Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Empowering Wireless Network Applications with Deep Learning-based Radio Propagation Models

Stefanos Bakirtzis, Cagkan Yapar, Marco Fiore, Jie Zhang, Ian Wassell

TL;DR

This work addresses the need for accurate yet efficient radio propagation modeling in next-generation wireless networks characterized by mmWave/THz bands, RIS, and high mobility. It surveys conventional propagation models and presents data-driven DL-based approaches that learn input–QoI mappings from environmental and system features using architectures such as MLPs, CNNs, and Transformers, with training/validation/testing pipelines and standard error metrics. Key findings include substantial planning-time reductions with DL surrogates, RIS pattern inference and beamforming acceleration, improved localization via synthetic radio maps and neural ray tracing, and faster uncertainty quantification compared to traditional methods. The study argues that DL-based propagation models can enable real-time network automation and reliability, while outlining open challenges and directions (datasets, learning ray trajectories, and LLM-based reasoning) to guide future research.

Abstract

The efficient deployment and operation of any wireless communication ecosystem rely on knowledge of the received signal quality over the target coverage area. This knowledge is typically acquired through radio propagation solvers, which however suffer from intrinsic and well-known performance limitations. This article provides a primer on how integrating deep learning and conventional propagation modeling techniques can enhance multiple vital facets of wireless network operation, and yield benefits in terms of efficiency and reliability. By highlighting the pivotal role that the deep learning-based radio propagation models will assume in next-generation wireless networks, we aspire to propel further research in this direction and foster their adoption in additional applications.

Empowering Wireless Network Applications with Deep Learning-based Radio Propagation Models

TL;DR

This work addresses the need for accurate yet efficient radio propagation modeling in next-generation wireless networks characterized by mmWave/THz bands, RIS, and high mobility. It surveys conventional propagation models and presents data-driven DL-based approaches that learn input–QoI mappings from environmental and system features using architectures such as MLPs, CNNs, and Transformers, with training/validation/testing pipelines and standard error metrics. Key findings include substantial planning-time reductions with DL surrogates, RIS pattern inference and beamforming acceleration, improved localization via synthetic radio maps and neural ray tracing, and faster uncertainty quantification compared to traditional methods. The study argues that DL-based propagation models can enable real-time network automation and reliability, while outlining open challenges and directions (datasets, learning ray trajectories, and LLM-based reasoning) to guide future research.

Abstract

The efficient deployment and operation of any wireless communication ecosystem rely on knowledge of the received signal quality over the target coverage area. This knowledge is typically acquired through radio propagation solvers, which however suffer from intrinsic and well-known performance limitations. This article provides a primer on how integrating deep learning and conventional propagation modeling techniques can enhance multiple vital facets of wireless network operation, and yield benefits in terms of efficiency and reliability. By highlighting the pivotal role that the deep learning-based radio propagation models will assume in next-generation wireless networks, we aspire to propel further research in this direction and foster their adoption in additional applications.
Paper Structure (11 sections, 3 figures, 1 table)

This paper contains 11 sections, 3 figures, 1 table.

Table of Contents

  1. Introduction
  2. Radio Propagation Modeling Under the Prism of Deep Learning
  3. Conventional Propagation Models
  4. Deep Learning Data-driven Propagation Models
  5. Applications
  6. Network Planning
  7. Reconfigurable Intelligence Surface Modeling
  8. Localization
  9. Uncertainty Quantification
  10. Open Challenges and Future Directions
  11. Conclusion

Figures (3)

  • Figure 1: Next-generation wireless ecosystem comprising a plethora of diverse wireless channels, each posing unique challenges and affected by different propagation mechanisms; radio propagation models have been evolving alongside to cope with these challenges
  • Figure 2: Characteristic pipeline of a data-driven propagation model: a set of input features are processed by a DL module which transforms it to a wireless channel QoI.
  • Figure 3: Visualization of the exploitation of DL-based model for empowering wireless network applications.