Wireless sensor network for Spectrum Cartography based on Kriging interpolation

Dynamic spectrum access with Cognitive Radio (CR) network is a promising approach to increase the efficiency of spectrum usage. To allow the optimization of resource allocation and transmission adaptation techniques, each CR terminal needs to acquire awareness of the state of the time-frequency-location varying radio spectrum. In this paper we present a Spectrum Cartography (SC) approach where CR terminals are supported by a fixed wireless sensor network (WSN) to estimate and update the Power Spectral Density (PSD) over the area of interest. The wireless sensors collaborate to estimate the spatial distribution of the received power at a given frequency using either a centralized or a distributed Kriging (DK) algorithm. We present an analysis of the semivariogram models used to estimate the spatial statistics of wireless PSD distributions. The performance of the centralized and DK algorithms are evaluated by simulating different realizations of the PSD and the results are compared with classical interpolating schemes varying the density of nodes in the area and the number of nodes used for local estimation.