Second Order Multiple Model Estimator for Target Tracking in Clutter

The modern multiple model (MM) estimators assume that the transition probability matrix (TPM) governing the mode jumps is known. In target tracking systems, mode sequences usually assumes Markov Chain model. In the paper, a new adaptive algorithm is proposed which, unlike the standard MM model, updates the probabilities of the model taking into account the states from the two previous iterative cycles. The model probabilities from the two previous cycles are combined into a single model probability, based on the likelihoods of each model. The performance of the proposed MM estimation algorithm is tested with a se t of numerical simulations.