Impact of imperfect channel state information on the performance of wireless sensor networks

The performance of wireless sensor networks (WSN) is often assessed without paying close attention to the effects of channel state information recovery, which is assumed to be exactly known. This work focuses on the effects of imperfect knowledge of the channel state information on a WSN whose sensors measure a target parameter and send it to a common fusion center by an amplify-and-forward technique. A baseline estimation rule, consisting in disregarding the presence of noise in the estimated channel gain, is considered. Two other estimation rules are also studied, based on the joint processing of the received samples during the sensing and training intervals: i) a maximum a posteriori rule consisting in maximizing the a posteriori probability of the target parameter; ii) a least-squares rule based on the minimization of the mean-square error of the estimated target parameter. In both cases, conditionally on the target parameter, the received samples are assumed to be jointly Gaussian distributed as far as concerns the derivation of the estimation rule. Numerical results illustrate the merits of the proposed estimation rules by showing the MSE performance with different network parameters.