A Bayesian Interpretation of Distributed Diffusion Filtering Algorithms [Lecture Notes]
Marcelo G. S. Bruno, Stiven Schwanz Dias · IEEE Signal Processing Magazine · 2018
In this lecture note, we use a Bayesian methodology to formulate the optimal solution to the problem of cooperative tracking of a time-varying signal over a partially connected network of multiple agents, where each agent has sensing, processing, and communication capabilities of its own. Subsequently, assuming a general state-space model for the hidden state vectors and the agents? observations, we present, also from a Bayesian perspective, the general form of the adaptthen-combine (ATC) and the random exchange (RndEx) distributed diffusion filters, contrasting them with the ideal, optimal network filter.