Abstract:

First, two approximate methods for calculating overall priori information of local nodes with interacting multiple model filter are developed. Secondly, to obtain global priori information of fusion center, global posteriori model probabilities are got according to the evidence combination rule of DempsterShaffer evidence theory, and based on this, the concept of the global equivalent target model is proposed. Based on above results, it is possible to apply a hierarchical memory fusion algorithm to make fusion of multiple interacting multiple model estimations. To improve estimation performance of local nodes by using the fusion results， a feedback mechanism that transports global posteriori model probability back to local nodes is put forward. Simulation results demonstrate that hierarchical memory fusion algorithm of interacting multiple model estimation is valid, and the developed feedback mechanism can actually improve the estimation performance of local nodes.