# How Generative Models Improve LOS Estimation in 6G Non-Terrestrial Networks

arXiv: https://arxiv.org/abs/2305.18845 · HTML: https://mlsanity.com/paper/2305.18845/

## Key facts

- Cited by 1 papers in the map (data through 14 Oct 2026).
- Builds on 2 of its 12 references that are in the map.
- First posted on arXiv 30 May 2023.

## Builds on (most cited first, 2 of 2)

- [Modeling Tabular data using Conditional GAN](https://mlsanity.com/paper/1907.00503/) (2019)
- [Generative Neural Network Channel Modeling for Millimeter-Wave UAV Communication](https://mlsanity.com/paper/2012.09133/) (2020)

## Built on by (most cited first, 1 of 1)

- [Agentic Satellite-Augmented Low-Altitude Economy and Terrestrial Networks: A Survey on Generative Approaches](https://mlsanity.com/paper/2507.14633/) (2025)

## Abstract (arXiv)

With the advent of 5G and the anticipated arrival of 6G, there has been a growing research interest in combining mobile networks with Non-Terrestrial Network platforms such as low earth orbit satellites and Geosynchronous Equatorial Orbit satellites to provide broader coverage for a wide range of applications. However, integrating these platforms is challenging because Line-Of-Sight (LOS) estimation is required for both inter satellite and satellite-to-terrestrial segment links. Machine Learning (ML) techniques have shown promise in channel modeling and LOS estimation, but they require large datasets for model training, which can be difficult to obtain. In addition, network operators may be reluctant to disclose their network data due to privacy concerns. Therefore, alternative data collection techniques are needed. In this paper, a framework is proposed that uses generative models to generate synthetic data for LOS estimation in non-terrestrial 6G networks. Specifically, the authors show that generative models can be trained with a small available dataset to generate large datasets that can be used to train ML models for LOS estimation. Furthermore, since the generated synthetic data does not contain identifying information of the original dataset, it can be made publicly available without violating privacy