete event systems (DES) from input-output sequences representing the observed behavior. The proposed off-line technique yields an input-output model expressed as an interpreted Petri net (IPN), which represents exactly the language than that generated by the observed system, which may include cyclic sequences. First, a sample of input-output vectors words are processed for obtaining sequences of output changes called events; then sequences of K-length event traces are built and represented by an IPN model composed by non measurable places. Then measurable places are added; they are related to transitions representing pertinent output changes. Finally inputs are associated to transitions and implicit non measurable places are removed.