Studies on multi-model Kalman filters for stochastic systems

This paper presents a multiple model Kalman filter method for stochastic systems. The method utilizes a group of linear models, for instance N models, to represent a nonlinear stochastic system. Kalman filters associated with the group of models are used so that an N estimated states are obtained. Then the N state-estimates, each of which is weighted by its possibility that is also calculated online, are combined to form an optimal estimate. Simulation tests were conducted for both a second-order system and a ship-control system, and it is demonstrated that the method is effective for systems with sharply changing parameters. Simulation results also demonstrate that the method is applicable to a number of engineering problems for state estimate and parameter identification for nonlinear stochastic systems