Suboptimal Kalman filtering for linear systems with Gaussian-sum type of noise

This paper develops several suboptimal filtering algorithms for discrete-time linear systems that have state and/or measurement noise of the Gaussian-sum type. These new computational schemes are modifications and generalizations of the well-known algorithms of Sorenson and Alspach and of Masreliez. Under the common minimum mean square estimation criterion, these new schemes are derived as recursive computational algorithms. Monte Carlo simulations have shown that these new filtering algorithms significantly improve the computational efficiency and/or filtering performance of the existing algorithms.