Fuzzy local linear models for target tracking
S. McGinnity · 1996
Four new filters for nonlinear estimation have been introduced based on a local linear modelling approach. Simulations on a highly nonlinear target tracking application suggest the performance of the four fuzzy Kalman filters (FKF) to be comparable to the extended Kalman filter, and FKF3 to be consistently better, especially when system noise is added. This highlights the effect of composite modelling, which in general leads to greater modelling accuracy and hence to more accurate tracking. A major advantage of the local modelling based approach is that further linearisation of the model is not required for application of Kalman filtering (except in FKF4). Also, as the approach is model based it lends itself to systems for which analytical equations may not be available.