No abstract

Spectrum maps, which provide RF spectrum metrics such as power spectral density for every location in a geographic area, find numerous applications in wireless communications such as interference control, spectrum management, resource allocation, and network planning to name a few. Spectrum cartography techniques construct these maps from a collection of measurements collected by spatially distributed sensors. Due to the nature of the propagation of electromagnetic waves, spectrum maps are complicated functions of the spatial coordinates. For this reason, model-free approaches have been preferred. However, all existing schemes rely on some interpolation algorithm unable to learn from data. This work proposes a novel approach to spectrum cartography where propagation phenomena are learned from data. The resulting algorithms can therefore construct a spectrum map from a significantly smaller number of measurements than existing schemes since the spatial structure of shadowing and other phenomena is previously learned from maps in other environments. Besides the aforementioned new paradigm, this is also the first work to perform spectrum cartography with deep neural networks. To exploit the manifold structure of spectrum maps, a deep network architecture is proposed based on completion autoencoders.

Spectrum cartography constructs maps of metrics such as channel gain or received signal power across a geographic area of interest using spatially distributed sensor measurements. Applications of these maps include network planning, interference coordination, power control, localization, and cognitive radios to name a few. Since existing spectrum cartography techniques require accurate estimates of the sensor locations, their performance is drastically impaired by multipath affecting the positioning pilot signals, as occurs in indoor or dense urban scenarios. To overcome such a limitation, this paper introduces a novel paradigm for spectrum cartography, where estimation of spectral maps relies on features of these positioning signals rather than on location estimates. Specific learning algorithms are built upon this approach and offer a markedly improved estimation performance than existing approaches relying on localization, as demonstrated by simulation studies in indoor scenarios.

No abstract