A state estimation method for multiple model systems using belief function theory

Multiple model methods have been generally considered as the mainstream approach for estimating the state of dynamic systems under motion model uncertainty. In this paper, a multiple model method based on belief function theory is proposed. This method handles the case of systems with an unknown and variant motion model. First, a set of candidate models is selected and an associated Dempster-Shafer mass function is computed based on the measurement likelihood of possible motion models. The estimated state of the system is then derived by computing the expectation with respect to the pignistic probability. In order to validate our work, we applied the proposed method to a vehicle localization problem. The comparison with other methods demonstrates the effectiveness of the proposed method.