| WSN is intended to be deployed in environments where sensors can be exposed to circumstances that might interfere with measurements provided. Such circumstances include strong variations of pressure, temperature, radiation, and electromagnetic noise. Thus, measurements may be imprecise in such scenarios. Data fusion is used to overcome sensor failures, technological limitations, and spatial and temporal coverage problems. The data fusion can be implemented in both centralized and distributed systems.;In the centralized fusion algorithms, we propose four algorithms to be implemented in WSN. As a case study, we propose a remote monitoring framework for sand production in pipelines. Our goal is to introduce a reliable and accurate sand monitoring system. The framework combines two modules: a Wireless Sensor Data Acquisition (WSDA) module and a Central Data Fusion (CDF) module. The CDF module is implemented using four different proposed fusion methods; Fuzzy Art (FA), Maximum Likelihood Estimator (MLE), Moving Average Filter (MAF), and Kalman Filter (KF). All the fusion methods are evaluated throughout the simulation and experimental results. The results show that FA, MLE and MAF methods are very optimistic, to be implemented in WSN, but Kalman filter algorithm does not lend itself for easy implementation; this is because it involves many matrix multiplications, divisions, and inversions. The computational complexity of the centralized KF is not scalable in terms of the : rk size. Thus, we propose to implement the Kalman filter in a distributed fashion. The proposed DKF is based on a fast polynomial filter to accelerate distributed average consensus. The idea is to apply a polynomial filter on the network matrix that will shape its spectrum in order to increase the convergence rate by minimizing its second largest eigenvalue. Fast convergence can contribute to significant energy savings. In order to implement the DKF in WSN, more power saving is needed. Since multiplication is the atomic operation of Kalman filter, saving power at the multiplication level can significantly impact the energy consumption of the DKF. This work also proposes a novel light-weight and low-power multiplication algorithm. Experimental results show that the TelosB mote can run DKF with up to 7 neighbors. |
