Discrete-Time Mode Filters for Markovian Jump Processes

The abrupt changes in a dynamical system due to internal and/or external disturbances can be characterized by a random step variable, called system mode. The simplest model for the mode variable is a finite state markov chain. The knowledge of the system mode is important for estimation, control, and prediction of such a system behavior. However, the mode can only be measured through noisy observations. This paper presents a comparative study of various mode filters based on continuous-valued and discrete-values observations in discrete time.