Hierarchical Decentralized State Estimation with Unknown Correlation for Multiple and Partially Overlapping State Vectors

Abstract

A fusion method for hierarchical decentralized state estimation is investigated with unknown cross-correlation. The process is divided into several local estimators for different subsystems that can operate independently from each other. Each of the local estimators estimates only part of the entire state vector, where some of the states overlap with each other. Using the proposed fusion method, the local state estimates are fused together, to reconstruct the global state vector and to get an improved state estimate. A comparison with other existing fusion methods was made through simulations and showed a reduction in estimation error when using the proposed method.