A neighbor selected diffusion Kalman filter over communication resource constrained networks

Constrained communication resource is a crucial issue for any task involving sensor networks. To satisfy finite communication constraints, the authors propose a reduced-communication method for diffusion Kalman filter, each node produces its intermediate estimate with the information from its selected neighboring subset, the final estimate of each node can be obtained by combining its selected neighbors' intermediate estimates. The reduced-communication diffusion Kalman filter algorithm is unbiased in mean sense and this algorithm is also stable and convergent in mean-square sense. Its theoretical steady-state mean-square deviation (MSD) is given in this article. A rule is proposed to select the neighbors to reduce the network MSD. Simulation results are presented to demonstrate the effectiveness of the proposed approach.