Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Geo2SigMap: High-Fidelity RF Signal Mapping Using Geographic Databases

Yiming Li, Zeyu Li, Zhihui Gao, Tingjun Chen

TL;DR

Geo2SigMap tackles RF signal mapping by fusing geographic databases with an open-source ray-tracing workflow and a cascaded U-Net that learns to produce detailed SS maps from synthetic data. The framework automates building-map creation, PG map generation, and SS map inference, incorporating sparse field measurements at deployment to adapt to unseen areas. In real-world CBRS experiments over six LTE cells, Geo2SigMap achieved an average RMSE of 6.04 dB for RSRP prediction, outperforming five baselines by about 3.6 dB on average. This approach offers a scalable, transferable path to accurate map-level RF propagation predictions without requiring extensive real-world measurements for each new area, with open-source tooling enabling broad adoption.

Abstract

Radio frequency (RF) signal mapping, which is the process of analyzing and predicting the RF signal strength and distribution across specific areas, is crucial for cellular network planning and deployment. Traditional approaches to RF signal mapping rely on statistical models constructed based on measurement data, which offer low complexity but often lack accuracy, or ray tracing tools, which provide enhanced precision for the target area but suffer from increased computational complexity. Recently, machine learning (ML) has emerged as a data-driven method for modeling RF signal propagation, which leverages models trained on synthetic datasets to perform RF signal mapping in "unseen" areas. In this paper, we present Geo2SigMap, an ML-based framework for efficient and high-fidelity RF signal mapping using geographic databases. First, we develop an automated framework that seamlessly integrates three open-source tools: OpenStreetMap (geographic databases), Blender (computer graphics), and Sionna (ray tracing), enabling the efficient generation of large-scale 3D building maps and ray tracing models. Second, we propose a cascaded U-Net model, which is pre-trained on synthetic datasets and employed to generate detailed RF signal maps, leveraging environmental information and sparse measurement data. Finally, we evaluate the performance of Geo2SigMap via a real-world measurement campaign, where three types of user equipment (UE) collect over 45,000 data points related to cellular information from six LTE cells operating in the citizens broadband radio service (CBRS) band. Our results show that Geo2SigMap achieves an average root-mean-square-error (RMSE) of 6.04 dB for predicting the reference signal received power (RSRP) at the UE, representing an average RMSE improvement of 3.59 dB compared to existing methods.

Geo2SigMap: High-Fidelity RF Signal Mapping Using Geographic Databases

TL;DR

Geo2SigMap tackles RF signal mapping by fusing geographic databases with an open-source ray-tracing workflow and a cascaded U-Net that learns to produce detailed SS maps from synthetic data. The framework automates building-map creation, PG map generation, and SS map inference, incorporating sparse field measurements at deployment to adapt to unseen areas. In real-world CBRS experiments over six LTE cells, Geo2SigMap achieved an average RMSE of 6.04 dB for RSRP prediction, outperforming five baselines by about 3.6 dB on average. This approach offers a scalable, transferable path to accurate map-level RF propagation predictions without requiring extensive real-world measurements for each new area, with open-source tooling enabling broad adoption.

Abstract

Radio frequency (RF) signal mapping, which is the process of analyzing and predicting the RF signal strength and distribution across specific areas, is crucial for cellular network planning and deployment. Traditional approaches to RF signal mapping rely on statistical models constructed based on measurement data, which offer low complexity but often lack accuracy, or ray tracing tools, which provide enhanced precision for the target area but suffer from increased computational complexity. Recently, machine learning (ML) has emerged as a data-driven method for modeling RF signal propagation, which leverages models trained on synthetic datasets to perform RF signal mapping in "unseen" areas. In this paper, we present Geo2SigMap, an ML-based framework for efficient and high-fidelity RF signal mapping using geographic databases. First, we develop an automated framework that seamlessly integrates three open-source tools: OpenStreetMap (geographic databases), Blender (computer graphics), and Sionna (ray tracing), enabling the efficient generation of large-scale 3D building maps and ray tracing models. Second, we propose a cascaded U-Net model, which is pre-trained on synthetic datasets and employed to generate detailed RF signal maps, leveraging environmental information and sparse measurement data. Finally, we evaluate the performance of Geo2SigMap via a real-world measurement campaign, where three types of user equipment (UE) collect over 45,000 data points related to cellular information from six LTE cells operating in the citizens broadband radio service (CBRS) band. Our results show that Geo2SigMap achieves an average root-mean-square-error (RMSE) of 6.04 dB for predicting the reference signal received power (RSRP) at the UE, representing an average RMSE improvement of 3.59 dB compared to existing methods.
Paper Structure (13 sections, 7 equations, 8 figures, 3 tables)

This paper contains 13 sections, 7 equations, 8 figures, 3 tables.

Table of Contents

  1. Introduction
  2. Related Work
  3. System Design
  4. Building Maps and 3D Meshes
  5. Ray Tracing-based Path Gain (PG) Maps
  6. Cascaded U-Net for RF Signal Mapping
  7. Implementation and Measurements
  8. Dataset Generation and Cascaded U-Net Model Training
  9. Real-World Measurements
  10. Baseline Methods
  11. Evaluation
  12. Discussions and Limitations
  13. Conclusion

Figures (8)

  • Figure 1: (a) An example 512m$\times$512m area on the Duke West campus, (b) the corresponding building map generated by OSM and rendered by Blender, (c) an example building object in the 3D mesh used as input to Sionna, and (d) the simulated path gain (PG) map with an antenna placed at the center of the map with a height of 24m, 160$^{\circ}$ azimuth orientation, and 10$^{\circ}$ downtilt.
  • Figure 2: Geo2SigMap achieves efficient and precise RF signal mapping via a proposed cascaded U-Net architecture, which is composed of U-Net-Iso and U-Net-Dir for generating coarse path gain (PG) maps and fine-grained signal strengh (SS) maps, respectively.
  • Figure 3: A 6.41 million km2 area in North America (left), from which a total number of 27,176 512m$\times$512m areas with a building-to-land ratio of at least 20% are selected to generate the building map and PG map datasets used to train the cascaded U-Net model in Geo2SigMap. The trained model is evaluated using measurements conducted on the Duke University campus (right).
  • Figure 4: (a) Six LTE cells operating in the CBRS band deployed on the Duke University West Campus (detailed cell information in Table \ref{['tab:cbrs_cell_info']}). (b) Sample RSRP measurements collected by the UE when served by two cells (PCIs A and F) within the corresponding 512m$\times$512m area.
  • Figure 5: Root-mean-square error (RMSE) of the RSRP values predicted by Geo2SigMap with varying number of measurement points in the sparse map for different PCIs (left) and device types (right).
  • ...and 3 more figures