Distributed State Estimation over Wireless MIMO Fading Channels with Imperfect CSI
We consider distributed state estimation of a linear plant under stabilizing feedback over wireless multiple-input multiple-output (MIMO) fading channels, where channel-estimation noise corrupts the channel state information (CSI) used for information fusion. We propose a distributed estimator that combines local prediction, communication over wireless MIMO channels, and a posterior update based on local information and communicated data. Each sensor uses noisy pilots to estimate its aggregated wireless MIMO fading channel, and a pilot-noise bias correction recovers the required neighboring-data differences in conditional mean. We first analyze the mean estimation error and show that a Chebyshev design of the communication weights achieves a prescribed decay factor $d\in(0,1)$ satisfying the sensing conditions, with finitely many communication updates per plant sample at every finite signal-to-noise ratio (SNR) gain. We then characterize mean-square stability and the effects of SNR gain and receive-array size on the state estimation mean-square error (MSE), including its transient and steady-state value. In the ideal-communication limit, we show that the square root of the absolute MSE deviation from its steady-state value decays as $o(d^k)$. Numerical simulations verify the theoretical results and show substantial performance gain compared to various state-of-the-art baselines.