An Identification Technique for Timed Event Systems

Compared to time-driven systems, few tools are available for identification of timed event-driven systems. Here we address the identification problem for those systems that require not only proper ordering, but also correct timing, of input and output events for their complete description. First we review well-known results from computational learning theory, noting their roots in classical systems theory and interpreting their relevance to the identification problem. Next, as a prerequisite to identification, we introduce a simple formalism for timed deterministic event-based system behavior that is also well-adapted to identification. Then we present the main result of the paper, an algorithm that under appropriate conditions computes an automaton whose behavior reliably imitates a given timed event system. Simulation results illustrate best-case and worst-case identification conditions.