A Novel Robust Distributed IMM Filter Based on Normal‐Exponential‐Gamma Distribution

The estimation accuracy of traditional distributed interacting multiple model (IMM) filtering algorithms will decrease or even diverge when heavy‐tailed measurement noise (HTMN) appears. To solve this problem, a robust distributed IMM filtering algorithm using weighted average consensus (WAC) based on normal exponential gamma distribution (NEGD) is designed. Firstly, the measurement noise is modeled as normal exponential gamma distribution, the inverse Wishart (IW) distribution is selected as the conjugate prior distribution of scale matrices. Secondly, the system state vector and noise related arguments are jointly speculated applying variational Bayesian (VB) technique. What's more, information pairs and model probabilities are updated using WAC to improve estimate behavior of the designed algorithm and enhance the reliability of the sensor network. Finally, a simulation experiment of a maneuvering target tracking sensor network was used to validate the proposed algorithm.