Centralized fusion algorithms based on EKF for multisensor non-linear systems

Aiming at a kind of nonlinear multisensor systems, we study three classic nonlinear centralized fusion algorithms based on the extend Kalman filter (EKF) and extend some fusion theories for linear dynamic systems to nonlinear systems. On the basis of extend information filter (EIF), three kinds of fusion algorithms such as augmented measurements fusion, measurements weighted fusion and sequential filtering fusion are presented. Afterwards, we compare estimate accuracies of the three nonlinear fusion algorithms, and discuss the exchanging property of measurement's update order. The results are as follows. Firstly, when measurement properties are identical, the estimate of the augmented measurements fusion algorithm and the measurements weighted fusion algorithm are equivalcent. Secondly, the estimate accuracy of the sequential filtering fusion does not hold completely functional equivalence in linear systems as the other two fusion methods. Thirdly, the exchanging property of the measurement's update order of nonlinear sequential filtering fusion can no longer be guaranteed. Four examples based on bearings-only tracking are shown to demonstrat the validity of the conclusions.