Scholarlink

<p>Accurately modeling wireless channel propagation is essential for the design and optimization of wireless systems. Large-scale propagation models predict path loss and received signal strength, with Recommendation ITU-R P.1546-6 being a key resource for scenarios involving extended distances and antenna heights, operating over 30 MHz to 4000 MHz and antenna heights from 10 m to 1200 m. This paper employs symbolic regression, a machine learning approach combining genetic programming with experimental data, to derive analytical expressions for the field-strength curves in this Recommendation. Unlike neural networks, which function as black-box models, symbolic regression yields explicit mathematical expressions that enable direct interpolation for unlisted configurations and provide analytical insights into the propagation behavior. Nevertheless, the derived formulas offer an alternative to tabulated data, bridging the gap between purely empirical approaches and black-box models such as neural networks. The findings contribute to the understanding of propagation challenges and highlight the trade-offs in analytical modeling for complex scenarios. Analytical formulas with minimal complexity were sought, under the constraint that the maximum absolute error in dB, compared to measured data, should be less than 2 dB for distances ranging from 1 km to 1000 km between the antenna and the receiver.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Symbolic regression approach for analytical modeling of ITU-R P.1546-6 propagation curves

  • 期刊论文

  • 发表日期：

    2025年5月18日

查看全文
  • Guilherme Kneitz de Oliveira,
  • Diego Barreto Haddad,
  • Gilson Antonio Giraldi,
  • Maurício Henrique Costa Dias

摘要

Accurately modeling wireless channel propagation is essential for the design and optimization of wireless systems. Large-scale propagation models predict path loss and received signal strength, with Recommendation ITU-R P.1546-6 being a key resource for scenarios involving extended distances and antenna heights, operating over 30 MHz to 4000 MHz and antenna heights from 10 m to 1200 m. This paper employs symbolic regression, a machine learning approach combining genetic programming with experimental data, to derive analytical expressions for the field-strength curves in this Recommendation. Unlike neural networks, which function as black-box models, symbolic regression yields explicit mathematical expressions that enable direct interpolation for unlisted configurations and provide analytical insights into the propagation behavior. Nevertheless, the derived formulas offer an alternative to tabulated data, bridging the gap between purely empirical approaches and black-box models such as neural networks. The findings contribute to the understanding of propagation challenges and highlight the trade-offs in analytical modeling for complex scenarios. Analytical formulas with minimal complexity were sought, under the constraint that the maximum absolute error in dB, compared to measured data, should be less than 2 dB for distances ranging from 1 km to 1000 km between the antenna and the receiver.