Analysis on adaptive filtering for nonlinear dynamic estimation systems

It is very difficult to get absolutely accurate estimation models for practical filtering and target tracking systems. Accordingly, adaptive filtering technology has been presented to deal with state estimation with inaccurate models and some typical adaptive methods have been established. These adaptive methods have also used to design various filtering or fusion algorithms most of which have applied in many engineering systems. Unfortunately, the current studies are short of deeply analyzing the principles and computing complexity of these adaptive filtering methods, which are exceedingly important to improve application levels of the adaptive algorithms. Aiming at the problem mentioned above, for a kind of nonlinear state estimation system with unknown or inaccurate covariances of additive noises, we introduce three kinds of adaptive filtering methods based on unscented Kalman information filter (UKF) and deeply discuss their principles and computing complexity in this paper. The study can help researchers to thoroughly understand the adaptive filtering theory and better develop further research in theory and engineering applications.