EUVIMEDEUVIMED entdecken
Crossref

Millimeter wave–3D massive MIMO: Deep prior‐aided graph neural network combining with hierarchical residual learning for beamspace channel estimation

Haridoss Sudarsan, Krishnakumar Mahendran, Srinivasan Rathika, Subburaj Nagan Yoga Ananth

International Journal of Communication Systems · 25.11.2024 · 37 (17)

DOI: 10.1002/dac.5918 ↗

Kurz gefasst

Fachartikel in der Zeitschrift International Journal of Communication Systems, online erschienen am 23.07.2024. Verfasst von Haridoss Sudarsan und 3 weiteren Autor:innen. Die Arbeit wurde bisher 3-mal zitiert (Crossref) und nennt 47 Literaturangaben.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Millimeter Wave (mmWave) communication has emerged as a transformative technology at the forefront of wireless communication. One of the key challenges in harnessing the potential of mmWave technology is overcoming the increased susceptibility to propagation losses and environmental obstacles. To address these challenges, Three‐Dimensional Massive Multiple‐Input Multiple‐Output (3D Massive MIMO) systems have gained traction. The 3D aspect extends this concept by considering the elevation dimension, allowing for enhanced spatial resolution and coverage. Accurate estimation of the channel in 3D Massive MIMO scenarios is particularly challenging because of the complex propagation characteristics of mmWave signals. This paper introduces an efficient‐Aided Graph Neural Network Combining with Hierarchical Residual Learning (DPrGNN‐HrResNetL), designed specifically for beamspace Channel Estimation (CE)in mmWave‐Massive MIMO environments. The proposed model leverages deep priors and GNN mechanisms to enhance the extraction of spatial features, while hierarchical residual connections facilitate effective information flow through the network. DPrGNN enables the model to capture and understand complex spatial relationships among different antenna elements. The incorporation of deep priors provides a mechanism for leveraging prior knowledge about channel characteristics. This enhances the efficiency of the learning process, allowing the model to learn and adapt more effectively. The integration of hierarchical residual connections facilitates effective information flow through the network. This is particularly important for modeling complex dependencies within the beamspace channel data, enhancing the learning capacity of the network. The performance of the DPrGNN‐HrResNetL model is evaluated across a range of Signal‐to‐Noise Ratios (SNRs), utilizing metrics such as Normalized Mean Squared Error (NMSE) to measure the accuracy of the estimation. The outcomes underscore the resilience and efficacy of the DPrGNN‐HrResNetL approach in achieving precise CE within demanding mmWave scenarios.

Zitieren

APA

Sudarsan, H., Mahendran, K., Rathika, S., & Ananth, S. N. Y. (2024). Millimeter wave–3D massive MIMO: Deep prior‐aided graph neural network combining with hierarchical residual learning for beamspace channel estimation. International Journal of Communication Systems, 37(17). https://doi.org/10.1002/dac.5918

Vancouver

Sudarsan H, Mahendran K, Rathika S, Ananth SNY. Millimeter wave–3D massive MIMO: Deep prior‐aided graph neural network combining with hierarchical residual learning for beamspace channel estimation. International Journal of Communication Systems. 2024;37(17). doi:10.1002/dac.5918

BibTeX
@article{Sudarsan2024,
  author = {Sudarsan, Haridoss and Mahendran, Krishnakumar and Rathika, Srinivasan and Ananth, Subburaj Nagan Yoga},
  title = {Millimeter wave–3D massive MIMO: Deep prior‐aided graph neural network combining with hierarchical residual learning for beamspace channel estimation},
  journal = {International Journal of Communication Systems},
  year = {2024},
  volume = {37},
  number = {17},
  doi = {10.1002/dac.5918}
}
Kostenlos starten →