IndoorDRaGon: Data-Driven 3D Radio Propagation Modeling for Highly Dynamic 6G Environments

Private 5G and future 6G networks offer significant potential for automation in vertical domains but must face the challenge of dynamic and rapid adaptation to new operational requirements. In this context, non-stationary, ad-hoc network operation is essential for continuously adapting reliable network solutions to rapidly changing environments. In this paper, we present IndoorDRaGon as a novel signal-strength prediction method for network planning of dynamic environments that combines expert knowledge from the mobile communications domain with lightweight machine learning methods based on random forests to achieve accurate and computationally efficient spatiotemporal quality of service predictions. In a comprehensive performance evaluation, the performance of IndoorDRaGon is compared to real-world measurements, ray tracing analysis and a vast range of state-of-the-art channel models. It is found that IndoorDRaGon achieves significantly better accuracy for unseen environments than the latter, even when only a tiny portion of measurements is considered in the training data.