Decentralized state estimation in discrete event systems under partially ordered observation sequences

Abstract

We consider the problem of decentralized state estimation, where two or more observation sites send information to a coordinator who aims to determine the set of possible current states of a given discrete event system (DES). More specifically, we consider a DES (modeled as a nondeterministic finite automaton) whose underlying activity is partially observed at multiple observation sites that can gather information about different subsets of events. When these sites report to the coordinator the sequences of observations that they have recorded, the goal is to fuse this information and infer the possible current states of the given system. We discuss an efficient recursive algorithm that can perform this task with complexity that is captured by the product of the lengths of the observation sequences reported by the observation sites. © 2016 IEEE.