Resilient Fusion of LMB Densities Based on Medoids

This article deals with the problem of resilient fusion of labeled multi-Bernoulli (LMB) densities, which arises in the situation that the sensor network (SN) undergoes abnormal behaviors like malicious attacks, resulting in the change of transmitted data from each sensor node. Compared to fusion algorithms based on perfect SN conditions, a detection procedure should be deployed before performing fusion so as to exclude abnormal data. To this end, we propose to decompose the LMB densities as the union of Bernoulli components (BCs), and then the medoids of BCs are exploited to form the fused LMB density. Besides, a new density-based spatial clustering of applications with noise (DBSCAN)-based label-matching algorithm is proposed. The performance of the proposed algorithm is verified via simulations.