A novel learning-based spectrum sensing technique for cognitive radio networks

Spectrum sensing is one of the most challenging issues in Cognitive Radio (CR) networks. It should be performed efficiently to reduce number of false alarms and missed detections. This paper presents a novel approach, which employs collective intelligence developed via learning agents, for spectrum sensing in CR networks. The approach is used to share the sensed information, then digest it and make intelligent decisions about the presence or absence of primary users (PUs), by exploiting the accumulated history. The usage of history thus results in reduced sensing, subsequently requiring minimum activity in the common control channel (CCC), to help secondary users (SUs) exchange information and switch to the chosen empty space(s). Paper provides implementation of the proposed approach based on maxminfunctions integrated with a probabilistic decision making process. The performance analysis of the proposed approach shows that the usage of accumulated history by CR nodes results in reduced spectrum sensing by fine tuning the scan threshold.