Rsearch On Dynamic Spectrum Allocation Based On Auction Theory In Cognitive Radio Networks

With the development of wireless communication and the increase of people’srequirements, spectrum resource has become scarcer increasingly, which highlights thedisadvantage of traditional static spectrum allocation scheme whose spectrum utilizationis inefficient. Cognitive Radio has provided a new method to use the radio resourceflexibly and efficiently. Moreover, DSA (Dynamic Spectrum Allocation，DSA), which isable to make use of the idle spectrum in spectrum hole to realize the effective sharingand dynamic management, could solve the inefficiency radically. The research on DSAis of high theoretic and practical importance.This thesis mainly researches the dynamic spectrum allocation based on auctiontheory in cognitive radio networks including two different network environments.Firstly, in the single radio access network environment, in order to improve theeffectiveness of traditional auction in DSA issue in cognitive radio networks, a newdynamic spectrum allocation algorithm based on one-band multi-winner auction isproposed. Compared with the original greedy algorithm, the new algorithm can achievebetter spectrum allocation solutions with lower computational complexity byintroducing the multiple greedy strategy. It increases the seller’s revenue by improvingthe VCG (Vickery-Clarke-Groves, VCG) mechanism, while retaining the dominantstrategy incentive compatible property. In addition, it suppresses the occurrence ofcollusion effectively. Simulation results show that the algorithm proposed can getspectrum allocation efficiency close to the optimal solution and increase the auctionrevenue significantly.Secondly, in the heterogeneous radio access network environment, in order toprevent secondary users from thronging in bidding, an improved continuous doubleprogressive auction scheme is adapted. Meanwhile, a fast convergence algorithm basedon MDP (Markov Decision Processes，MDP) is proposed to accelerate the convergencerate. It is able to decrease the number of asking and bidding beacause it could predictsystem’s future state through the analysis of the current state. Simulation results showthat the algorithm proposed can accelerate the convergence rate and increase the systemutility significantly.