Self-Coexistence in Cognitive Radio Networks Using Multi-Stage Perception Learning

In this paper, we study the self-coexistence problem among competitive Cognitive Radio (CR) networks in an uncoordinated distributed wireless environment of homogeneous and heterogeneous bands. This problem can be correlated with famous optimal foraging theory, where the humming birds forage to explore islands in search of food sources to survive. The behavior of learning from observations leads them to find island of optimal resources. The proposed perception based learning mechanism for homogeneous spectra, helps the CR networks to strategize their choice of actions on the basis of rewards gathered from the accessed spectrum bands and successfully grab a clear chunk of spectrum. However, in heterogeneous bands scenario, the CR networks inadvertently choose the best suitable band greedily which lead them to collision. We incorporate a regret minimization technique with the proposed learning mechanism to resolve the contention among them and maximize system performance. Experimental results conclude that the networks could achieve the objective of finding a free spectrum with maximized system utility using the proposed heuristic within limited number of interactions.