Diagnosis of cyclic discrete-event systems using active acquisition of information

Abstract

This paper extends the active acquisition of information approach developed in Thorsley and Teneketzis (2004) from the case of acyclic, timed automata to the more general case of cyclic, asynchronous automata. Conditions for the existence of optimal solutions at finite cost are presented for both logical and stochastic systems. The information state method developed in the previous paper is reduced to a "diagnoser state" method wherein actions are computed for each potential set of states, as opposed to each potential set of strings. After developing a method of finding an optimal policy, a limited lookahead algorithm is presented to produce a suboptimal solution with less intensive computation