An efficient particle filter for jump Markov nonlinear systems

We present an efficient particle filtering algorithm for state estimation of jump Markov nonlinear systems. The standard particle filter can easily show particle degeneracy around mode changes. In the new algorithm, the number of particles in each mode is a tuning parameter that can be selected, for instance, a priori by the designer without distorting the posterior probability density function. This new algorithm offers a more accurate estimate with the same number of particles. Although the results of this paper are generally applicable, the paper focusses on state estimation for jump Markov systems in a radar target tracking application.