Modèles de Markov triplet et filtrage de Kalman (Triplet Markov models and Kalman filtering) (Q1417121)

The authors propose generalizations to the classical linear state-space model and the Kalman filter. Instead of having two separate equations for the state and the observations, their models merge them in a single equation with additional parameters which cater for more flexible modelling of the state-observation dynamics. The more general variant of the model (the triplet model) introduces, besides the state and observations, auxiliary (artificial) variables in the equation. The authors show that their models are indeed strictly more general than the classical model.