Adaptive estimation over distributed sensor networks with a hybrid algorithm

Estimation of unknown parameters associated with a distributed sensor network using its noisy measurements has been an active research area recently. Several estimation algorithms, such as the incremental and diffusion algorithms, have been proposed to address this problem. Incremental algorithms require less communication among nodes of the networks while diffusion algorithms are more robust and require large amounts of energy for communication. In this study, we have proposed a hybrid methodology that combines incremental and diffusion algorithms based on the property of a priori error, where is the difference of output error and noise variance of each sensor. The proposed network started with an incremental communication scheme and switched to diffusion scheme to complete the rest of the estimation. Simulation results showed that the proposed algorithm largely improved the convergence rate as well as the estimation accuracy.