Decentralized estimation over imperfectly-estimated channels in wireless sensor networks

This paper considers the decentralized estimation of an unknown parameter in a wireless sensor network (WSN) over orthogonal Rayleigh fading channels in the presence of channel estimation error (CEE). Two classes of estimators are proposed to mitigate the adverse effects caused by CEE according to whether the unknown parameter is modeled to be deterministic or random. In the deterministic case, a maximum likelihood estimator (MLE) is derived. Though optimal, the MLE typically has a high complexity preventing its practical implementation. To get a tradeoff between performance and complexity, we propose several low complexity alternatives such as the first order Taylor approximation of the MLE and the suboptimal best linear unbiased estimator (BLUE). We further develop the maximum a posteriori (MAP) estimator and the linear minimum mean squared error (LMMSE) estimator for the random case in parallel to those in the deterministic case. Simulation results show that the proposed estimators achieve a significant performance gain over that without considering CEE.