Fault-tolerant distributed fusion of PDFs using KLDs-induced functional FCM clustering
Zheng Wei, Zhansheng Duan · Chinese Journal of Aeronautics · 2025
In distributed fusion, when one or more sensors are disturbed by faults, a common problem is that their local estimations are inconsistent with those of other fault-free sensors. Most of the existing fault-tolerant distributed fusion algorithms, such as the Covariance Union (CU) and Fault-tolerant Generalized Convex Combination (FGCC), are only used for the point estimation case where local estimates and their associated error covariances are provided. A treatment with focus on the fault-tolerant distributed fusions of arbitrary local Probability Density Functions (PDFs) is lacking. For this problem, we first propose Kullback–Leibler Divergence (KLD) and reversed KLD induced functional Fuzzy c -Means (FCM) clustering algorithms to soft cluster all local PDFs, respectively. On this basis, two fault-tolerant distributed fusion algorithms of arbitrary local PDFs are then developed. They select the representing PDF of the cluster with the largest sum of memberships as the fused PDF. Numerical examples verify the better fault tolerance of the developed two distributed fusion algorithms.