Multiple Hypotheses Mixing Filter for hybrid Markovian switching systems

This work addresses the problem of stochastic state estimation for hybrid Markovian switching systems. The proposed Multiple Hypotheses Mixing Filter (MHMF) combines the Generalized Pseudo Bayes´ (GPB) multiple hypotheses tracking with the Interacting Multiple Model´s (IMM) estimates mixing in order to improve performance, the later being a particular case of the MHMF. A hypotheses pruning step prevents the filter´s output to be degraded by estimates coming from very unlikely hypotheses and the mode transition probabilities are estimated online based on the measurements´ likelihoods. A target tracking application shows the MHMF´s utility as a stochastic filter for hybrid systems.