Enhancing Estimates Over WSN Using Robust Distributed Unbiased FIR Filtering With Consensus on Noise Power Gain

Abstract

Signal processing over wireless sensor networks (WSNs) often deals with big data collected from local nodes and sensors distributed over a wide area, where measurements are affected by uncontrollable natural effects, such as weather, electromagnetic noise, and physical limitations of the devices. Therefore, robust estimators have become a top priority to mitigate undesired effects and improve accuracy. In this article, a distributed unbiased finite impulse response (dUFIR) filter is developed with consensus on the generalized noise power gain (GNPG), as being a robust estimator that ignores noise and initial values. Based on numerical simulations and experimental measurements of ambient temperature, it is shown that the dUFIR filter outperforms the distributed Kalman filter (KF) in terms of accuracy and robustness under uncertain noise conditions.