Physics-Inspired Machine Learning for Radiomap Estimation: Integration of Radio Propagation Models and Artificial Intelligence
Songyang Zhang, Brian Choi, Feng Ouyang, Zhi Ding
Abstract
Radiomap captures the geometrical distribution of radio frequency signal power. As an important tool to qualitatively and quantitatively describe radio propagation behavior and spectrum occupancy, radiomap has found broad applications in deployment and configuration of modern wireless and Internet-of-things networks. Practically, a high-resolution radiomap can be reconstructed from partial or sparse observations collected by mobile devices or sensors. To leverage the power of radiomap, efficient radiomap reconstruction from sparse samples has emerged as an urgent challenge. To capture the underlying data statistics, and also to estimate the physical radio frequency models, this work introduces three types of physics-inspired machine learning approaches to radiomap reconstruction. The experimental results demonstrate the potential of integrating data-driven artificial intelligence with model-based radio propagation behavior for radiomap reconstruction.