Centralized and Distributed Millimeter Wave Massive MIMO-Based Data Fusion With Perfect and Bayesian Learning (BL)-Based Imperfect CSI

This paper presents low-complexity decision rules as well as the pertinent analysis for data fusion in millimeter wave (mmWave) massive multiple-input multiple-output (MIMO) wireless sensor networks (WSNs). The proposed framework considers both unknown and known parameter scenarios, and the spatial correlation arising due to close proximity of the sensors for both the centralized MIMO (C-MIMO) and distributed MIMO (D-MIMO) antenna configurations. The resulting detection performance is characterized by determining the closed-form expressions of probabilities of detection and false alarm for both antenna configurations. The optimal sensor gains are also determined for both the D-MIMO and C-MIMO architectures to further improve the detection performance. Additionally, asymptotic analysis is presented for both antenna configurations to determine the power scaling laws for the mmWave massive MIMO WSN, which lead to an improved sensor battery life without sacrificing the system performance. Furthermore, decision rules are also derived along with the pertinent analysis for a practical scenario with uncertainty in the channel state information (CSI) at the fusion center, wherein CSI of the mmWave massive MIMO channel is estimated using the novel sparse Bayesian learning (SBL) framework. Simulation results are presented to illustrate the performance of the proposed schemes and to validate the analytical results.