Resilient labeled multi-Bernoulli fusion with peer-to-peer sensor network

This paper addresses fusion of labeled multi-Bernoulli (LMB) densities in a peer-to-peer sensor network subject to misbehaviors due to abnormal conditions of the signal propagation channel (i.e. low signal-noise-ratio) or to malicious attacks. In such situations unintentional or intentional modifications of the transmitted messages can occur which degrade the performance of standard fusion methods. In order to overcome the aforementioned difficulties, we propose a resilient approach to distributed fusion of LMB densities in which each node exploits the local density as trusted source of information to detect the modified/false data coming from neighboring nodes. The implementation issues of the proposed method as well as possible cyber-attack models in the context of multi-object filtering are also discussed. The performance of the proposed method is verified via simulation experiments concerning fusion under cyber-attacks.