Nonlinear Kalman filtering using fuzzy local linear models

A local linear modelling based approach to nonlinear state estimation is introduced. The local models are defined using the Sugeno fuzzy inference framework and constructed using neurofuzzy modelling techniques. Two new fuzzy Kalman filters, which do not require further linearisation nor analytical system equations, are derived. Simulation results presented for a highly nonlinear target tracking problem suggest potential improvements when compared with conventional extended Kalman filtering.