A general framework for decentralized estimation

Work by various authors has established decentralized theory for Kalman filters and H_∞ Filters. In this paper, the authors extend the existing theory to the entire class of Luenberger observers and give a necessary and sufficient condition for decentralized estimation. It is shown that decentralized estimators work by combining local estimates through weighting matrices. In special cases, this creates a global estimate which looks much like a weighted average. The authors conclude with a proof of a necessary and sufficient condition proposed earlier by Willsky et al. (1982)