Distributed Data Fusion: Neighbors, Rumors, and the Art of Collective Knowledge

The Dark Side of Decentralized Target Tracking : Unknown Correlations and Communication Constraints

“…In particular, network-induced effects such as packet delays and losses or quantization are studied. Distributed estimation is a key tool for multisensor data fusion, and [31] presents a broad overview of the underlying theory and the required Bayesian inference techniques. A concise discussion and review of the past forty years of distributed estimation presented in [40].…”

Distributed Kalman filter

“…Therefore, this chapter also presents a solution based on the accumulated state density (ASD), which is closely related to the DKF but does not require the measurement models to be known. Surveys that reflect the history of research in distributed estimation can be found, for instance, in [2,3]. This chapter is structured as follows: Section 2 summarizes the problem formulation.…”

Bayesian data fusion for distributed learning

“…Distributed data fusion research can be categorized into Bayesian or consensus-based approaches [26]. Bayesian methods focus on preserving the full distribution of the unknowns given the data, called posterior, over the estimated process at each agent, so that sensor data can be easily and recursively merged with prior knowledge and does not need to be stored.…”

“…Although on the other side, DDF is intrinsically more adaptive and resilient to failures. When dealing with distributed agents, some of the challenges involve both dealing with observations impacted by the same process noise [35] and also non-independence of local estimates due to multiple counting of data [26,30,36,37], which essentially means that in a distributed architecture local agent estimates may be correlated. To maintain optimality and consistency, a distributed fusion method should account for such cross-correlation issues.…”