Distributed Sequential Filtering for Nonlinear Systems With Heavy‐Tailed Noise Under Binary Sensor Networks: A Variational Bayesian Approach

ABSTRACT In this article, the distributed sequential filtering problem is investigated for a class of nonlinear systems (NS) subject to non‐Gaussian heavy‐tailed noises through binary sensor networks. Since only one bit of output data is valid for binary sensors, a novel Gaussian tail function is proposed to gather valuable information from binary sensors for filtering purposes. Furthermore, a unified distributed sequential filtering framework for handling non‐Gaussian heavy‐tail noise with inaccurate statistics is developed by a variational Bayesian (VB) strategy combined with cubature Kalman filtering (CKF), which is according to the spherical‐radial cubature rule. To be more specific, the posterior distribution functions of system states together with the noise covariance (NC) and the auxiliary variable are jointly estimated under such a framework. In addition, the distributed sequential filter is received by Metropolis weights and arithmetic average fusion. Finally, an example of target tracking is utilized to reveal the effectiveness and applicability of the proposed distributed filtering algorithm.