Event-based state estimation for stochastic hybrid systems

This study presents a state estimation algorithm for the stochastic hybrid system with event-based sampling. In event-based sampling, sensors transmit their measurements to an estimator only when predefined events happen, to reduce the communication cost. On the basis of the event-based sampling, the hybrid state estimation problem is formulated as to compute the probability density of the hybrid state with the sequence of noisy measurements generated at certain events. This hybrid state estimation problem is challenging since it requires computation of the exponentially increasing number of probabilities of the discrete state histories and evaluation of the multivariate integration. The proposed event-based hybrid state estimation algorithm utilises the interacting multiple model approach and pseudo measurement generation method to overcome these difficulties. The algorithm is then demonstrated with an illustrative aircraft tracking example.