Beamformers employ an array of antenna elements to collect the electromagnetic wave in the spatial domain and filter the corrupted signal in the element-space or beam-space. The spatial filtering performance of both element-space and beam-space beamformers is jointly determined by two key factors, i.e., beamformer geometry and excitation weights. In this work, we propose a cognitive sparse beamformer, which is capable of swiftly adapting entwined beamformer geometry and excitation weights according to the environmental dynamics through a “perception–action” cycle. In the “perception” step, situational information is extracted from the collected real-time data, and the sparse beamformer is updated in the “action” step via a regularized switching network, which divides the large array into groups and one antenna or beam is replaced with other candidates in the same group in the metric of array gain (AG) during each iteration. To circumvent the prohibitive computations resulted from matrix inversion accompanying with each update, closed-form formulas are derived to quantify the AG variation, thus facilitating efficient online beamformer reconfiguration. Extensive simulations show the effectiveness of the proposed cognitive sparse beamformer design method.
Cognitive Sparse Beamformer Design in Dynamic Environment via Regularized Switching Network
Greco, Maria;Gini, Fulvio
2022-01-01
Abstract
Beamformers employ an array of antenna elements to collect the electromagnetic wave in the spatial domain and filter the corrupted signal in the element-space or beam-space. The spatial filtering performance of both element-space and beam-space beamformers is jointly determined by two key factors, i.e., beamformer geometry and excitation weights. In this work, we propose a cognitive sparse beamformer, which is capable of swiftly adapting entwined beamformer geometry and excitation weights according to the environmental dynamics through a “perception–action” cycle. In the “perception” step, situational information is extracted from the collected real-time data, and the sparse beamformer is updated in the “action” step via a regularized switching network, which divides the large array into groups and one antenna or beam is replaced with other candidates in the same group in the metric of array gain (AG) during each iteration. To circumvent the prohibitive computations resulted from matrix inversion accompanying with each update, closed-form formulas are derived to quantify the AG variation, thus facilitating efficient online beamformer reconfiguration. Extensive simulations show the effectiveness of the proposed cognitive sparse beamformer design method.I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.