System prediction combining state estimation with an evidential influence diagram

摘要

In this paper we develop a system state estimation model for combining partial information regarding the state of a system of interest. In addition we develop an evidential influence diagram representing our a priori knowledge of system relations. Both system state estimation and a priori knowledge are represented by belief functions. A predicted future system state is obtained by combining the fused estimated system state with the fused a priori knowledge. The predicted system state can be marginalized to give specific state predictions of all variables of interest of the system state estimation model. Finally, we may compare predicted system states with later actual states to highlight any deviations from expected developments.