Clustering-Based Distributed Fault-Tolerant Target Tracking for Sensors With Decreased Detection Accuracy

The distributed state estimation problem of fault sensors with decreased detection accuracy is investigated. The biggest challenge is to reduce or even eliminate the influence of local state estimation of fault sensor on the fused state estimation in consensus fusion. The clustering-based distributed cubature information filter (DCIF) is proposed to deal with the challenge. To improve the accuracy of distributed state estimation, density-based spatial clustering of applications with noise (DBSCAN) clustering is used to distinguish the fault sensors with decreased detection accuracy from the normal sensors. After the normal sensors are selected by DBSCAN clustering, an improved consensus-based fusion method is designed to obtain distributed consensus estimations on the premise of keeping the communication topology of wireless sensor network (WSN) unchanged. The proposed method can delete the local state estimations of the fault sensors and realize the distributed consensus fusion. The boundedness of the estimated error is proved by the stochastic stability theory. Finally, the effectiveness of the proposed clustering-based DCIF algorithm is shown by simulation example.