Multiple model tracking with intermittent mode observations

This paper is focused on the estimation of both mode and state of a hybrid system when intermittent mode observations are available. These intermittent data are modeled by a marked point process. The exact hybrid filter based on this formalism is elaborated by use of the reference probability method. A recursive form of the exact conditional density is given in the general case as well as in the initial Gaussian case. A suboptimal and finite dimensional filter extracted from the previous form is used to design a radar plus imaging sensor maneuvering target tracker. Simulations show the efficiency of this new tracker