Hierarchical fusion with maximum correntropy decentralized extended information filtering for target tracking in clustered WSNs

Abstract

In this paper, a novel hierarchical fusion based on maximum correntropy decentralized information filtering and covariance intersection has been proposed for clustered WSNs. First, the maximum correntropy extended information filtering is derived in a decentralized manner, which combines the advantages of the insensitivity to the impulsive noise from maximum correntropy with the stability of the decentralized algorithms. Then, the fusion center with an MCC-based covariance intersection algorithm is designed to fuse the local estimation from different clusters and retain the estimation consistency. Moreover, both simulations and physical experiments in the mobile target tracking application prove that the proposed method is capable of achieving a more precise estimation than the local estimators and the decentralized extended information filtering against impulsive noise and random link failures.