Overlooked potential of systems with Markovian coefficients
Henk P. Blom · 1986
The usual system of linear stochastic difference equations with Markovian coefficients has been generalized to cover Rn valued processes which jump simultaneously with and due to jumps of the coefficients. The additional modelling potential is illustrated for a manoeuvring aircraft. Once being familiar with this overlooked potential of systems with Markovian coefficients, their use is remarkably simple. To filter partial observations of the generalized system candidate algorithms are reviewed, in particular the Generalized Pseudo Bayes algorithms and the more recent Interacting Multiple Model (IMM) algorithm. The IMM algorithm for the generalized system is applied to tracking a manoeuvring aircraft. The results indicate that the generalized models and the associated IMM algorithm has the potential to yield trackers that perform better than the existing decision-directed trackers.