| 基于强化学习的定向无线通信网络抗干扰资源调度算法 |
| |
| 引用本文： | 谢添,高士顺,赵海涛,林沂,熊俊.基于强化学习的定向无线通信网络抗干扰资源调度算法[J].电波科学学报,2020,35(4):531-541. |
| |
| 作者姓名： | 谢添 高士顺 赵海涛 林沂 熊俊 |
| |
| 作者单位： | 国防科技大学电子科学学院, 长沙 410003 |
| |
| 摘 要： | 为了在无线网络中进行高效的链路资源调度、减小网络干扰、提高网络容量，提出了一种利用回溯天线并考虑干扰环境的链路资源分布式智能调度算法.首先，结合通信的路径损耗模型设计卷积核，对节点密度矩阵进行卷积来衡量干扰链路强度，从而避免对所有干扰链路进行信道估计产生巨大的计算代价；然后，结合强化学习的思想设计了与通信环境交互的链路调度学习模型，每个链路利用神经网络进行独立的训练，将训练所得的决策结果反馈到环境中进行状态更新，模型在不断更新的环境中迭代来学习最优的调度策略.该方法能分布式的运行，可有效衡量无线网络中的链路干扰强度，结合衡量结果进行高效的链路资源分布式调度，从而最大化网络容量.仿真结果验证了该调度算法无论是在算法迭代收敛还是网络容量性能上都能很好地逼近全局的调度算法，达到全局算法最优结果的92%~100%.
|
| 关 键 词： | 链路资源调度 定向通信 分布式智能调度 强化学习 神经网络 |
| 收稿时间： | 2020-04-13 |
An anti-jamming resource scheduling algorithm for directional wireless communication networks based on reinforcement learning |
| |
| Affiliation: | College of Electronic Science and Technology, National University of Defense Technology, Changsha 410003, China |
| |
| Abstract: | In order to schedule link resources in wireless network efficiently and reduce network interference and improve network capacity, a distributed intelligent scheduling algorithm of link resources using directional communication is proposed. First, the convolution kernel is designed by combining the path loss model of communication. The density matrix of nodes is convolved to measure the strength of interference links, thereby avoiding the huge computational cost of channel estimation for all interference links. Then the link scheduling learning model interacting with the communication environment is designed based on the idea of reinforcement learning. Each link is trained independently by using neural network, and the decision results obtained from training are fed back to the environment for status update. The model runs iteratively in a constantly updated environment to learn the optimal scheduling policy. This method can measure the link interference intensity effectively in wireless network, and distribute the link resources according to the measured results, so as to maximize the network capacity. Simulation results show that the distributed scheduling algorithm can approach the global scheduling algorithms well both in iterative con- vergence and network capacity performance, reaching 92%-100% of it. |
| |
| Keywords: | |
|
| 点击此处可从《电波科学学报》浏览原始摘要信息 |
|
点击此处可从《电波科学学报》下载全文 |
|
