中文摘要：

为了提高认知无线网络中频谱共享所产生的效益，研究了认知网络的频谱管理技术，提出了一种基于代理的频谱交易算法。该方法以代理商作为频谱交易和分配的中介，减轻了多主用户、多次用户频谱交易过程中的系统开销；在各主用户服务提供商之间采取相互竞争或相互合作的模式以最大化自身利益或总体利益为目标与代理商达成最终的频谱交易价格和频谱出售量，代理商把交易得来的频谱在次用户之间进行拍卖，各次用户向代理商投标，代理商根据其投标分配相应的频谱，次用户自适应地调整投标以使自身利益最大化。主用户服务提供商之间的竞争或合作和次用户之间的竞拍均采用纳什均衡作为最终的结果。基于MATLAB对所提算法的性能进行了仿真，仿真结果验证了该算法的有效性。

英文摘要：

In order to improve the benefits generated by spectrum sharing, the spectrum management was studied and the agent-based dynamic spectrum trading algorithm was proposed. In the algorithm, the agent was used as a medium for spectrum trading, which reduced the huge system overhead among multi-primary service providers （PSP） and multi-secondary users （SU）. All PSP competed or cooperated with each other to achieve the maximum self-profit or total profit respectively, and come to terms with the agents on the final transaction price and quantity of spectrum. And then, the agent auctioned the spectrum among all SU. Each SU submitted a bid to the agent, then the agent allocated corresponding spectrum quantity to the SU. All SU adaptively adjusted the bids to achieve the maximum selfrofit. Nash equilibrium was considered as the solution to both the game among PSP and the auction among SU. Simulation results show that the proposed algorithm performs well.