AirMAP: Scalable Spectrum Occupancy Recovery Using Local Low-Rank Matrixapproximation

We propose AirMAP, a framework for enabling scalable database-driven dynamic spectrum access and sharing. We bring together the merits of compressive sensing and collaborative filtering to provide accurate radio occupancy map while reducing the network overhead cost and overcome the scalability issue with conventional approaches. We start from an observation that close-by users have a highly correlated spectrum observation and we propose to recover the spectrum occupancy matrix in the borough of each sensing node by minimizing the rank of local sub-matrices. Then, we combine the recovered matrix entries using a similarity criterion to get the global spectrum occupancy map. Through simulations, we show that the proposed framework minimizes the error while reducing the network overhead. We also show that the proposed framework is scalable when considering high frequencies.