Distributed Bernoulli Filter Based on Weighted Conservative Fusion

This letter presents a novel distributed Bernoulli filter for tracking a target over a peer-to-peer sensor network. For the case where the Bernoulli random finite set is either empty or singleton, two types of fusion weights are designed based on the information of sensor capabilities and environmental factors. Further, novel weighted conservative fusion approaches including geometric average and arithmetic average are derived to fuse local Bernoulli densities. The Gaussian mixture implementations of the proposed algorithms are given, which are subsequently combined with distributed flooding protocol for distributed averaging. Simulation results demonstrate the outperformance of the proposed approaches compared to the state-of-the-art approaches.