A Learning-Based Distributed Spectrum Sensing Mechanism for IEEE 802.22 Wireless Regional Area Networks

It is now indisputable that the performance of cognitive radio networks is closely subject to the accuracy and reliability of the inherent spectrum sensing process. In this regard, the development of an efficient sensing mechanism is an imperative task, the performance of which not only relies on the choice of the sensing function, but it substantially depends on the efficiency of the sensing data fusion, i.e. the combining of outputs from individual sensing functions. Due to its importance as well as the lack of efficient algorithms, the spectrum sensing data fusion was left as open issue in the cognitive radio IEEE 802.22 standard for wireless regional area networks (WRANs). In this research, we address this open issue by proposing a novel distributed sensing algorithm for WRANs, named single-channel learning-based distributed sensing (SC-LDS). This algorithm is self-trained, stable, and compensates for fault reports using a reward-penalty approach. Moreover, it exhibits more uniform performance in all traffic regimes, is fair (reduces the false-alarm/mis-detection gap), adjustable to different application needs, and bandwidth efficient. Simulation results unanimously corroborate that the proposed SC-LDS algorithm outperforms other techniques such as the AND, OR and VOTING rules.