基于强化学习的合作频谱分配算法  

Cooperative spectrum allocation algorithm based on reinforcement learning

作　　者：李冠雄 李桂林 LI Guanxiong;LI Guilin(School of Microelectronics,Tianjin University,Tianjin 300072,China;School of Electrical Information Engineering,Dalian Jiaotong University,Dalian 116021,China)

机构地区：[1]天津大学微电子学院,天津300072 [2]大连交通大学电气信息工程学院,大连116021

出　　处：《电波科学学报》2022年第1期8-14,共7页Chinese Journal of Radio Science

摘　　要：为了解决认知无线电网络中的频谱分配问题,提出了一种基于用户体验质量的合作强化学习频谱分配算法,将认知网络中的次用户模拟为强化学习中的智能体,并在次用户间引入合作机制,新加入用户可以吸收借鉴其他用户的强化学习经验,能够以更快的速度获得最佳的频谱分配方案;并且在频谱分配过程中引入了主用户和次用户之间的价格博弈因素,允许主用户根据自身情况对授权频谱进行定价;研究了不同频谱价格对于次用户收益的影响,使算法更加贴近现实场景.在系统评价方面,采用了平均意见得分模型对系统用户的服务质量进行直观展现.仿真结果表明,该算法可有效提升用户服务质量和系统的通信性能,为认识用户间的频谱分配提供了一种有效的方案.In order to solve the problem of spectrum allocation in cognitive radio networks, we propose a cooperative reinforcement learning spectrum allocation algorithm based on user experience quality. which simulates the secondary users in the cognitive network as agents in the reinforcement learning, and introduces a cooperation mechanism between the secondary users. New users can absorb and learn from the reinforcement learning experience of other users, and obtain the best spectrum allocation plan at a faster speed. In addition, the price game factor between the primary user and the secondary user is introduced in the spectrum allocation process, allowing the primary user to price the authorized spectrum according to their own situation, and the impact of different spectrum prices on the income of the secondary user is studied, making the algorithm closer to the real scene. In terms of system evaluation, the average opinion score model is used to visually display the service quality of system users. Simulation results show that the algorithm can effectively improve user service quality and system communication performance, and provides an effective solution for understanding the spectrum allocation among users.

关 键 词：认知无线电 频谱分配 强化学习 平均意见得分 价格博弈 

分 类 号：TN92[电子电信—通信与信息系统]