基于神经网络计算的无线容量高实时预测

发布时间:2020-07-31 作者:赖昱辰,钟祎,王建峰 阅读量:

 

 

基于神经网络计算的无线容量高实时预测
 
赖昱辰1, 钟祎1, 王建峰2
(1.华中科技大学,中国 武汉 430074;2. 微软公司,美国 雷德蒙德 98052 )
 
摘要:提出了一种基于卷积神经网络(CNN)计算的无线网络容量高实时预测方法。针对不同的网络部署环境,考虑路径损耗、信道衰落、墙损等因素,分别建立无线网络模型以获取数据集。将无线网络中接入点的部署方案视为二维矩阵像素图,并作为神经网络的输入,将无线网络容量标记为标签。使用CNN处理矩阵并输出数值,与标签值对比进行权重优化,再仿真验证CNN的不同架构和参数的影响。CNN可以更智能和高效地进行无线网络性能的评估与优化,实现大规模物联网(IoT)网络的部署和监管,具有高准确性和鲁棒性。
关键词:卷积神经网络;容量预测;网络部署;干扰


High Real-Time Capacity Prediction Based on Neural Network Evaluation
 
LAI Yuchen1, ZHONG Yi1, WANG Jianfeng2
(1.Huazhong University of Science and Technology, Wuhan 430074, China; 2. Microsoft Corporation, Redmond, 98052, USA)
 
Abstract: Based on convolution neural network (CNN) evaluation, a high real-time prediction method for wireless network capacity is proposed. According to diversified aspects such as path loss, channel fading and wall loss, wireless network models in different deployment environments are established to obtain data sets. Then the deployment patterns of access points are regarded as 2-dimensional matrix pixel maps, which are the inputs of the neural network, and the values of the wireless capacity are marked as labels. CNN is used to handle matrices, output numeric, compare with the label value for weight optimization, and verify the performance of CNN models with different architectures and parameters through simulation. CNN can enable more intelligent and efficient wireless network performance evaluation and optimization, realize the deployment and regulation of massive Internet of things (IoT) networks, and prove high accuracy and robustness.
Keywords: convolutional neural network; capacity prediction; network deployment; interference

 

 

在线PDF浏览: PDF