Diffusion Kalman filter based on local estimate exchanges

We consider the problem of distributed Kalman filter design for a linear state-space model, where a set of nodes is required to collectively estimate the unknown state in real time, only through local estimate exchanges. This is very important in the cases when the local measurements are not available or expensive to transport. We further study the relationship between individually and locally linear optimal state estimates, and introduce the principle of information-sharing to distribute the local information among individual nodes. The proposed solution is totally distributed based on the diffusion policy, with no requirement of the centralized fusion center, data are processed and exchanged locally in the neighborhood. Simulation results verify the effectiveness of the proposed algorithm as compared with the existing techniques.