Time- and coefficient- selective diffusion strategies for distributed parameter estimation

We present an innovative approach to distributed estimation, featuring selective update of parameter estimates. In distributed parameter estimation, sensor nodes consume energy not only in processing data, but most costly, in communicating and diffusing updated parameter estimates. Reducing the number of parameters to be updated and reducing the frequency of updates are thus effective ways to save in energy consumption. The approach presented in this paper features advantages of set-membership adaptive filtering (SMAF) and those of partial updates in adaptive filtering. Simulation results show that the proposed algorithm offers substantial reduction in energy consumption without much, if any, performance degradation.