An information theoretic approach to interacting multiple model estimation

We present an information theoretic approach to develop an interacting multiple model (IMM) estimator. In the mixing and output steps of the proposed estimator, the weighted Kullback-Leibler (KL) divergence is used to derive the fusion of conditional probability density functions. A lower bound and an upper bound are derived for the error covariance of controllable and observable Markov jump linear systems. Simulation results are provided to verify the effectiveness of the proposed estimator.