Observers for discrete-event manufacturing systems

Abstract

The author describes the design of observers for a class of discrete-event dynamic systems where events can become unobservable in real time. He defines a novel temporal model that augments a finite state machine to describe both the temporal and behavioral attributes of the system. The behavioral model is analyzed to generate ambiguity trees that contain unobservable states due to certain events becoming unobservable in real time. The depth of an ambiguity tree is a measure of the degree of ambiguity in a normally observable system due to the loss of an event. The author describes an observer that changes the temporal attributes of the model in real time to identify ambiguous states.< >