Based On Game Theory In The Cognitive Wireless Network Dynamic Spectrum Allocation Technology

As a scarce resource for wireless communications, spectrum may be overcrowdedin the future to allow rapidly increasing users as well as various types of newapplications. Cognitive radio (CR) networks have been proposed as one of the mostpromising solutions for the scarcity of spectrum resource. Equipped with variousflexible abilities, including sensing/estimation of channel conditions and dynamicspectrum sharing/allocation, CR technology can observe, learn, optimize, and changethe transmission parameters to improve network performance in a time-varying wirelessenvironment. This dissertation focuses on the dynamic spectrum resources allocationtechnology of CR networks.In this dissertation, we consider a cognitive radio communication system withmultiple primary services and a secondary service. In this system, each primary servicewants to sell its unused spectrum to the single secondary service for gaining a certainprofit. Because each primary service is selfish, e.g., they are belonging to differentservice providers, it has to compete with all other primary services for maximizing itsprivate profit. Based on a repeated game model, an existing literature has formulated aspectrum sharing scheme to form collusion among all primary services for achieving thehighest total profit and also proposed the triggering strategy to punish those primaryservices deviated from the collusion. However, when the primary services do not valuetheir future profits enough, they cannot form such collusion even if a punishment isapplied to the deviating primary service. In this case, the primary services can onlyobtain the profit at the Nash equilibrium, which is lower than the highest profit atcollusion. To remedy this, the present dissertation introduces two new strategies to formcollusion among the primary services. One strategy supports the primary services toobtain a profit that, albeit lower than the highest profit, is higher than the Nashequilibrium profit, even if they do not value their future profits enough. The otheradopts the so-called “strongest credible punishment” to prevent possible deviation fromcollusion and can support those primary services, which cannot form collusion in theexisting scheme, to obtain the highest profit. Numerical simulation shows that the proposed two strategies outperform the existing scheme in maximizing the total profits.Meanwhile, when the primary services share their unused spectrum with the singlesecondary service, they may not be able to exchange their strategies fully due to theconstraint of wireless channels. Moreover, because of individual selfish behaviours,each primary service in the aforementioned non-cooperative competition may choose toconceal its strategy information, which hence is incomplete for others primary services.In this case, we propose a dynamic spectrum allocation strategy to increase the totalprofit of all primary services, that is, each primary service gradually adjust its spectrumprice to approach the Nash Equilibrium according to its current marginal profit function.Also, we deduce the convergence condition for this dynamic strategy.