Radio Map Interpolation using Graph Signal Processing

Interpolating a radio map is a problem of great relevance in many scenarios such as network planning, network optimization and localization. In this work such a problem is tackled by leveraging recent results from the emerging field of signal processing on graphs. A technique for interpolating graph structured data is adapted to the problem at hand by using different graph creation strategies, including ones that explicitly consider NLOS propagation conditions. Extensive experiments in a realistic large-scale urban scenario demonstrate that the proposed technique outperforms other traditional methods such as IDW, RBF and model-based interpolation.