Nonlinear sequential fusion on direct filtering and information filtering based on the CKF

Nonlinear filtering is a lasting and challenging topic in information fusion filed all along. Due to the complexity of nonlinear systems, it makes that data fusion based on nonlinear filtering distinctly differs from the traditional linear data fusion. For the linear Kalman fusion, there are some important properties, for example, equivalence of the Kalman filter and its information filtering form and equivalence between sequential fusion algorithms based on the direct Kalman filtering and the information filtering. However, we cannot directly and subjectively ensure that all of conclusions or properties in linear fusion can be followed for nonlinear systems because performance of nonlinear fusion depends closely on the concrete nonlinear filters. For these, we study two equivalent problems on estimation fusion accuracies. The first is estimation equivalence of the direct cubature Kalman filter and its associated information filter and the second is to theoretically prove fusion accuracy equivalence of two multisensor sequential fusion methods based on the direct cubature Kalman filter and the information filter. The results on theoretical proof and simulation examples show explicitly and clearly that the two equivalence properties existed in the current linear Kalman estimation fusion can be completely succeeded in the cubature Kalman estimation fusion for nonlinear systems.