Robust Channel Modeling of 2.4 GHz and 5 GHz Indoor Measurements: Empirical, Ray Tracing, and Artificial Neural Network Models

Robust channel models for indoor areas are a crucial part of network planning and are immensely valuable for the small cell and indoor 5G network evolution. As the main input for many resource allocation and network planning problems, the accuracy of the path loss model can improve the overall accuracy of these techniques. Previous measurement campaigns exist for outdoor areas and higher frequencies; however, extensive indoor measurements at these frequencies are missing from the literature. Both WLAN and LTE networks use 2.4- and 5-GHz bands. For this work, indoor measurements were carried out in two distinct indoor environments, at two frequencies, and various models were compared. The measurements were made at the Deutsches Museum Bonn and the ICT cubes, an office space at RWTH Aachen University. Both empirical and deterministic models are tested on the data; the free-space path loss model, the single and dual-slope models with line-of-sight and nonline-of-sight, ray tracing models, and artificial neural network models were all tested and evaluated. Overall, the artificial neural network combined with the free-space path loss model proved to be the most robust model, which accurately predicted the propagation in the indoor environments, at both frequencies.

Automated summary

Robust channel models for indoor areas are crucial for network planning and can improve the accuracy of resource allocation and network planning techniques. This study conducted measurements in different indoor environments and compared various models, finding that the combination of artificial neural networks and the free-space path loss model is the most robust for accurately predicting propagation in indoor environments.