Sammendrag
Spectrum cartography constructs maps of metrics such as
channel gain or received signal power across a geographic
area of interest using measurements of spatially distributed
sensors. Applications of these maps include network planning,
interference coordination, power control, localization,
and cognitive radio to name a few. Existing spectrum cartography
methods necessitate knowledge of sensor locations,
but such locations cannot be accurately determined from pilot
positioning signals (such as those in LTE or GPS) in indoor
or dense urban scenarios due to multipath. To circumvent this
limitation, this paper proposes localization-free cartography,
where spectral maps are directly constructed from features
of these positioning signals rather than from location estimates.
The proposed algorithm capitalizes on the framework
of kernel-based learning and offers improved prediction performance relative to existing alternatives, as demonstrated by a simulation study in a street canyon.
Vis fullstendig beskrivelse