NeTS: Small: Dynamic Spectrum Access by Learning Primary Network Topology

University Of California-Los Angeles, Los Angeles CA

Investigators

Abstract

Primary user networks provide wireless coverage over a large area by using multiple geographically separated transmitters operating on licensed frequency bands. On the other hand, cognitive radios are opportunistic users of unoccupied spectrum. In order to coexist with the transmitters in its vicinity, they need to sense the spectrum, i.e., detect the primary users' transmissions, and avoid causing them interference. This project focuses on the dynamic spectrum access in the presence of primary networks that are cellular, have anisotropic antennas, or employ frequency reuse, and aims to significantly increase the range of spectrum that cognitive radios can use. Through higher-layer radio scene analysis we aim to maximize the cognitive radio network throughput by increasing the detected spatial resolution and utilization of spectrum holes. This combination of increased spectrum and geographical spread makes large-scale cognitive radio networks a viable candidate for implementing smart grids, environmental networks, and traffic sensors. For defense purposes, the learned primary network topology provides unprecedented information about commonly deployed communication networks. This project research is based on development of cooperative algorithms for the identification of spectrum holes and classification of primary network activity by geographically large-scale cognitive radio networks. A novelty of the proposed methods is that they will not rely on knowledge of the channel propagation models or the location of the radios. This blind nature of the algorithms allows for more diverse applications. Instead of geographical vicinity, correlations in the received signals will be used for distinguishing between primary transmitters and learning the shape of their footprints. For the case that footprints of two or more transmitters overlap, message passing based cooperative spectrum sensing methods will be formulated for the joint estimation of the multiple transmitters spectrum occupancy. The learned footprint and the detected spectrum occupancy will be used to analyze the primary user activity over time. Primary users that are part of the same infrastructure-based primary networks will be identified. Further, learning their traffic statistics will enable the classification of their channel access methods and protocols.