Comparison of Fusions Under Unknown and Partially Known Correlations

Abstract

Abstract Approaches to decentralised estimation seek balance between proximity to the optimal centralised solution and complexity of the designed solution. Fusion of two estimates is a fundamental technique. Since unknown cross-correlation of estimation errors induces a loose evaluation of the fused estimate quality, a partial knowledge of correlations is beneficial. This paper inspects a recently introduced type of an implicit partial knowledge. The implicit knowledge is compared with other types of the knowledge in terms of families of upper bounds of mean square error (MSE) matrices. Next, an existence of a non-trivial lower bound of MSE matrices is discovered. Finally, the analytical results are illustrated graphically.