基于博弈学习的动态频谱分配算法研究.pdfVIP

2010 年5 月 第27 卷第5 期 计算机仿真 文章编号:1以炀幽9348(2010)05 -01∞ -05 基于博弈学习的动态频谱分配算法研究 徐浩漫，唐伦，陈前城 (重庆邮电大学通信与信息工程学院，重庆4仪嗣5) 摘要t 为了解决认知无线网络中的动态频谱分配问题，提高频谱利用惑和避免于扰，提出了…种基于学习的库诺(ωm时 博弈模型，将授权用户对于空闲频谐的分配行为模拟为动态的博养过稳.并赋予授权用户学习的能力，通过对博弈过程中博 弈者行为的学习和总结形成新的博弈策略，而且远比较了将最优反应学习算法和模拟退火算法应用到自适应博弈学习中系 统的性能和用户的收益。仿真结果表明，两种算法均能够使授权用户通过学习达到策略的均衡，而有限理性下基于模拟退 火的自适应博弈学习算法的鲁糠做更强，收敛攸更好，且能够使授权用户获得更高的收益。 关键词:动态频谱分配;库诺博弈;模拟i昆火;最优反应动态学习 中图分类号:1啊91 1. 5 ,TN92 文献标识码:A An Investigation of Dynamic Spectrum Allocation Algorithm ßased on Game Learning XU Hao - man ,TANG Lun ,CHEN Qian 四 bin (Schω，1 of Comrnunication and Information Engineeri吨， Cho吨qing University of Pωts， and Telecomrnunications , Chongqing 翩翩5. China) ABSTRACT :With the special purpose of d严旧JIlÎc spectrum all∞atíon on wireless co.伊itíve network , a general ωurnot 伊mes fz割nework ís pro阳se ba晶ed on learning. The model of the behavíor of 由e primaryωers 咐10 allocate their fr咽 spectrum to 阳附ond us师as a collection of dynamic g刷创P阴阳 is 耐up. At the same time , we sup- pose that the primary users have the learning ability , that is , they can make s伽ategic decisions by 1侃rning from other players in the 伊m创. We apply and com归re the best - response leaming algorithm and the adaptive learning algo- rithm hωed on simulated annealing. The simulation results show that h创h methods enable the primaryωers to stabi- lize their strategy , hut the latter one is more steady and can give primary u肥rsabe阳r profit. KEYWORDS: Dynamic spectrum alloc刷ion; Co町not 伊mω; Simulated annealing; Best - response learni吨 alg，伊 ri