Research On Spectrum Sharing Algorithm For Cognitive Radio Based On Game Theory

With the rapid development of wireless communication technology, the newwireless devices and applications makes the communication system need moreradio spectrum. Under the current fixed spectrum assignment policy, most of theavailable spectrum is already licensed to services providers, which causes thespectrum scarcity problem. However, a large portion of the licensed spectrum isunderutilized in the most of time and region. The dramatic increase in wirelessaccess services and low spectrum utilization necessitate a new communicationparadigm. Dynamic spectrum access and cognitive radio are emergingtechnology enabling open spectrum sharing flexibly and efficiently.The function in cognitive radio that realizes the dynamic spectrum sharingis called spectrum sharing. This paper will focus on this topic in our research.The behavior of the licensed networks, cognitive networks and secondary usersare analyzed in an economic way. A spectrum trading model is established basedon game theory, and an algorithm is proposed intending to improve the spectrumefficiency, which is also more realistic with the current policy and compatiblewith the existing legacy infrastructure.The spectrum trading between multiple licensed networks and multiplecognitive networks is modeled as a double auction in this paper. Based on thismodel, the paper first analyzes the performance variation under different numberof networks. Then, considering that the licensed and cognitive networks mayhave their own preferences in the trading process, the market equilibriumsolutions are compared under various preference combinations of2licensednetworks and2cognitive networks. As the vacant spectrum is dynamic changedwith the time, the channel status is also uncertain in each round of spectrumtrading. So this paper proposed a dynamic pricing algorithm in which the marketclearing price is associated with the exact bidding and asking prices in eachround. The performance of the spectrum double auction algorithm consideringboth the networks’ preference and dynamic pricing is compared with the solutionunder the market equilibrium, and the results reveal that the numbers of tradedchannels are very close to numbers of the equilibrium solution and meanwhilewith no rapid rise of the trading price.After buying the spectrum from the licensed networks, the cognitivenetworks will compete with each other to attract more cognitive users under acircumstance that the cognitive users are free to choose which network they access for their own applications. The competition is to set the proper price ofthe shared spectrum between different cognitive networks in this paper, which iseasily modeled as a Bertrand game. To build the Bertrand game model, thespectrum demand function of cognitive users and the utility function of cognitivenetworks are investigated which help us to form the normal representation of thegame. And the Nash equilibrium and Pareto optimality solution of the game isdeveloped and compared. The result shows that the Pareto optimality solutionhas the highest total profit of cognitive networks. However, this state is not astable equilibrium and a cognitive network may deviate from it to achieve ahigher utility. To establish collusion between the cognitive networks, a repeatedgame with a punishment mechanism is formulated in which a cognitive networkis more concerned about the long-term profit. The numeral results show that theutility function can reflect the saturation of satisfaction and the pricing of thespectrum is just the best-response for the cognitive networks.