A novel robust distributed IMM filter for jump Markov systems with non-stationary heavy-tailed measurement noise

To cope with the issue of decreased estimation accuracy in the presence of non-stationary heavy-tailed measurement noise (NSHTMN) using traditional distributed IMM filtering algorithms, A novel robust distributed interactive multiple model (IMM) filtering algorithm is designed in this paper. Firstly, the NSHTMN is selected as Gaussian-Student's t-mixture (GSTM) distribution employing a Bernoulli variable, and the prior information of the mixed probabilities is calculated from the estimated value at the previous moment. Then the state vector, mixed probabilities, auxiliary parameters, and Bernoulli variables of the system are jointly estimated employing the variational Bayesian (VB) method. What's more, to increase the stability of sensor networks and the estimation performance of filtering algorithm, the weighted average consensus has been applied to update information pairs and model possibilities. Finally, a target tracking simulation experiment is conducted to illustrate that the designed algorithm has better estimation performance compared with cutting-edge algorithms.