5G mmwave network planning using machine learning for path loss estimation
- Author
- Yosvany Hervis Santana (UGent) , Rodney Martinez Alonso (UGent) , Glauco Guillen Nieto, Luc Martens (UGent) , Wout Joseph (UGent) and David Plets (UGent)
- Organization
- Project
- Abstract
- New frequency bands, such as the mmWave at 28 GHz for fifth Generation networks in Frequency Range 2, are being introduced to accomplish the required throughput for new services such as remote surveillance, object tracking, and factory automation. These high-frequency bands are sensible to reflections and diffraction. Therefore, accurate path loss calculation relies on advanced models and ray tracing. However, this is time-consuming, and evaluating a large set of candidate solutions is no longer possible when planning the optimal number and location of base stations in the network planning process. This paper investigates the use of machine learning to approximate a complex mmWave ray-tracing-based path loss model in indoor scenarios. The much lower calculation time allows us to approximate the ray-tracer path loss estimation well and to apply a genetic algorithm for realizing network planning. The machine learning model is trained and validated for two buildings and tested with another, with an average of the Mean Absolute Error of 2.8 dB over all cases. It is shown that the combination of Machine Learning and Genetic Algorithm is able to find a network deployment in the FR2 band accounting for the minimum number of access points and minimum electromagnetic exposure, while still providing a predefined coverage percentage inside the building.
- Keywords
- MULTIOBJECTIVE OPTIMIZATION, WIRELESS NETWORKS, GENETIC ALGORITHM, INDOOR, GHZ, Planning, Ray tracing, Computational modeling, 5G mobile communication, Millimeter wave communication, Estimation, Machine learning, 28 GHz, 5G, FR2, genetic algorithm, network planning, mmWaves, modeling machine learning, path loss, ray tracing, wireless InSite
Downloads
-
WICA 1207.pdf
- full text (Published version)
- |
- open access
- |
- |
- 4.90 MB
Citation
Please use this url to cite or link to this publication: http://hdl.handle.net/1854/LU-01JAFSHB33A0X014ZQE429NGQW
- MLA
- Hervis Santana, Yosvany, et al. “5G Mmwave Network Planning Using Machine Learning for Path Loss Estimation.” IEEE OPEN JOURNAL OF THE COMMUNICATIONS SOCIETY, vol. 5, 2024, pp. 3451–67, doi:10.1109/OJCOMS.2024.3405742.
- APA
- Hervis Santana, Y., Martinez Alonso, R., Nieto, G. G., Martens, L., Joseph, W., & Plets, D. (2024). 5G mmwave network planning using machine learning for path loss estimation. IEEE OPEN JOURNAL OF THE COMMUNICATIONS SOCIETY, 5, 3451–3467. https://doi.org/10.1109/OJCOMS.2024.3405742
- Chicago author-date
- Hervis Santana, Yosvany, Rodney Martinez Alonso, Glauco Guillen Nieto, Luc Martens, Wout Joseph, and David Plets. 2024. “5G Mmwave Network Planning Using Machine Learning for Path Loss Estimation.” IEEE OPEN JOURNAL OF THE COMMUNICATIONS SOCIETY 5: 3451–67. https://doi.org/10.1109/OJCOMS.2024.3405742.
- Chicago author-date (all authors)
- Hervis Santana, Yosvany, Rodney Martinez Alonso, Glauco Guillen Nieto, Luc Martens, Wout Joseph, and David Plets. 2024. “5G Mmwave Network Planning Using Machine Learning for Path Loss Estimation.” IEEE OPEN JOURNAL OF THE COMMUNICATIONS SOCIETY 5: 3451–3467. doi:10.1109/OJCOMS.2024.3405742.
- Vancouver
- 1.Hervis Santana Y, Martinez Alonso R, Nieto GG, Martens L, Joseph W, Plets D. 5G mmwave network planning using machine learning for path loss estimation. IEEE OPEN JOURNAL OF THE COMMUNICATIONS SOCIETY. 2024;5:3451–67.
- IEEE
- [1]Y. Hervis Santana, R. Martinez Alonso, G. G. Nieto, L. Martens, W. Joseph, and D. Plets, “5G mmwave network planning using machine learning for path loss estimation,” IEEE OPEN JOURNAL OF THE COMMUNICATIONS SOCIETY, vol. 5, pp. 3451–3467, 2024.
@article{01JAFSHB33A0X014ZQE429NGQW,
abstract = {{New frequency bands, such as the mmWave at 28 GHz for fifth Generation networks in Frequency Range 2, are being introduced to accomplish the required throughput for new services such as remote surveillance, object tracking, and factory automation. These high-frequency bands are sensible to reflections and diffraction. Therefore, accurate path loss calculation relies on advanced models and ray tracing. However, this is time-consuming, and evaluating a large set of candidate solutions is no longer possible when planning the optimal number and location of base stations in the network planning process. This paper investigates the use of machine learning to approximate a complex mmWave ray-tracing-based path loss model in indoor scenarios. The much lower calculation time allows us to approximate the ray-tracer path loss estimation well and to apply a genetic algorithm for realizing network planning. The machine learning model is trained and validated for two buildings and tested with another, with an average of the Mean Absolute Error of 2.8 dB over all cases. It is shown that the combination of Machine Learning and Genetic Algorithm is able to find a network deployment in the FR2 band accounting for the minimum number of access points and minimum electromagnetic exposure, while still providing a predefined coverage percentage inside the building.}},
author = {{Hervis Santana, Yosvany and Martinez Alonso, Rodney and Nieto, Glauco Guillen and Martens, Luc and Joseph, Wout and Plets, David}},
issn = {{2644-125X}},
journal = {{IEEE OPEN JOURNAL OF THE COMMUNICATIONS SOCIETY}},
keywords = {{MULTIOBJECTIVE OPTIMIZATION,WIRELESS NETWORKS,GENETIC ALGORITHM,INDOOR,GHZ,Planning,Ray tracing,Computational modeling,5G mobile communication,Millimeter wave communication,Estimation,Machine learning,28 GHz,5G,FR2,genetic algorithm,network planning,mmWaves,modeling machine learning,path loss,ray tracing,wireless InSite}},
language = {{eng}},
pages = {{3451--3467}},
title = {{5G mmwave network planning using machine learning for path loss estimation}},
url = {{http://doi.org/10.1109/OJCOMS.2024.3405742}},
volume = {{5}},
year = {{2024}},
}
- Altmetric
- View in Altmetric