Gossip-Based Distributed Filtering Over Networks Using Projection

This paper studies distributed state estimation for a dynamic system in a connected sensor network. Each node observes the system state, collects measurement, and works collaboratively to make state estimation. We propose two algorithms, namely weighted gossip-based distributed filtering (WG-DF) and convergent (as communication round increases) gossip-based distributed filtering (CG-DF). The information transmission across the network abides by an asynchronous gossip protocol, in which only two neighboring nodes are allotted to exchange information in each communication round. First, we convert our problem to a consensus problem. Then, we design the communication protocol and fusion weights by using projection. Finally, we analyze transient and steady-state behavior.