Decentralized estimators for interconnected systems using the interface information

Abstract

A method for designing decentralized linear state estimators for large-scale systems is proposed. The system under consideration, called the global system, consists of a set of interconnected subsystems called local models. The coupling among the subsystems is described by an algebraic equation. The information about the system consists of noisy measurements taken at the subsystem level, called local measurements. Each local measurement comprises measurement on the local state vector and on the interaction variables. Decentralized estimators that minimize the performance index are sought. Decentralization is obtained from the fact that only local models and information are used in the design. The optimality conditions are derived through the connection structure, and the constraints that these conditions impose on either the connection variable measurements or the interaction equations are developed. The global state estimate of the overall state vector is obtained by stacking local state estimators.<>