Parameter Estimation in Distributed Sensor Networks and CFO Estimation in OFDM Systems

The research of this dissertation addresses two important issues relevant to parameter estimation: maximizing network lifetime (NLT) of distributed wireless sensor networks deployed to estimate and track a common parameter; and carrier-frequency-offset (CFO) estimation in orthogonal frequency-division-multiplexing (OFDM) systems.

In practice, the sensor nodes (SNs) of most wireless sensor networks are energy-constrained. To estimate and track a common parameter using such energy-constrained networks, we propose to employ set-membership adaptive filter (SMAF) in each of the SNs. The SMAF updates (thus transmits) parameter estimates only if the magnitude of the estimation error exceeds a predefined threshold, which affects the frequency of updates and the overall estimation performance. This approach renders a nicely formulated trade-off mechanism, resulting in more frugal energy use for SNs, and prolonged NLT. This approach also leads to a solution framework that relates maximizing NLT to network performance, i.e., meeting a performance constraint defined based on the mean-squared-deviation (MSD) of the consensus estimate. The NLT maximization is posed as a constrained optimization problem whose solution yields the optimal error thresholds for the SMAFs that maximize the NLT. The optimal solution is obtained by using an iterative binary search algorithm, which also solves a problem of node selection. The robustness of the proposed solution is investigated with respect to the spatial correlation and the impact of the uncertainty in the knowledge of spatial correlation on the NLT. We also solve NLT maximization for the case of optimum energy with a constraint on total energy. Simulation results using data from real-world applications show that the proposed approach offers substantially prolonged NLT over conventional tracking algorithms such as normalized least mean-squares (NLMS) adaptive filters.

The second part of the dissertation focuses on estimating CFO in OFDM systems taking into account power amplifiers (PA) nonlinearity in time-varying multipath fading channels, like those in mobile environments. We derive Cram`{e}r-Rao lower bound (CRLB) and approximate maximum-likelihood-estimators for CFO estimation in the static and time-varying channel scenarios. Analysis and simulation reveal that Doppler fading introduces a floor on the accuracy of CFO estimation. We then study the impact of PA nonlinearity on the accuracy of CFO estimation and present the modified CRLBs for the same. Performance of an ideal predistortion (PD) scheme is compared to that of a practical PD scheme employing unscented Kalman filter (UKF). Simulation results corroborate our theoretical analysis and prove the efficacy of our proposed PD filter to compensate for nonlinearity.