TY - JOUR
T1 - State fusion with unknown correlation : ellipsoidal intersection
AU - Sijs, J.
AU - Lazar, M.
PY - 2012
Y1 - 2012
N2 - This article focuses on the problem of fusing two prior Gaussian estimates into a single estimate, when the correlation is unknown. Existing solutions either lead to a conservative fusion result, as the chosen parametrization focuses on the fusion formulas instead of correlations, or they are computationally expensive. The contribution of this article is a novel parametrization, in which the correlation is explicitly characterized a priori to deriving the fusion formulas. Then, maximizing the correlation ensures that the fusion result is based on independent parts of the prior estimates and, simultaneously, addresses the fact that the correlation is unknown. In addition, a guaranteed improvement of the accuracy after fusion is attained. An illustrative example demonstrates the benefits of the proposed method compared to an existing fusion method.
AB - This article focuses on the problem of fusing two prior Gaussian estimates into a single estimate, when the correlation is unknown. Existing solutions either lead to a conservative fusion result, as the chosen parametrization focuses on the fusion formulas instead of correlations, or they are computationally expensive. The contribution of this article is a novel parametrization, in which the correlation is explicitly characterized a priori to deriving the fusion formulas. Then, maximizing the correlation ensures that the fusion result is based on independent parts of the prior estimates and, simultaneously, addresses the fact that the correlation is unknown. In addition, a guaranteed improvement of the accuracy after fusion is attained. An illustrative example demonstrates the benefits of the proposed method compared to an existing fusion method.
U2 - 10.1016/j.automatica.2012.05.077
DO - 10.1016/j.automatica.2012.05.077
M3 - Article
SN - 0005-1098
VL - 48
SP - 1874
EP - 1878
JO - Automatica
JF - Automatica
IS - 8
ER -