Centralized and distributed LCMP beamforming based on the block conjugate gradient method

Abstract

This dissertation investigates the reception of multiple signals of interest (SOIs) amid interfering signals (ISs) using a collaborative network of spatially distributed nodes that form a virtual antenna array. Focusing on receiver beamforming for the Earthto-satellite link, both centralized and distributed linearly constrained minimum-power (LCMP) beamformers are presented, each using the block conjugate gradient method. The centralized LCMP beamformer is intended for scenarios in which a central fusion node estimates the weight vectors. In contrast, the distributed LCMP beamformer is developed by incorporating the average consensus (AC) protocol into the block conjugate gradient method, enabling each node to estimate its weight vectors from its own raw data and limited information exchange, thereby eliminating the need for a central fusion node. To further reduce computational complexity and inter-node data communication, the B mini-batch is incorporated into the proposed distributed LCMP beamformer. Numerical simulations are performed for both a planar scenario with regular node distributions and a spatial scenario with random node distributions, representing a swarm satellite network. The results indicate that both centralized and distributed LCMP beamformers converge as the number of snapshots increases, and that their performance remains consistent across the scenarios considered. For the distributed LCMP beamformer, numerical results show that increased network connectivity reduces execution time, and that the B mini-batch can substantially reduce it with a reasonable impact on symbol error rate. Numerical results further reveal that when the inter-node spacing exceeds half-wavelength, beam efficiency is limited by elevated side lobe levels, while the constraints on SOIs and ISs remain satisfied. Additionally, the numerical results confirm that the proposed LCMP beamformers effectively manage multiple SOIs in the presence of ISs. Finally, it is shown that the proposed LCMP beamformers reduce to the minimum power distortionless response (MPDR) beamformers under a single unity-gain constraint in the absence of interfering signals.