基于Q-Learning的认知无线电系统感知管理算法  

Q-Learning Based Sensing Task Management Algorithm for Cognitive Radio Systems

作　　者：李默[1] 徐友云[1] 蔡跃明 

机构地区：[1]解放军理工大学通信工程学院,南京210007

出　　处：《电子与信息学报》2010年第3期623-628,共6页Journal of Electronics & Information Technology

基　　金：国家973计划项目(2009CB3020402);国家863计划项目(2007AA01Z267;2009AA01Z249)资助课题

摘　　要：认知无线电系统不仅是一个自适应系统,更应该是一个智能系统。该文将智能控制中的Q-Learning思想引入到认知无线电系统中,用于解决感知任务在认知用户之间的分配问题,给出了一种基于Q-Learning的感知管理算法。该算法在不知道信道状态信息以及不需要对主用户业务进行估计的假设下通过不断地与环境进行交互和学习来给认知用户分配感知任务。仿真表明,该算法能够提高感知效率,并且收敛速度较快,可作为未来认知无线电系统走向智能化的一种尝试。More than an adaptive system,the cognitive radio system is an intelligent system.The Q-Learning of the intelligent control theory is adopted in the paper,to solve the sensing task allocation problem among cognitive users.And a Q-Learning based sensing management algorithm is proposed.The algorithm allocates sensing tasks to users through times of interaction with the environment and self-learning.The scheme of the paper works without any channel state information and estimation of primary traffic.From the simulation result,the algorithm could improve the sensing efficiency compared to the static allocation algorithm and attain to the convergence in a short time,which could be an attempt to the future intelligent cognitive radio systems.

关 键 词：认知无线电 Q-LEARNING 感知任务管理 

分 类 号：TN92[电子电信—通信与信息系统]