Diffusion in Networks by Cooperative Particle Filtering

In this paper, we propose a diffusion method for estimation of hidden processes by cooperative agents in a network. We adopt particle filtering for estimating the hidden processes. Each filter has it own observation process, unknown to the rest of the agents, but the hidden process is the same for all the agents. The agents track the hidden process, form their posteriors of the process and then approximate it with Gaussians. Then they exchange these Gaussians to form fused posteriors and to allow for diffusing of information across the network. The Gaussian approximation of the posterior distributions simplify the calculations in the diffusion steps. The simulation results show that the proposed method has similar performance as the centralized method.