Robust Linear Decentralized Estimation in IoT Networks Relying on Imperfect CSI

This work conceives an optimal minimum mean square error (MMSE) robust transceiver design considering imperfect knowledge of the observation and channel matrices in a wireless sensor network (WSN). In order to model the uncertainty in the observation and channel matrices corresponding to each sensor, popular stochastic uncertainty model is being employed. The proposed transceiver design takes individual sensor power constraint into consideration which makes the proposed design much more apt for practical implementation. The resulting average MSE optimization subject to individual sensor power constraint is non-convex in nature and hence block coordinate descent (BCD) based alternating minimization framework is proposed which yields an optimal MSE combining and precoding matrices in each iteration. Simulation results corroborate our analytical findings, and the proposed robust transceiver design outperforms the uncertainty agnostic design in the existing literature.